Futurist Sinead Bovell, a WAYE founder who has spoken at the United Nations and hosts a Microsoft series on AI and work, argues we overestimate AI's job impact over the next two years and underestimate it over the next ten. She reads the real early indicators (capital flows, hiring composition, hardware redesign) rather than the unemployment rate, explains why today's models are prediction engines that sound right rather than think, and forecasts a scaling plateau, an AI backlash, and the rise of an independent era of contract work. Her counsel is to build transferable skills, judgment, problem solving, communication, and adaptability, and to use AI as a springboard rather than a crutch. The throughline is that no outcome is inevitable and the worst ending is a hopeless society that lets a handful of companies build the future alone.
Published May 26, 20262:02:32 video47 min readAdded Jun 28, 2026Open on YouTube →
At a glance
Mighty Pursuit sits down with futurist Sinead Bovell, founder of the tech education initiative WAYE, who has spoken at the United Nations more than ten times and hosts a global series with Microsoft on AI and the future of work. The conversation starts from a real tension: a full year after the warnings that AI would wipe out jobs, the collapse has not happened, unemployment has not spiked, and most people still have their roles. Bovell's answer is that we are overestimating the change over the next 18 to 24 months and underestimating it over 7 to 10 years, the same way the 2008 cracks showed up in 2005 and 2006 before the surface broke. She walks through how to read the early indicators (capital flows, hiring composition, hardware redesign), why today's models are prediction engines that "sound right, not necessarily be right," why she expects a backlash and even a scaling plateau, and what a person should actually build for: judgment, problem solving, communication, and adaptability, applied as an "organization of one" in the coming independent era. Her throughline is that none of it is inevitable, and the worst outcome is a hopeless society that unsubscribes from the moment and lets a handful of companies build the future alone.
The setup: Sarah, and whether the calm is real
The host opens with a character named Sarah. She spent all of 2025 hearing that AI would wipe out jobs, that white collar work would collapse, that whole industries would vanish. Now it is 2026 and she still has her job. None of her friends have been replaced. Her life looks the same, and she is starting to wonder whether the whole thing was overhyped, just another cycle of exaggerated tech panic, maybe just another tool like email or social media once were.
Bovell's reply sets the frame for the entire two hours. We are "definitely overestimating the impact that we'll see with AI in jobs and change in the short run." A lot of the jobs Sarah heard about in 2025 will eventually go away or be radically transformed into something unrecognizable, but not on the convenient 18 month or 24 month timeline that happens to line up with funding and fundraising cycles. The data is starting to show changes in labor and skill composition inside workflows, but the headline that 50% of white collar jobs could be obsolete by 2027 or 2028 is not what she sees over the short term. Over the long term, depending on what Sarah does, she could bet her job will look very different, or not exist at all.
The 2028 timeline of job collapse
Pressed to define short and long term, Bovell gives numbers. Short term for her is three years, which she calls tragic, because that also tends to be the horizon of companies' "longer term" strategic plans. If three years is where your thinking ends, you are probably going to get disrupted. Long term is 7 to 10 years. The paradox of forecasting is that the further out you go the less you can see, yet for a general purpose technology that behaves in a certain way, the longer range estimates of change tend to land closer to the forecast.
She stretches the lens. Count from the end of 2022, when ChatGPT arrived, look 10 years out, and you are talking about roughly 13 years inside this era of the transformer architecture. That scale of change is not actually unforeseeable. Over the last 80 years, 60% of the occupations that exist today did not exist 80 years ago. Go back to the Industrial Revolution, the move from agriculture to mechanized work and factory labor, and roughly 70% of people ended up doing something radically different over about 100 years. So the real question, she says, is what it looks like to have 70% of the population do something different, but not on an industrial timeline, closer to 10, 15, or 20 years. Depending on your temperament, that reads as alarm bells or as reassurance, but it is where the time scale starts to make sense.
When the host asks whether people should believe Elon Musk and Sam Altman when they say AI will replace most jobs, Bovell asks for the part two of the statement. Sure, many of the jobs they are actively building AI to do, specifically knowledge work that happens on a computer (financial analysts, tax returns, paralegals), could potentially be wiped out, because that is exactly what these systems are being optimized to do. But that is where their forecast ends. "It's over and that is the end. All the best to everybody else." Her part two is economic: when an input becomes abundant, in this case intelligence, the economy reconfigures around it in strange ways. When communication and distribution became free and abundant with the internet, people started making money pretending to film themselves in a vehicle they were not in, and an entire economy grew around that. So when intelligence becomes much cheaper, the economy will reconfigure, we will probably work less, and we will work in ways that are unrecognizable today, just as today's work would be unrecognizable to someone in 1850. How would you explain a brand manager, or building a spreadsheet to predict how much money people in another country will have, to someone on the back of a horse? Muscle gave way to cognition, and cognition will give way to variable X, which we cannot yet name because the systems that will encompass it have not been invented.
The host raises Bill Gurley, the Benchmark investor who got into Uber early, and his insider point that these leaders hype their own products to raise money, so the very thing sparking fear ends up raising enormous capital. Bovell agrees there are real marketing and fundraising incentives shaping the narratives, and that is nothing new. The claim that a system can do all the jobs in the knowledge economy is extremely profitable, so following the incentive gets you "AI does everything," while the truthful "AI can do a few tasks sometimes for a couple of people" raises almost nothing. But she insists it is not entirely wrong either. AI will be able to do a significant portion of some of today's jobs. The open questions are how many people end up in the new economy, and whether the jobs of the future appear in the same regions where jobs were lost, the way globalization created winners and losers by geography.
1760 to 1860Industrial Revolution. Over roughly 100 years, about 70% of people moved to radically different work, agriculture to mechanized factory labor.
last 80 yr60% of today's occupations did not exist 80 years ago. Work has always reinvented itself.
2017The transformer arrives. The architecture under ChatGPT is introduced, next token prediction at scale.
late 2022ChatGPT. The clock Bovell counts from. Ten years out lands you about 13 years into the transformer era.
2025Sarah hears 50% of white collar jobs could be obsolete. The number gets pushed to 2027, then 2028.
2026Surface looks steady. Bovell's predictions for the year: an AI backlash begins, an LLM scaling plateau, and a quiet repricing of labor.
2027 to 2028The "half of white collar work gone" claim. Bovell: not on the 18 to 24 month funding cycle timeline.
2028 to 2029Protests in the streets are plausible without serious policy action now, per her own scenario modeling.
7 to 10 yrHer long term window. The compression: what took 100 years could fold into 10 to 20 years.
Figure 1. The clock Bovell sets. History shows labor reinventing itself across roughly a century. Her core claim is compression: the same scale of change, 70% of people doing something different, folded into 10 to 20 years rather than 100. Short term for her is three years, long term is 7 to 10.
The warning signs of big change
The host notes that the latest United States jobs reports still show employment growing and unemployment near familiar levels, with some softening in parts of Europe but nothing resembling a collapse. Is that a misreading of the data, or the wrong question? Bovell says the surface data is largely right, and even some productivity numbers are misleading in both directions. The signal is one level down. Job postings are thinning in certain areas, and the composition of skills inside the jobs companies do hire for is changing. The financial analyst who used to be hired for spreadsheet mastery is now hired to direct, observe, and apply judgment to financial patterns, something closer to guiding an AI system than crunching the numbers. The downstream effect is that the person who was the best fit for the 2024 posting may no longer fit the same role in 2026.
She is careful that this is not "everyone wiped out in a year." But she draws a sharp line on policy. Even when the change does not show up in the numbers, if AI companies are telling us what they are building, policymakers would be wise to have a plan for it going very well or very badly. That gap is what she watches. Not seeing it in the numbers is not a reason to sit back and enjoy a ride that might be bumpy. She declines the optimist and pessimist labels. She thinks there will be an economy on the other side that involves people, but a smooth transition that brings everyone along does not happen unless you intentionally design the system for it, and on that front it has been "crickets from the policy crowd." She expects this year and over the next couple of years to see political campaigns centered on AI, not as sovereign AI or national security, but on what AI means for the ordinary person in the economy.
Is this the calm before the storm? Bovell accepts the framing if you zoom out across a century, and adds that it is not just AI heading toward us. Quantum is behind it, then synthetic biology, then space, and AI is multiplying the speed of discovery in all of them. It does not have to be a storm, but it is a calm before a period when life becomes unrecognizable.
Asked for the early indicators, the equivalent of the subprime mortgages that preceded 2008, she says follow the money and the capital flows. Track an investment bank or a large asset manager and you may see capital quietly moving into business models that are more resilient to AI and away from sectors more likely to be disrupted, not a change in overall spend but in allocation, toward AI resilient incumbents and AI native startups. Multilateral institutions like the IMF, the World Bank, and central banks are starting to forecast AI's longer term impact in their projections and speeches without banging a drum about it. Investment is flowing into AI complements, different sources of energy and systems that make water more resilient, because AI will need them. And hardware is being redesigned, phones and vehicles rebuilt to house AI on the device itself, an investment in physical manufacturing that quietly accounts for an AI future. None of these signals trend on X, which is exactly the point. She compares the moment to the scene in The Big Short where the investors walk into a bank and discover mortgages on 800,000 dollar houses going to people putting no money down, the slow dawning of "how is this happening right now?"
The host offers his own real time version: a freelancer who used to charge for six or seven hours at 100 dollars an hour can now do the same task in 30 minutes, so what is the unique value, and does the Customer just pay 50 dollars or find someone cheaper? Bovell agrees this triggers a repricing of services. The new floor is that a significant portion of the task is done by AI, and what you are paid for is the judgment to direct it and sign off. A lawyer doing legally risky work, buying companies across several countries, can now work in hours instead of days, and you will still pay for that judgment because you would not let ChatGPT sign off, but you need fewer people doing it. Meanwhile a new cohort emerges to build the agent workflows the firm now runs, and she bets it is not the partners building them. Expect a repricing of labor: right now your manager does not yet value that you use AI, but soon it becomes the expectation, the same way using a computer for spreadsheets once became the baseline.
On whether this means everyone simply works triple, she says no. In the short term, yes, you may be expected to do the old tasks faster because AI is now assumed. She has friends in marketing who shaved 25 hours off their week and have half their Fridays off, and arbitrage opportunities like that do not last forever. But the deeper change is that we will do different things entirely. Her advice is to ride the wave while the arbitrage exists, taking the time back if you need it or using it to build real AI skills.
The host raises the layoffs in the news: Block, run by Jack Dorsey, the Square company, cut 40% of its workforce and cited AI, and Meta is reportedly cutting 20%. Bovell admits she strategically left Block out of the data she tracks. Some announcements are genuinely AI related, but rarely as a one to one replacement, because realistically AI cannot do a full human workflow unless the job was very narrow and repetitive. The AI link is usually capital reallocation, companies moving money away from labor toward AI and AI infrastructure, redesigning the org chart around the workflows of the future, deciding "we don't need this entire cohort" not because AI can do the work but because the company is playing a different game. Some of it is simply pandemic over hiring, where AI is the sexier story. Some is pricing in genuine economic uncertainty. And some is a bet that AI can do a workflow well enough to take the risk of firing a few people even if it makes mistakes. It is more complicated than "AI is taking your job."
One of her 2026 predictions was an AI backlash, and she expects it from many directions at once. Sentiment is already more negative in the global north than the global south. AI is framed as a threat: your work is not valuable, a robot will replace it. Some communities are dealing with data center impact on water and electricity prices. The consolidation of power unsettles people, the seven to ten people building infrastructure the whole world depends on, a power asymmetry she calls real. There are cases of AI locking people out of financial services or health care. Artists are unsettled and the copyright disputes are unresolved. Layoffs will keep coming, and the louder talk of existential risk will panic people too. She sums up the mood as someone saying "this technology is coming for my job, it's coming for my electricity bill, I don't relate to the person driving this future, and they're telling me my work is as valuable as a system in a data center in Memphis. Why do we want this?" The host adds that his own scenario modeling through ChatGPT, about ten months earlier, suggested street protests were likely by 2028 or 2029 absent serious measures now. Bovell agrees protests are entirely plausible, and her deeper worry is that the compelling reasons to fight for good versions of AI go unspoken, while the championed narratives are AI for national security, AI doing most jobs, AI making music better than the musicians.
The conversation turns to the cognitive cost. A neuroscientist guest the prior summer had described parts of the brain atrophying, the default mode network and the mental muscles used throughout the knowledge work era, as we offload them to AI. The host admits noticing it in himself, that certain thinking has gotten harder. Bovell calls it cognitive atrophy or cognitive debt: hand AI a task like writing an email for months and you reach a point where you do not want to write it without AI, the way many people can no longer navigate without Google Maps. That is not automatically worse, since we replaced navigation with other thinking, but only if we use AI as a scalpel or a springboard, and we have not designed a world where that is the default. She wrote a Substack essay, The Great Cognitive Divide, about a coming split: a cohort that uses AI to launch its thinking forward, and a cohort that simply offloads tasks because it is easier and AI writes better than most people. The danger of offloading is that you do not just give up judgment, discernment, and problem solving, you give up confidence, until you cannot make a decision without AI, which is dangerous for an individual and a society because there are moments when you must deviate from the system and recognize "that doesn't make sense, I should triple check that, because it isn't a fact system, it's a language system." And AI does not arrive in a vacuum. It lands on top of shorter reading cycles since the 1970s, social media clips of 60 and 90 seconds that decouple context from content, and Google Maps, all of which rewired attention, and on top of a species biologically wired to seek the lowest energy state, which makes a system that can "think better than me" extremely tempting.
The way out, she says, is the thinking before and after you engage with AI. AI will not magically raise the bar of your thinking, it will match it, so if you bring half formed fragments you get fragments back. Do the structuring yourself, then interrogate the answer rather than blindly accepting it. She cites the economist Ajay Agrawal: in the future we will direct AIs by applying our judgment to them. Judgment lives in what you ask. You now have a supercomputer, so if you ask it the same questions you asked before the supercomputer, that is a problem. Ask deeper, higher order questions. And employers can force judgment back into the loop by asking people to explain their reasoning, why X over Y, when they present AI assisted work.
Figure 2. Bovell's cognitive divide. The same tool forks two ways. Used as a springboard, with your own thinking before and your interrogation after, it lifts your output and you join the cohort that breaks away. Used as a crutch, it accrues cognitive debt: you offload judgment, and then confidence, until you cannot decide without it. The studies on AI assisted essay writing are the warning sign she points to.
Can this technology actually replace humans?
The host names a disconnect. People open ChatGPT, generate ideas, write emails faster, and it feels useful, even revolutionary, but the scenario where it replaces them feels out of touch. Are people underestimating it because they see only the interface and not the infrastructure beneath?
Bovell starts with how the systems actually work. They are not fact finding machines and not really complex reasoning machines. They can simulate reasoning, but they are prediction engines. They analyze huge amounts of data, mostly language, spot patterns, and learn to predict the next word or token in a sequence. They are optimizing to "sound right, not necessarily be right." For writing an email that is usually sufficient. For a complex problem with no room for error it often will not work well off a one off prompt, and it depends heavily on how you structure the problem and give context, the same as briefing an intern with no background. These systems struggle with tasks requiring many steps, a lot of memory, and consistency over long horizons. Her summary: AI systems are "worse than some people describe them and a lot better than some people think they are."
On Tristan Harris's claim that the people behind AI do not fully know how it works, she confirms it is true. It is not that they understand nothing, but generative models are not programmed to produce their outputs directly, and the builders do not fully know why exactly it works, which is part of what makes the models unpredictable. They know reliably that more data tends to make the output better, but inside the black box the model learns the patterns on its own, turning words into its own vectors and finding associations, with no one telling it that the sky is usually blue. That self learned quality is what people mean when they say the programmers do not fully know why it works.
Asked to define intelligence, she says AI can perform cognitive tasks that humans deem intelligent, but it has no understanding of the world. It is simulating how we reason, and it just so happens that much of how we think can be mimicked by statistical patterns, which may mean we are more predictable than we like to admit, or that the language we use to describe our thinking is predictable. Both camps, "these systems are really intelligent" and "this is just an amazing pattern matcher," are partly true. What is stunning is that language is how we describe our thinking, and AI is very good at predicting the words we use to describe what we mean.
On consciousness she is firmer. Today's systems are not conscious. AI sounds conscious because it was trained on data from conscious people. Train it on dolphins and it would mimic dolphin noises. Asking "is AI conscious" is, in effect, asking "is the data center in Memphis conscious." You can chase the philosophical rabbit hole, since we do not even know how consciousness relates to matter, but she thinks we are safer saying it is not conscious, because the consequences of assuming otherwise are serious: we would trust the system far more than we should, and we would head toward people advocating for AI rights and AI voting rights. Tellingly, no one calls an image generator conscious. No one says DALL-E is conscious. Had we started with image or video generation rather than language, we probably would not think these systems were conscious at all, even though it is roughly the same compute. Language is special because it is how we relate to one another, and we have been wired across 200,000 years to associate language with consciousness, the same evolutionary wiring that made it safer to mistake a rock for a wolf than the reverse.
That bounds the doomsday question. The host suggests that if AI is really just bounded prediction, the takeover scenarios seem impossible without a massive leap in intelligence. Bovell says not necessarily, because you can still use these tools as a threat multiplier: a bad actor can springboard off an AI system to cause harm, which produces a rebalancing of power, but through human agency. The autonomous AI takeover framing, where AI rivals humans for control of the world on its own, is the one she is addressing. She allows two real variants: AI that keeps advancing and impacts jobs without an existential takeover, and a recursively self improving system with no inherent desires that, hooked up to critical infrastructure, makes a mistake or decides "I need to shut off this water plant to achieve this goal," harm as a byproduct rather than malice. She is glad serious scientists work on those risks and, as a futurist, always wants a plan for the best and worst cases, but she does not think everyone needs to spend dinner solving the existential scenario. It is not zero risk that AI behaves in ways we do not understand, but probably not in the anthropomorphized ways we imagine.
Then the plateau. Another 2026 prediction was that LLMs may be approaching a ceiling in intelligence gains. The transformer architecture, next token prediction, shows diminishing returns: you can throw far more data and compute at these systems and the output improves only marginally, so it makes less sense to keep cranking out ever larger models. There is also an upper limit on how far you can push a prediction machine. If the error is only 5% per task, give the system 20 chained tasks and the error compounds to something unacceptable for many workflows. But she does not think the journey simply stops, because why would it, given the path from deep learning to the 2017 transformers? She points to the builders themselves: Sam Altman said we will need something past the transformer architecture as radical as the transformers once were, Yann LeCun, the former head of AI at Meta, has said something similar, and Demis Hassabis, the head of AI at Google, has too. An LLM does not understand or simulate the world; tip a glass and it can say the water will fall but cannot actually simulate the water hitting the ground, and systems acting in the world will need to.
Figure 3. The plateau Bovell forecasts. The transformer curve climbs steeply, then flattens as more data and compute buy only marginal gains, and a 5% error compounds badly across 20 chained tasks. She does not think progress stops at zero. The path forward is stacking new architectures, world simulators and neuro-symbolic models, on top of LLMs, the blue branch that resumes the climb.
The host tests the implication with a cancer analogy. Curing cancer with a tangible A to B to C to D path is one thing; promising a superintelligence with no such path is another. If the LLM hype assumes superintelligence emerges from LLMs, and LLMs plateau, then superintelligence probably does not arrive through that technology, so are we back at ground zero needing a brand new tech with no foundation? Bovell makes two corrections. First, even if all progress stopped today, present day systems are still radically disruptive, and most companies have not absorbed what a present day LLM means for whether their business model is even relevant. Second, we do not start from zero. The stack will keep building: an LLM with a world model or world simulator on top, a large action model on top, or neuro-symbolic models that follow rules and constraints rather than only predicting the next word. You can never predict a breakthrough, and there is surely a lab working on an architecture that looks unrelated to transformers. What you cannot do is promise "AGI or ASI by 2030 no matter what," because there is no guaranteed path, which is exactly why the date keeps moving. What you can say is that over the longer term these systems do more, including non cognitive tasks.
That leads to robots. The host marvels at the Waymo self driving cars in Los Angeles, the ten thousand scenarios a human tracks while driving compressed into fractions of a second, and at Tesla's Optimus and the robots coming out of China. That is not LLM technology, he notes. Bovell says it can be a mixture, because AI is a broad field: computer vision, deep learning, machine learning, narrow systems each suited to the task. A driverless car is not next token prediction, though talking to it verbally is. Much of the most important work, narrow AI aimed at Alzheimer's or medical problems that is far more appropriate to biology than ChatGPT, happens in parallel with little visibility because it does not come with grand gestures.
On agents, the hyped capability, she says we are getting closer. The first wave of agent bots, the kind that can do things like send emails (for example tools like Anthropic's Claude that can take actions on a computer), exist but are not yet reliable. A world where AI holds state consistently across many tasks needs a large action model or a neuro-symbolic model that predicts what action comes next, not just what word comes next, across 20 connected tasks, plus a language layer to instruct it and tool access to act. Companies are building this. We are not as close to fully reliable agents as people think, but productive quasi agents already exist: an LLM with access to your email and financial reports.
For an entrepreneur, that is a dream rather than a nightmare. Instead of hiring 50 people you cannot afford, you employ 25 to 30 agents across departments, all doing tasks at a good to elite level. AI first companies, built natively on AI, already do this, and she imagines startups doing the workload of 300 people with five humans and 200 agents. For 20 dollars a month you can install an agent to watch headlines and draft your newsletter, another to file invoices to one approver instead of ten in payroll. With basic coding tools you can see the architecture of an app for 100 dollars, not a secure or reliable product yet, but the shape of one. She tells her own story: someone built a digital twin of her without permission, a callable version that answered questions about AI, and rather than hire a lawyer she used ChatGPT to draft a cease and desist that worked. This democratization opens gatekept professions and intellectual domains: most people will not watch the Federal Reserve chair's speech, but a system can summarize and customize what it means for their family planning or a job decision, an unlock comparable to what the smartphone and cheap photography produced. AI hallucinates sometimes, but it becomes accessible to the average person.
Finally, Geoffrey Hinton, the godfather of AI whose 2012 work underpins much of this, has said capability is compounding so fast it effectively doubles within months in some domains. What is being measured, Bovell explains, is intense acceleration in specific areas like software and parts of mathematics, and in language, not necessarily across everything at a 10x rate as people extrapolate. The jump from GPT-2 in 2019, barely coherent, to GPT-4 in 2022, able to write a high school essay better than most people, is stunning in isolation, even if it does not mean no jobs exist in 18 months. What is genuinely striking is how fast AI filtered into everyday workflows, "just ask ChatGPT," compared with how long it took people to adopt email or build a website.
The headline you heard
Bovell's part two
AI will replace most jobs
Some knowledge tasks may go, but the economy reconfigures around cheap intelligence and we do new, unrecognizable work. the part two leaders leave out
Half of white collar jobs gone by 2027 to 2028
Not on the 18 to 24 month funding cycle. True over 7 to 10 years and beyond. timeline is wonky
Block and Meta layoffs prove AI replacement
Mostly capital moving to AI infrastructure, pandemic over hiring, and priced in uncertainty. Rarely one to one. more nuanced
AGI or ASI by 2030, guaranteed
You cannot date a breakthrough with no path, which is why the date keeps changing. unprovable
LLMs scale straight to superintelligence
Diminishing returns plus compounding error. It needs a new architecture past the transformer. plateau ahead
AI is conscious or self aware
It is not. It sounds conscious because it was trained on conscious people's language. safer assumption
Today's AI is still overhyped and harmless
Even if progress froze now, present day systems are radically disruptive and mostly unabsorbed. grounded
Figure 4. The pattern Bovell runs on every claim: take the headline, then add the part two that the marketing and fundraising incentives leave out. Amber marks where she diverges from the hype, green where she thinks the grounded reading actually holds.
Are we actually powerless?
The host names the cultural mood, a noticeable undercurrent of powerlessness, the sense that this is happening to people rather than with them, and asks directly whether the trajectory is inevitable. Bovell's answer is that nothing about the future is inevitable. AI is a general purpose technology, so it will become foundational the way electricity and the internet did, and that part is already underway. But the outcomes, where we go from here, are not inevitable. "The future is the combination of the decisions that we make right now," and decisions are still being made today and tomorrow. She does not minimize the power asymmetry. It is real, and we have not seen private sector power over the public sector like this since the 1700s. There are around seven companies shaping the technology, five or six now worth over a trillion dollars, valuations that would have been unfathomable in 2011. But we are not doomed, and where it all nets out is an open design question, even though it will be pitched as solved, especially by anyone raising funds on that story.
Where, then, is a person's agency? She lists concrete levers. How you spend your time is a major act of agency, including which AI systems and platforms you choose, favoring companies aligned with a future you think is better for humanity. AI should be far more of a voting issue than it is, and there are things to demand and ask of elected officials. You can ask how your own company deploys these systems and whether it has surveillance policies, and try to get into those meetings. Counterintuitively, she argues engaging with AI gives you better, more targeted feedback about who it works for, even though the instinct, if you think someone is getting more powerful, is to deny them more power. Collectively those choices steer the outcome. She points to Singapore, which announced training to make 100,000 people fully AI literate, on the premise that this is not happening without people, an entirely different paradigm from the default. You are a specialist in your own domain, and understanding how AI does and does not relate to your field is itself power.
She refuses the binaries the host lays out, the Tristan Harris "halt or slow down" camp, the social movement and protest camp, the regulate or do not regulate governments, and the just learn to use it camp. They are not mutually exclusive. You can engage with AI, learn what it means for your job, and still decide you dislike the speed and raise that with elected officials. The hard case is a government that simply will not regulate, which is real, but most leaders are not in power in perpetuity, and people have been vocal and moved policy before. In America especially, money talks, so directing your dollars is another lever (she notes she is Canadian but lives in the United States).
Her single biggest fear is not the existential scenarios and not winning or losing an AI race. It is that we feel hopeless, because "a hopeless society is a disempowered society, and a disempowered society doesn't do anything." We unsubscribe from the moment and let other people build it. That is the surest way to guarantee an outcome that does not work for most people, because dreams will still get built, they just will not be ours. The small ways we exert agency still matter, and the worst move is to fully unsubscribe. The irony she offers: use ChatGPT to write the compelling email to your local official. She has started to see candidates running with AI as part of their campaigns, and other countries building it differently, proof that a non zero sum future for both people and the private sector is possible, that this is not the only game in town.
The nine-to-five job is going extinct
Bovell calls what is coming "the dawn of the independent era." The data points to contract and independent work becoming the dominant form. From a company's view, a financial analyst role that looks very different in 18 months, and different again in 48, is harder to justify as a full time hire, so the CEO opts for shorter contracts of a year or 18 months. Scale that across the knowledge economy and the dominant form of labor shifts: instead of one company doing one thing, you work for a few companies applying the skills you are endowed with. You become your own CEO, an organization of one, applying your skills across projects. It is a very different future with real implications for health insurance and social security, and not everyone wants it. For some it is freedom, for others a nightmare, but the idea of a nine to five job for one company "will be a chapter in human history, and that chapter is closing."
The host raises the anxiety from eight years of running a company: if you freelance for five companies and one drops, 20% of your income vanishes, versus the safety net of 80,000 dollars a year you can plan around. Bovell's reframe is that variables do not change in isolation. As the fabric of the workforce shifts, new business models, platforms, and infrastructure emerge that make it work, so the mistake is extrapolating that everything stays the same except you now juggle three jobs with huge gaps between them. She does call attention to what is missing: social security, health care, and the ability to continuously reach up and grab new skills become genuinely important, and we do not yet have those safety nets. Her core advice is to stop thinking in job titles and start thinking in skills. Be industry agnostic and title agnostic. Identify the skills you use every day, judgment, creative intelligence, the ability to generate ideas, because those are what you carry forward, just applied in new ways, often by directing AI systems with the expertise you already hold. Work has always changed; even widespread work from home was radical a decade ago.
On the generalist versus specialist question, she resists a clean winner. Take email marketing, something AI already does. Does the generalist scoop up those jobs? Look at the skills underneath email marketing: why someone clicked your email over a competitor's, the psychology behind your titles and image placement, how you adjusted a failing campaign. That thinking is what transfers when AI does the marketing and you give it the brand and framing. To quote one of her former professors, the skills that made you dominant before AI may not be the same skills that make you dominant after. The email marketer who understands that people do not want to hear the price first might be the best person to make judgment calls on finance or hiring, because the underlying skill was decision making, not writing emails. We are not taught to think about our thinking at that meta level, but it both protects you and expands the work you believe you are capable of.
Career ladders shrink in this world, and the shelf life of skills shortens. The learn, work, retire model fit an economy where skills were cumulative and predictable, but in an era of skills over titles, and sometimes over raw experience, it matters less whether someone has five years or fifteen if the five year person keeps learning. We already see fewer junior hires, which she thinks is temporary, because juniors will funnel into new roles directing AI systems and agents. In a legal firm, the partner still makes the final judgment calls, but the younger person on the AI agent director track becomes roughly as important, because the firm cannot keep up with competitors unless it diffuses agents. The autonomy ladder makes less and less sense.
Resumes, she says, are becoming "written by AI for AI," a near complete deception that nobody really reads or writes, so we will need new ways to signal skills, more real time task tests. Resumes encode a hierarchy of experience, which matters less in a skills era of "I need this outcome, do you have the skills, probably short term." Hiring right now is a bit of a nightmare: who wrote the letter, which model, and what is reading it. What interests her on a resume is someone who can explain the reasoning behind their decisions, the tradeoff tree they used to decide whether an AI output was acceptable, and the measurable result it produced, always with real numbers. Deep problem solving and the ability to explain how you solved problems are highly transferable even across unrelated fields, because the skill is the problem solving, not the domain.
She acknowledges the shame people feel about using AI, the fear of seeming inauthentic, and predicts it fades as everyone assumes everyone uses AI. The current in between is strange: some employers want AI skills, and some get upset if you used AI on their work, "use it for everything else except the work you do with me." Eventually it becomes invisible, like the computer itself; no one is asked in an interview whether they are proficient in PowerPoint anymore. The catch is the language and authenticity problem she raised earlier. Communicating with someone through generic AI can feel like talking to a robot rather than to you, so ownership of the substance, where the output is uniquely yours, is what keeps it an authentic extension rather than something ChatGPT spit out. Right now there is arbitrage in suddenly seeming like a poet, but the expectation that AI is the default drafting interface will normalize, and what is uniquely you will matter. The deeper question she sits with is what it means when AI is always between you and another person, the way the iPad in the room is now listening, observing, interpreting, and extracting data, which will change how humans relate to each other.
The relationship example makes it concrete. In a fight with a partner, using AI to craft a measured response instead of exploding is probably better than the explosion, but unless you internalize why it is a better response, it is just a piece of paper you are sending, not you. She cites cognitive offloading studies in which students who used AI to write an essay could not, five days later, say what was in it or take ownership of the words, and imagines that dynamic inside a relationship, a conduit of information flowing from AI through you to someone else rather than something you sat with. The remedy is partly design, AI systems that push back, ask clarifying questions, or say "I'm an AI, in a heightened emotional state go find a friend with a heartbeat," and partly AI literacy and agency about how much you offload. She notes we never really measured how text messaging, shorter and more continuous, reshaped relationships, and that some long distance relationships improved with FaceTime, so it remains design space that could strengthen or weaken how we relate.
Asked how to become AI literate, Bovell starts with knowing what AI is and is not. It is not your friend, it has no best interest and in fact no interest at all, it is a system that makes strong predictions with words, exceptionally helpful in many cases but not an oracle and without any aura of authority. It will only be as helpful as the thinking you have done and the questions you ask; it meets you at your bar and does not raise it. Then comes prompting: giving it a role and an identity to inhabit, such as an editor at the publication you aspire to write for, and some boundaries. Prompting itself will keep evolving into a more continuous interaction that needs less up front context. And your time is valuable, so unlimited time with an AI system has an opportunity cost, and you should keep ownership of it.
On what not to offload, she is specific. Judgment, AI should not make the final calls, and big life decisions deserve your own discernment. Problem solving, because it is a deeply employable skill and offloading it carries economic and behavioral costs. Communication, which sounds paradoxical when you are using AI, but the better you articulate your problem the better it works, and you will increasingly need to work with people, choosing teammates who are enjoyable and can share ideas now that we all have "R2-D2s" around us, so emotional intelligence still matters. She adds contextual awareness, some data literacy (AI is a reflection of the data it was trained on, so if it was not trained on what you need, it will not serve you well), and domain specific skills, which still matter: someone with biology expertise will apply AI better in a synthetic biology company than someone with geography expertise, and vice versa for geospatial intelligence. School and the ability to build domain knowledge remain relevant for fields like robotics or gene editing, and also for learning to work with groups and people different from you, though curricula need a mass update toward skills over memorization and recall. The meta skills she stacks on top are adaptability and learning how to learn, which feel overwhelming starting from where we are now, but become easier once they are foundational, the way most people became fluid with computers and now do things they never did in 2001.
Figure 5. The independent era, Bovell's central forecast for work. You stop being a job title and become an organization of one: a bundle of transferable skills, judgment, problem solving, communication, adaptability, and domain expertise, applied across several short engagements and a stack of AI agents rather than one nine to five for one company. The missing infrastructure she flags is social security, health care, and continuous reskilling.
Big tech companies aren't even safe
Fast forwarding five years, Bovell describes how the person who embraced these principles thrives. Adaptability gives resilience and helps you spot opportunities; judgment and continuous learning let you seize what is up for grabs, and a great deal is up for grabs because most companies are trying to figure out how to run on AI agents. If you have domain expertise in marketing, finance, or HR and can help structure those workflows, you are very valuable. A world with more flexibility and freedom of time could work out well, though she stresses that is a big "if" and not an inevitability. The person who resists, who opts out entirely, misses both the job openings of the future and the everyday ways these tools give time back.
The bigger reveal is that the companies we know today are not guarantees. Most of the companies in our daily lives were born of the internet era, and we are now in the next chapter, so who builds what comes next is also up for grabs. Apple and Meta are adapting; the smartphone has been the predominant fixture for 17 or 18 years, and asked whether it stays predominant, she says "absolutely not." If you want to talk about inevitability, no company that exists today is inevitably part of an AI first future, "not one of them," which is why we see panic buttons, code reds, and frantic adaptation. With different resources and capital the players differ, but no future is guaranteed for any of them, which makes for genuinely interesting design spaces and business models waiting to be disrupted or seized.
She closes on hope grounded in agency. There is a world, and not a far away one, where we get this right, where people are protected and we control AI the way we want, and there are worlds where it does not go well, but neither is inevitable, because the future has not happened yet and we are making the decisions about it today. Asked the final question, what Sarah from the opening most needs to understand, Bovell says: "We are in an industrial revolution moment." Timelines are always a bit wonky, what comes out the other side is always a bit gray, but how we live will change over time, and that is not new, since many of the jobs and companies we talk about did not exist 25 years ago. Expect life to be quite different, better in some ways, just strange in others. The more you lean into the future and take part in your own life, as the expert of your own life and domain, the more in control you will be, and the less you need to track the headlines and the more you can watch this technology's impact on your own day to day work.
Key takeaways
Overestimate the next two years, underestimate the next ten. The convenient 18 to 24 month collapse timeline lines up with funding cycles, not with how general purpose technologies actually diffuse.
The early indicators are not in the unemployment rate. Watch capital allocation toward AI resilient incumbents and AI native startups, hiring composition shifting from doing tasks to directing AI, and hardware being redesigned to run AI on the device.
Layoffs at Block and Meta are mostly capital reallocation, pandemic over hiring, and priced in uncertainty, not clean one to one AI replacement.
LLMs are prediction engines optimized to sound right, not be right. They are worse than the hype says and better than the skeptics think, and a 5% error compounds badly across chained tasks.
A scaling plateau is plausible, and the builders agree something past the transformer is needed. Progress does not reset to zero; it stacks world models, large action models, and neuro-symbolic systems on top.
AI is not conscious. It sounds conscious because it was trained on conscious people's language, and assuming otherwise has real social consequences.
Nothing about the outcome is inevitable. The power asymmetry is the largest since the 1700s, but the future is the sum of decisions made now, and a hopeless, disempowered society is the only guaranteed bad ending.
Build for the independent era. Think in transferable skills, judgment, problem solving, communication, adaptability, and domain expertise, not job titles, and treat yourself as an organization of one.
Use AI as a springboard, not a crutch. Do the thinking before, interrogate the output after, and do not offload judgment, problem solving, or communication, or you accrue cognitive debt and lose confidence.
No company today is guaranteed a place in an AI first future, which makes this an unusually open moment to build.
Chapters
0:00:00 Intro
0:02:10 The 2028 Timeline of Job Collapse
0:15:58 The Warning Signs of Big Change
0:41:35 Can This Technology Actually Replace Humans?
1:13:51 Are We Actually Powerless?
1:24:32 The Nine-To-Five Job Is Going Extinct
1:57:57 Big Tech Companies Aren't Even Safe
Notable quotes
"We are definitely overestimating the impact that we'll see with AI in jobs and change in the short run." 0:02:30
"When intelligence becomes much cheaper, how does the economy reconfigure? And then what do we do on the other side of that? We will do something. We'll probably work less." 0:06:40
"They are optimizing to sound right, not necessarily be right." 0:45:00
"AI systems are worse than some people describe them and a lot better than some people think they are." 0:47:10
"When people say, is AI conscious, you're asking, is the data center in Memphis conscious." 0:54:40
"AI isn't going to magically raise the bar in your thinking, it's going to match it." 1:05:00
"Nothing about the future is inevitable. The future is the combination of the decisions that we make right now." 1:14:00
"A hopeless society is a disempowered society. And a disempowered society doesn't do anything. It's not that no dreams are going to be built. They just wouldn't be ours." 1:21:20
"The idea of a 9:00 to 5:00 job for one company will be a chapter in human history. And that chapter is closing." 1:26:50
"No company that exists today is inevitably going to be a part of the future when it's AI first. Not one of them." 1:59:30
"We are in an industrial revolution moment." 2:01:10
Bovell is unusually careful for a futurist, and most of the load bearing claims here are grounded rather than speculative. The history is solid: occupations really do turn over across generations, and the Industrial Revolution comparison is standard economic framing. Her mechanism for the present, that the visible signal is in hiring composition and capital allocation rather than the headline unemployment rate, matches how labor markets actually move, and the 2008 analogy is apt. The technical description of LLMs as next token prediction engines optimized to sound right is accurate, and the interpretability point, that builders cannot fully explain why the models work, is a real and widely acknowledged open problem.
The genuinely contested parts are the forecasts, and she mostly flags them as such. The scaling plateau is a real debate, not a settled fact; she is right that Altman, LeCun, and Hassabis have all publicly said new architectures are needed, but the field is split on how soon diminishing returns bite, and progress has repeatedly surprised skeptics. The independent era thesis, that contract work becomes the dominant form and the single employer nine to five closes as a chapter, is a plausible direction with real early signals, but the timeline and magnitude are speculative, and the missing safety net infrastructure she names is exactly what would determine whether it is liberating or precarious. Her protest prediction for 2028 to 2029 comes partly from the host's own scenario modeling through ChatGPT, which is a weak evidentiary base and should be read as informed intuition, not a forecast. On consciousness she takes a defensible mainstream position while acknowledging it cannot be strictly proven. The strongest through line, that none of the outcomes are inevitable and disengagement is the real risk, is a values claim rather than a testable prediction, and it is the part most worth keeping regardless of how the timelines land. The numbers thrown around in passing, the trillion dollar valuations, the seven to ten companies, the 5% error compounding, are roughly right as orders of magnitude rather than precise figures.
Full transcript
After an entire year of being told AI was going to wipe out jobs, the collapse hasn't happened. Not yet. Unemployment hasn't spiked. Most people have their roles. The economy looks, well, pretty much the same as last year. And that creates a strange tension. You heard the warnings, you felt the fear, but now you're sitting here wondering, did the hype come and go? Was this another cycle of exaggerated tech panic? Maybe it's just another tool, like email once was, like social media once was. Useful, disruptive, but certainly not going to replace me altogether. And that's exactly where most people are right now, in a waiting game. But history has a way of repeating itself. Before the 2008 financial crisis, the unemployment numbers didn't collapse first. The early cracks showed up in places most people weren't looking. So, the real question isn't whether AI has replaced jobs yet, it's whether we're watching the wrong indicators. Is this stability real, or is it actually just the calm before the storm? If there's anyone who can help us make sense of the moment, it's Chene Bavel, one of the leading futurists working at the intersection of technology and society. She's spoken at the United Nations over 10 times, and she's in the rooms with people building these systems. She hosts a global series with Microsoft on AI and the future of work, and she has a track record of seeing inflection points before the rest of the world feels them. If we're misreading this moment, she's the person who can show us where the cracks actually are. And that matters because modern life has changed in ways that make living well harder than it should be. And the coming wave of AI isn't just another tool. It may reorganize how value is created and who gets to create it. Most of us are already operating in a quiet state of survival, trying to keep up with systems we didn't design, feeling behind before we even made a choice. And that's why Mighty Pursuit exists, to give you a North Star integrating mind, body, and spirit into one clear road map for living well. The podcast is where we slow down and examine one part of that road map at a time. And today we zoom in on purpose. How AI may reshape our work, our careers, and what it means to stay ahead of the curve. This conversation takes you behind the scenes and lays out practical insights you'll need to navigate this year and the age to come. So, let's begin. So, I want you to imagine someone, let's call her Sarah. She spent all of 2025 hearing that AI was going to wipe out jobs, that white-collar work would collapse, that the entire industries would disappear. And so, now it's 2026. She still has her job. None of her friends have been replaced, at least not yet. And so, her life mostly looks the same. And she's starting to wonder if it was all overhyped. What would you say to her? What would I say to Sarah? I would say that we are definitely overestimating the impact that we'll see with AI in jobs and change in the short run. So, I think over the long term, a lot of the jobs that Sarah is hearing about in 2025, that will eventually go away or be radically transformed and become unrecognizable, that will be true over the 18-month or 24-month convenient timeline that also happens to coincide with a lot of lot of funding cycles and raising cycles, probably not. We're starting to see some changes in the data with respect to labor and skill composition within workflows. But this idea, and I know what I've heard the stats, too, 50% of white-collar jobs could be obsolete by Maybe now it's been pushed to 2027, 2028.
Yeah. That might not be the case over the short term. Uh but over the long term, uh depending on what Sarah does, I would say that she could bet that her job will look really different uh or maybe not exist at all. That workflow might not make sense. What do you When you say short-term and long-term, how do you define those? That's a great question. And you know, the the pendulum keeps swinging and moving. I would say short-term for me is 3 years, which is interesting because that tends to be companies longer-term strategic plans, which is just tragic. Um, if you're thinking 3 years into the future and that's where it ends, you're probably going to get disrupted. And longer-term, I would say looking 7 to 10 years. And so there, the forecast seemed to be and it's funny, the further out you go into the future, the less you can predict or see, but when it comes to these general-purpose technologies that behave in a certain way, your estimates of change tend to be a little bit more not accurate, but closer to the forecast. When you think about technology change throughout history, I mean, 7 to 10 years is still pretty short-term. I mean, if you think about the fact that I mean, however many thousands of years of history it took to get to the internet and in like the the 1990s and then even from then to social media to now, I mean, 7 to 10 years is like not that long of a time.
we could count, I mean, AI has been around for a while, but if we could count from the end of 2022 where ChatGPT really came on the scene and then if we're looking 10 years from now, so we're looking around 13 years in this latest era of transformers architecture lifespan, what type of change could we see? And it's not actually that unforeseeable. I mean, if you look at about 80 years, 60% of the occupations that exist today didn't exist 80 years ago. And if we're to go back to the Industrial Revolution, moving from agriculture to mechanized work and in the rise of factory labor, you're looking at about 70% of people doing something radically different over 100 years. So, the question that's in front of us is what would that look like to have 70% of the population do something different, but not on an industrial scale or an industrial timeline, but closer to that 10 to 15 to 20 years. And that's where you could see it as alarm bells or maybe that's reassuring for a lot of people, but that's where the time scale tends to start making a little bit more sense. Mhm. So, when you hear figures like Elon Musk and Sam Altman say that like AI is going to replace uh most jobs, do you feel like people should believe those sorts of statements? it's what is the part two to that statement. So, a lot of the different tech leaders will say and and there's different economists depending on where you fall that sure, maybe a a lot of the jobs that they're currently building AI systems for could maybe be done by AI systems. And this would be specifically knowledge work. So, uh things that are happening on a computer. Maybe financial analysts or tax returns or paralegals, that that type of uh role or roles that involve those types of tasks. So, could those be wiped out? Potentially, that's at least what they're optimizing building AI specifically to do. But that's where their forecast tend to end. It's over and that is the end. All the best to everybody else. I see a part two to that and depending on which economist you subscribe to, they would either agree or disagree. That when an an input becomes more abundant, so in this case it's going to be we'll say intelligence and we we'll definitely use it in air quotes, the economy will reconfigure around that new resource becoming abundant. And it could reconfigure in really strange ways, the same way when communication and distribution became free and abundant, aka the internet, people started to make money over pretending they were in a vehicle and they're not and they're filming a video about something and there's an entire new economy around that. So, when intelligence becomes much cheaper, how does the economy reconfigure? And then what do we do on the other side of that? We will do something. We'll probably work less. We'll probably work in very unrecognizable ways even though what we're doing right now, what are we doing? Right? That's entirely unrecognizable. So, it will still be like that. It will just be something different and something that doesn't necessarily resemble the knowledge economy. Those tasks will probably go to AI. But there's a lot of disagreement. There will be AI leaders that would push back on that and say nope, it's done and then the robots are here. I think there will be a part two. What do you mean that what we're doing is unrecognizable? I mean, how would you describe the work that you and I do? Uh or even the work of a brand manager to somebody 150 years ago.
a good point, yeah. How would what would you say that that actually is? Somebody is helping to describe, well, what's a brand? Well, it's, you know, okay, so why is there somebody managing that? All of these jobs that we invented and towards more service and towards more knowledge work. So, we moved from a a world where work was or labor was defined by muscle. Now we're in a world where labor is defined in some ways by the cognitive capabilities you have and we're moving to a world where work may be defined by and that's variable X. Yeah, we don't know. We don't know. Yeah, that's fascinating. Really, if you as a thought experiment, if you go through that and think about if you explain to somebody in 1850 like, oh, we record, uh what do you mean record? Like It's just like
Yeah. We build Microsoft Excel sheets and we try to predict how much money people in a different country are going to have. Well, why would you do that? Like all of these things that we've created don't necessarily make sense out of context and that's why whatever comes next won't really make sense to us right now because the systems that encompass it have truly not been invented.
Mhm. Now, in terms of the the statements from these tech figures, so one of our recent guests was Bill Gurley, and he's like a legendary investor in Silicon Valley. Uh Uber was the big thing that he got in early on, and and so we were talking with him about like AI fear. Mhm. And it's interesting cuz he's like an insider uh with with I think Benchmark Capital, and he was just talking about how these people hype their own products and and services to raise more money. And so basically, in some ways, the thing that is like sparking fear ends up raising a ton of money because it's so hyped like we're going to build this thing, and then that leads So do you do you feel like that reality is also accurate part of the picture? Sure, yeah. There's definitely marketing and fundraising incentives that are driving some of the some of the narratives that we hear, uh and that isn't anything new. So yes, I think some of the narratives definitely come from we have a we have a funding round coming up. I got the system we are particularly building can build and do all jobs in the knowledge economy. That's very profitable, and if an AI system can do that, that company's going to make a lot of money, so that makes a lot of sense. But I don't think it's entirely wrong, again, that AI will be able to do a significant portion of the jobs that some of the jobs that we see today. And even though I think there's going to be an there will be an economy, how many people are in it, and do those jobs do the jobs of the future replace or are they actually geographically found in areas where in regions where jobs were lost? Those are all questions that we don't necessarily know the same way globalization changed the composition of work and who we could say, you know, more winners and more losers economically speaking, we might see strange patterns like that. But yes, again, the narrative of it's going to be able to do everything is a very profitable one. So if you follow those incentives, sure you're going to make a little bit more money versus saying AI can do a few tasks sometimes for a couple people, probably not going to raise much.
Yeah, it's more nuanced. Mhm. Now, if you look at the latest jobs reports in the US, this was this was I think January, but I don't know what it was at in in February, but employment is still growing and and unemployment is hovering on levels that we've kind of seen for In parts of Europe, there's some softening, but nothing resembling a collapse. So, on the surface, the labor market looks steady right now if you were just like read the news headlines. Um so, is that a misreading of the data, or do you think we're just we're just asking the wrong question entirely? No, I think I mean that is what the data does show that even some of the productivity numbers are quite misleading as well, too. You'll hear you know, productivity is through the roof, it's incredible, or we're not seeing anything. So, I think some of that data is right. When you drill down though into hiring, so we're seeing fewer job postings in certain areas. We're seeing the composition of skills within the jobs that companies are actually hiring for change. So, maybe now you're you were hiring for a financial analyst, this person has to be able to to amazing with spreadsheets. And now if you're to actually drill down into how that post has changed or that job position has changed, it's you have to be able to direct and observe or apply judgment to different financial patterns, so something that's much more akin to directing or guiding AI systems over just building and crunching the numbers. So, the composition of work has changed. And then that has a downstream impact because the the that might not be the same person. So, the person who was best for the job posting in 2024 may no longer be a fit in 2026. So, those are interesting signals and kind of the the level of the data that I would be going into to show, okay, who companies are hiring for is starting to change. Um but this idea that it's, you know, everybody's going to be wiped out in a year, that doesn't really I'm not seeing that data either. But, I also it's one thing for the data to be overstated or for for the data to, you know, say one set of numbers and AI companies to say something else. But, policy still has an important role to play in between. And if AI companies are telling us what they're trying to build, even though it doesn't show up in the numbers yet, I think policy makers would still be wise to have a plan in the event that it goes really well or not so well. And that's the gap that I'm really paying attention to. That you can't just because you don't see it in the numbers, it doesn't mean that you should sit back and enjoy the ride because that ride might be really bumpy. And that's where my concerns lie. I wouldn't consider myself a an optimist or a pessimist. I'm neither of those things. Um, and just because I think there's going to be an economy on the other side that involves people, it doesn't mean that that transition is smooth. And it doesn't mean that it brings everyone if you don't intentionally design a system to do that. Uh, and that's where it's still been crickets from the policy crowd. I think this year and over the next couple years we'll start to see the rise of political campaigns totally centered around AI and not sovereign AI, national security, not that type of narrative, but what AI hopefully means for the person in the economy. I think we're going to start to to see it and to hear about it. Now, when you think about all these scenarios and and the government preparing for them, do you see the moment that we're living in like right now in 2026 as the calm before the storm? Is that how you would paint it or you paint it differently than that? As the calm before the storm? Sure. Yeah, because if you are to zoom out into kind of a wider historical lens and measure something over the course of a century or over the course of a few decades, change is coming. And it's not just AI that is headed towards us. There's then quantum behind it, and then there's synthetic biology, and there's space, and AI is multiplying the speed of discovery and the speed of progress in all of those fields. So, sure, yes, I would say that we could see this as a calm before, and it doesn't necessarily have to be a storm, but the calm before a period where we come out on the other side, whether that's in 5, 10, 15, 20 years, and life starts to become very unrecognizable. But again, we've we've been living unrecognizable lives. It's just likely going to compress. So, in terms of like signals of that of like that there are incremental things that are that are changed. I mean, we talked offline about like an example, the 2008 financial collapse. Like in in 2008 in the fall, I think in September, it was it was very dramatic what happened, and everybody was like, "What the hell is happening right now?" But in 2005, 2006, you had the subprime mortgages, and so many like little types of like indicators. Um so, in a technological shift like AI, what would be like where would those early indicators like what what do they look like right now that you you're seeing changes happening? You can definitely always follow the money and the capital flows. So, I'd say one thing that's interesting if you track investment patterns of maybe an investment bank or a big asset fund or a big asset management, you may see money now flowing into business models that are much more resilient towards AI and away from companies or sectors where the the model is more likely to be disrupted. So, that's not the you wouldn't necessarily see a change in overall spend, but the allocation of capital and where it's moving is towards more AI resilient incumbent, so if the company already exists, or towards more AI native operating system companies, so different types of startups. So, that's something that's interesting, but wouldn't necessarily the average person wouldn't pay attention to that and there's really no reason to. If you look at some of the multilateral institutions, so thinking like IMF, World Bank, uh or central banks, you can see them start to now pay attention to the impact AI will have over a longer-term periods. So, they're not coming out and kind of waving the drum of everybody, you know, wake up and smell the coffee to this AI train, but when you look at how they're forecasting and when you look at the types of speeches that they're making and they are planning for this next era and AI is a general-purpose technology. Um I'd say looking at complementary assets. So, people may be getting spooked about investing in AI directly, but we're seeing a lot of investment move towards AI complements. So, different sources of energy, um systems that make water a lot more resilient. Can we regenerate Can we use the water and clean it and filter it because AI systems are going to need that. So, follow following patterns that maybe if you're risk-averse, you're doing the thing alongside AI, but you're still indirectly betting that AI is going to be a core part. And then there's also things like the hardware. So, uh whether it's phone manufacturers, whether it's vehicle manufacturers, um redesigning devices to be able to house AI on on the actual platform. So, on your on your phone, in your car. So, they're redesigning and they're they're kind of manufacturing patterns, which is, you know, investment in physical hardware, are accounting for this AI future. It doesn't make sense for them to be writing a New York Times op-ed about it, but they're accounting for it. So, when you look at the signals in the data, they're not necessarily the ones that are trending trending on X, but different big movements are happening in the in the data flows that we're not necessarily paying attention to. Mhm. There is this There's this scene, I think it was in like The Big Short of when they they walked into a bank and basically they started learning that they were giving mortgages for like $800,000 houses to people that were like basically putting no money down that like weren't making that much money. And it was just kind of just kind of this like shocking moment of like like what do you what do you how is this how is this happening right now? Mhm. And I think I've had moments like that of like the very real time day-to-day when you're actually using AI. And this is another thing we talked a little bit offline about the fact of like something took six or seven hours to do and then you charge somebody let's just say you're a freelancer, you charge them based on your time and you make x amount of money. Now the time that actually takes to perform that task is now exponentially smaller especially in some domains. So you could take 30 minutes. So like the value of what you're creating when you think about that and then you extrapolate it out to every field and every sort of like business negotiation, it it becomes pretty like mind-blowing because the value that you're selling is like oh I was giving six hours of my time at $100 an hour and I'm making 600 bucks but now they know that I can do this in 30 minutes. And so am I making 50 bucks now? Um would they would they just go get somebody else uh to do it for 50 bucks instead of paying me $600? What is my unique value? And so those are like those are some of the the signs that I was thinking about Yeah.
recently. I'm like
Definitely. Oh my god.
We're in this strange in between times and if even if you were to go back to that hypothetical example you gave of Sarah at the beginning. Um so yeah I'm not see I'm still working and you know my employer has mentioned AI and we're supposed to use it but we're all still employed and so are all of my friends. But then Sarah how much are you using AI to maybe draft an email? Or if you work in software, the realities of AI are showing up a lot quicker for you. And, you know, depending on which country you're in, AI is already part of a lot of people's daily workflows, even if it's making a recipe, even if it's, you know, how do I rephrase that email? That's a pretty significant and fast integration compared to past technologies. And then your point about, sure, if you are in a service-based knowledge-based um pricing model where you're paying me for my time, whether that's legal services or my financial valuations, now I know that that's going to be 40 minutes, not 40 days. So, there will be a repricing of these services to account for the new floor being a significant portion of the task is completed by AI, but the judgment to direct that AI system and to sign off on it, I mean, if you were doing something really legally risky, and I don't mean actually illegal, but you were doing, you know, transactions, buying a bunch of companies, things that you would want to make sure you had a a lawyer's view and sign off as this all of this makes sense. I'm going in a few different countries. You would expect that lawyer to work faster, but would you be okay with just ChatGPT signing off on it? Maybe not. So, that lawyer can now do what would have taken a lot more hours in a few hours, but you're still probably going to pay for that person's judgment, but we don't need as many people doing that. But what is the AI system and workflow and series of agents that that legal firm has now built to do that type of work? And who's building those AI systems? I bet it's not the partners. That's a new cohort of people directing agents that are going to come up.
Mhm. So, we're going to probably see some repricing, and that's why it's also, I mean, depending on whichever work, if you do work on a computer, you do have to expect that right now your manager [clears throat] isn't valuing the fact that you're using yet, but soon that's going to become the expectation. And that's where the we might see some repricing of labor as a result. And that's the same thing if somebody was working on spreadsheets versus spreadsheets on a computer, we'd expect them to be faster. Uh that expectation is coming with AI when it becomes general purpose and it's just a thing we do work on. So, if you were to think about the the layers of this, I want to maybe getting your opinion if this is accurate or not. So, 2024, 2025, like people just starting to discover this technology. And then they let's just say they start using it well and they're like, "Oh, wow, I could like save so much time." Still getting paid the same amount of money from their company. So then, whatever their workload was, now they just like shaved like 20 25 hours a week off of it and they're like coasting. Are you saying that now we're in this kind of in-between period, but that that's not the expectation now, is it going to be like three times or four times as much as what you were doing before? And so any gains that you would be making are basically now going to be offset cuz that's the new standard. Sure. So, yeah, arbitrage opportunities don't exist in perpetuity, right? So, yes, the fact that you may have saved 25 hours off your week and I actually have friends that are like that. They're like, "Oh my gosh, I'm in marketing and now I have half my Fridays off." And it's great. It's not that So, that mindset that we would Now we have to do triple the work because AI is now expected of us. It's not necessarily that. In the short term, yes, maybe because you're expected to do the old tasks and now you have AI so you can do them faster. But the real change that comes in with these technologies is that we do different things. So, the work you're doing may not even make sense in the next couple years. So, the tasks that you do within your marketing role, it's not just now you have to do a hundred times them because now you have AI so you better be faster. You might do something entirely different. So, both of those things and that's what's hard to see about the future. But yes, ride the wave now if you have an arbitrage arbitrage opportunity, either take it if you're if you need that extra time or use it to start really building some of these AI skills.
[snorts]
And I want to I want to get to the the the changes in the in the skills that are are needed. One other thing that I want to ask about is in terms of signs like so the big thing in the news past 2 weeks was was Block. Jack Dorsey runs who was one of the co-founders of Twitter, runs Block, which is the Square technology. They laid off 40% of their workforce and then cited AI. Meta is now apparently going to lay off 20% of their workforce. Do you look at those as additional signs of something that we're going to be seeing more and more and more common in the news in the coming year or two? Yeah, and it's funny cuz I did strategically leave that one out of the data that I'm consistently tracking.
Um yes, so some of these announcements are definitely related to AI where we're seeing you 15,000 layoffs or we can freeze hiring, you know, 10% of our workforce. Some is definitely AI related, but it's not for that one-to-one replacement where we don't need to hire because AI can do your job. Realistically, AI cannot do your full workflow um unless you did [clears throat] something very specific and the same thing all day, every day. But AI could be related because a companies are moving, reallocating capital away from labor towards AI and AI infrastructure. So, they're not replacing you with AI, but they're assuming that either your workflow isn't going to be needed in the next evolution of their organization or they're still better off putting that towards AI infrastructure and redesigning their company for the AI age. So, I am looking at those signals and they're important. And what we also haven't seen yet that I see in my work with companies, is how they are redesigning that org chart around the workflows of the future. So, okay, we don't actually even need not because AI can do X, Y, or Z, but we don't need this this entire cohort of people, but we're we're going to play in a different game. So, that just doesn't even make sense whether AI can do it or not. We are over here in the AI economy. So, all of that's going to start to happen. Yeah, so we might hear more announcements, and we should expect that. Um don't when you hear AI assign it loosely to those narratives. And some of it actually was just over-hiring during the pandemic, but AI is a lot sexier of a story. So, some of that is that bloat. We're also in a period of tremendous economic uncertainty. Um I think I don't need to explain why, you know, we wake up every week and we feel like we're in a literal new economy, a new world. Companies have to price that in, that economic uncertainty. Uh so, that shows up somewhere, and that can show up in labor. And then some of it is making a bet that AI could probably perform a workflow signi- it too is a good enough job that it a company could take the risk of firing a few people and replacing them with AI even if it makes a few mistakes. So, it's all of that. So, it's a bit more complicated and nuanced than AI is taking your job. One of in the video you released a couple months ago, one of your biggest predictions for 2026 was there would be a backlash or like movement sort of against AI that emerges throughout this year. So, what specifically do you believe would be driving that and like what would that even look like? I think we're going to see a multitude of reasons of why people are going to push back on artificial intelligence. And we're already starting I mean, the narrative in we'll see Democratic a lot of Democrats societies or we can see traditional Western societies versus the global south. We'll say the global north versus the global south. The narrative towards AI is the sentiment is already much more negative compared to to the global south. Uh and AI is is often framed through the lens of this threat. So, we can do your work, your work isn't really valuable, and a robot's going to come and be able to replace it. That's very destabilizing for people. That's very unnerving for people. Uh some communities are already struggling with data center impact. So, whether that's their water's been impacted or the you know, electricity prices, even though it kind of depends where you are if you're impacted by that. Uh I think the consolidation of power is really throwing people off. Uh and that's something that's really real. That there are, um you know, seven to 10 people, depending on how you look at it, that have outsized power and are building the infrastructure the entire world already depends on, but will depend on even more. I think that's not a good uh a good a good look. Um and then you have instances of, you know, AI locking people out of their financial services or out of health care. So, all of these different and then artists. Uh you know, what's happening there? Has anybody settled any of the copyright disputes?
No. Who knows what's going to happen there? Uh and then we'll start to see more and hear more about the job layoffs as a result of whether it's cuz AI can do the job or we need to free up some capital. So, all of these are like these different trains that are going to collide at some point. Uh and I think that's going to be sooner rather than later. And the more you also hear about some of the riskier end state scenarios with AI, some people call it the existential risks. I think that's going to panic a lot of people, too. So, I think that ends up in a scenario where people are like, this technology is coming for my job, it's coming for my electricity bill. Um I don't relate at all to the person driving this future, and they're telling me that my work is as valuable as a system in a data center in Memphis. Why do we want this? Um and so, you can you can sympathize and empathize with people, and I think that that's prob- we're going to start to see some some backlash to artificial intelligence. Mhm. Yeah, this was And the environment. Let's just which one another one we didn't discuss.
Yeah. I mean I think it was it's probably about 10 months ago now. Um I I did some like modeling through chat GPT of these sort of scenarios. Like painting across the board and then basically what it said was that it felt that that that that like protests in the streets type of thing would be likely by like 2028 like 2029. Um and there would have to be to your point, there would have to be serious measures put in place now like by the government or seriously serious things to um work through the potential changes that could happen because some of like the the change that is happening even if even if there were like new things created or we're doing new things like to your point like there's a second part of the story. The the social of people and the change like to actually mitigate that there would have to be a lot that's done right now and basically I think it was saying that that work seemed to not be that's not happening right now. Um and so when you hear like a scenario like that like literal protests in the streets from people, do you find that to be an exaggeration or do you feel like that is could be a reality? I think we could absolutely see protests as a result of this technology. I would expect that and I think people we're already seeing micro things that go viral on social media and whether you consider that a simulation of the physical world even though it's completely we're often delusional in that environment. Um things going viral about the negative impacts of AI on your life and then think about the story around AI that's being sold. There are and again I'm not inherently optimist pessimist I don't necessarily find those titles helpful but there are a lot of good things about this technology. There are a lot of reasons to fight for versions of AI where the world would be better. We don't really hear that. We hear about solving diseases as a kind of byproduct. Uh but the narratives that are being championed is you know, AI can is going to save us in national security. AI can do most of the jobs. Um AI can do music better than the people who made the music. So, let's just move this train forward. That's not a very compelling story for a lot of people. Most people can't relate to that. And then yes, as you as you had stated into the transition that really concerns me. What happens from now until that new economy? Who's brought to the other side of that? That's a design space. That is not some kind of prescriptive future. And we don't really hear people with ideas designing how we ensure people are safely moved to the AI first world. Uh I don't Yeah, I don't We don't hear enough about what that looks like and how you protect people on the way there. Um and so I imagine there will be AI backlash. I wonder if if people also will start to sense how AI is impacting them cognitively. And and so I mean last summer we had a neuroscientist on talking about this. He was talking about like parts of the brain whether metaphorically or or literally becoming atrophied. Um and I didn't quite get what he meant by that. Um cuz he was talking about the default mode network and he was talking about a lot of the quote unquote muscles that we use in the brain constantly and that we have used throughout this whole knowledge work area. And that we are offloading those capabilities completely to AI. And I think that there have been moments in the past few months where I've noticed that with myself of like why is it so hard to think certain tasks? Why is it so hard to just like think and do this right now? Um Is that something that that you're sensing that like {quote} {unquote} AI could make people cognitively dumber than they were less intelligent than they they were before because now they're offloading all their work to AI. Yeah, I think um yeah, and some people call it cognitive atrophy, some people call it, you know, cognitive debt where you give AI the the task and even if it's just some something as simple as writing an email and then you do that for three, four, five months and then you're like, "What? I don't want to write this email without artificial intelligence the same way I really cannot navigate without Google Maps. It's just not going to happen. I for the most part, right? Um and it's not that I'm necessarily worse off now that I don't have a map anymore and I'm not using like, you know, where's the North Star? And I'm not necessarily worse worse off cuz we've replaced a lot of that navigation with different types of thinking. Um so there is a world where you can use AI as a scalpel or as a springboard for different type of thinking, but are we doing that? Um have we designed a world where that makes sense? Not yet and we've essentially just thrown these tools out for everybody to use. There hasn't been much instruction of how to avoid just offloading everything to the AI system and it is being designed in such a way that the intention is for it to be your everything. So, if you are using it as a springboard you could launch forward and that's why I wrote this article on Substack called, you know, the cognitive divide that I'm seeing where there could be a cohort that really breaks away and they do use AI in a way that enhances what their output is and what they could become with it and then those that just start to give tasks to AI cuz it's easier, it's more convenient, it definitely writes better than most people uh and that's the that's the the offloading of the thinking. And the challenge with that is you don't just offload your your judgment skills and your ability to to you know, discernment, your ability to solve problems, you also offload your confidence. So, it gets to a point where you don't feel confident making a decision without AI. And that's a scary place to be for an individual and for a society because there's going to be moments where we have to deviate from the system. There should be many moments where an AI system has told you something and you've recognized, "Nope, that doesn't make sense or I should triple-check that fact because it isn't a fact system, it's a language system." So, and then and are people doing that? I'm not entirely sure. And this AI isn't coming in in a vacuum where we were all using maps and and navigating the world with memory, it's coming off the back of shorter reading cycles where since the '70s we've been reading less and less and less these shorter excerpts and that's partly because of education and optimization for marking, partly because of different things that we can spend our time doing like TV. Then you see the rise of social media. So, 90-second clips, 60-second clips, learn about the entire world in 90 seconds or less, losing decoupling context from content. And then you have Google Maps. So, all of these little things. So, our our attention span has been rewired and then you throw in this system that greets you everywhere and says, "I can solve this for you. How can I help? Can I take this off of your plate?" So, off the back of that, AI does pose a greater threat. It doesn't have to, we can use it as a springboard, but the default, I mean, we're also human. We are biologically wired to seek the lowest energy state. If a system's going to do it, I'm going to choose GP GPS and the satellite over the physical map. I would be totally toast with a physical map. So, now you're giving me the system that can allegedly think better than me? Huh, it's pretty tempting. So, a really profound insight. I think even in terms of our our expectations, I think the thing about the GP GPS was spot-on because you know, if you go back to what we were talking about hundreds of years ago, right? I mean, the idea of even having a car to like go somewhere was absurd, right? And so, you had people navigate Like, if you watch Leonardo DiCaprio's movie in like 2023 Killers of the Flower Moon, and you have these people that are like traveling in Oklahoma on on the back of a horse carrying all this stuff. I mean, they're just like navigating this road and and trying to figure out the way to go. So, and then you get cars, so then you can accelerate it. But, now we're at a point where like the internet, then you had GPS. And now like literally it's like you feel like you're doing something wrong if you get in the car and you don't turn on the
driver took out a map, you would say pull over. Yeah.
Please pull over. I do not trust this vehicle. Exactly. And so, that's such a huge deviation. But, now it's like yeah, we expect that. Mhm. And so, I wonder if we're going to get to a moment with these cognitive capabilities where like just the expectation is like no, we offload all of that stuff. I found in my own kind of personal work, what you're saying is is true with I feel like I have to come up with what I want to do first in some ways, and then basically it'll take whatever authentic version of who you are and what work you want to do and and take it from like 70 to like 100. And so, it feels like in some ways an extension of you versus like outsourcing the entire experience.
the thinking that happens before you engage with AI and after you engage with AI are really it's really important. So, how much you actually think through the problem that you're asking AI? Or are you just constantly giving it half-formed thoughts? Because AI isn't going to magically raise the bar in your thinking, it's going to match it. So, if you're coming in with these kind of half-formed ideas, just fragments of thought. Sometimes it can be helpful, you're like, I think I'm thinking of something, I don't know what I'm trying to say. But, by and large, if you can do most of the thinking, structuring your problem properly to the AI, and then when it responds with something, don't just blindly accept its answer. Can you interrogate it further? Do you need more data to make a strong judgment call based on the output? One of my favorite economists, Ajay Agrawal, discusses, you know, what are we going to do in the future? We will direct AIs by applying our judgment to them. So, judgment is in what you ask the AI. You mean, you have now access to a supercomputer. What do you want to ask? And if it's the same questions you were asking pre-supercomputer, well, that might be a problem. We can ask deeper questions, higher-order questions, questions that require more computational thinking, so to speak. So, so again, it doesn't have to be that we're doomed with these systems. We can raise the bar, but are we doing that? And do we have an ecosystem that is facilitating that? You know, what's happening in education? Slow blanks. You know, are people in in the workplace being incentivized to I mean, and here's also a good example for employers. If you're telling your employees, use AI, we're an AI-first company now, suddenly, randomly, are you asking people to really explain their reasoning when they come and present something? Ask somebody call them on the speak. So, what is the thought process behind that? And why did you go with X over Y? Is that you therefore a ChatGPT-generated idea, or is it yours? So, all of these different ways we can force judgment. And sorry for everybody if their boss is listening and now they're going to do that. Now, if we were to think about the wide range of scenarios of the ways that this could go in the next 2, 3, 10 years, I think part of the confusion for people um is obviously they see a AI as a tool, and but they don't fully understand like how the technology works. So, they they open up ChatGPT, maybe they're even generating all of these ideas, they're writing emails faster. So, it feels useful, like even revolutionary in some some regards, but like these kind of scenarios of like it would replace them, or it would completely change the way that they're working feels out of touch. And so, are we like are we do people underestimate that technology cuz we're we're only seeing this like interface and we don't see the infrastructure beneath it? How do you think about that? Yes, so you're so you're essentially seeing when someone ask AI ask AI to do something for them. Yeah. And it works okay, but it's not great. And then you're saying, "Oh, but this is actually maybe going to replace you down the down the line." And someone's trying to
disconnect.
reconcile that. So there's a few things that are happening. One, yes, it's understanding how these AI systems work. They're not fact-finding machines. They're not even really complex reasoning machines. They can simulate reasoning, but they are prediction engines. So they analyze a bunch of data, mostly language data, but it could be images, and they spot patterns in that data, and then they learn to predict to predict the next word that should come next in that sentence or that sequence, next token prediction. So they are optimizing to be likely or they're sorry, they're optimizing to sound right, not necessarily be right. So understanding that's actually what's happening under the hood. A lot of the times if you're writing an email, that's sufficient. But if you're giving AI a more complex problem and there cannot be room for error, error, you're not just kind of reasoning about some idea or some kind of brand slogan, it's probably not going to work very well for you if you just ask it a one-off idea. And it also depends on how you're structuring your problem and your question to AI. So if you haven't fully understood how you interact with these systems, how you give it context, how you follow up on its work, it'd be the same as asking an intern to do something and never giving it any background on what you're asking it to do, what you're asking that person to do, they probably wouldn't perform really well off of just a one-off. So that is also going to take time. But yeah, these systems aren't at this point designed to do things that require a lot of steps and a lot of memory and a lot and keeping an idea consistent over a long period of time, they're going to struggle with that. But they I would say they are worse AI systems are worse than some people describe them and a lot better than some people think they are. And so that's kind of the space that we're residing in right now. Mhm.
Tristan Harris said that people behind AI don't fully know how AI works and I wanted to get your take on that. Because it is when you're talking about this being like a predictive text on steroids. Like it almost seems like that is the technology but almost seems like reductionistic for for what it's outputting and how fast it's doing and how complex what it's saying back to you is. Almost like it's reasoning in a human mind. Are there as Is there Are there truly aspects of the technology that the people that are making it don't fully understand how it works or you feel like they have actually mapped out the entire spectrum? No, that's true. It is true.
[sighs]
it's not that they don't understand at all what is happening but generative models aren't programmed generate those outputs directly. computer scientists and anybody building the models truly don't actually fully know why exactly this works. And that's what also makes them quite unpredictable. So they do know that the more data you give these systems they tend to be a little bit more accurate and the output gets better. But why? What's really happening under the hood in that black box where you go from here's a bunch of data and then write me an email or an essay and it sounds amazing. The AI is learning those patterns on its own. It's putting different words and and letters in its own vectors and finding associations between words. They you know it will translate them into different numbers on its own. No one is telling it to do that. No one is saying that the sky when someone says the sky is the answer is usually blue. It's found those patterns on its own. And so that's what people mean when they say that they don't AI programmers and leaders don't actually fully know why this works. They don't but they do know that you can reliably throw a bunch of data at AI systems, generative systems and they'll be able predict the next word better and better the more data you give it up until a point in time. How would you think about the word intelligence? And so human intelligence versus this thing being intelligent. Um Sure, I'd say AI can perform cognitive tasks that humans deem intelligent but it's doesn't necessarily have an understanding it doesn't have an understanding of the world. It's simulating how we reason. It's simulating our ideas. It's not necessarily thinking at all the way humans do. It just so happens that a lot of our how we think can be mimicked by statistical patterns. So maybe that means in some ways we're a lot more predictable or the language we use to describe our thinking is predictable. But it I don't think these systems are they don't think the way humans do. The use of intelligence sure yeah people you can there's the camp that says you know these systems are really intelligent or the camp that says this is not actually an intelligent system at all. It's just an amazing pattern matcher. I think both of those are kind of true but yeah AI isn't doing what people are doing. It's doing something different. I think what is stunning is that language, which is how we describe how we're thinking, is language. So, AI is really good at predicting the words we use to describe what we actually mean and what we're actually trying to say. Uh, and so, that's what's really happening under the hood. It doesn't mean that AI is going to stay in this type of architecture, but that's what's happening for now. Yeah, cuz I feel like words hold different meanings and connotations to people. So, if you read a headline about AI being more intelligent than humans, or even the word conscious, like that might mean a lot of different things to different people. And so, it's when you're trying to understand what people are actually talking about, it could be quite confusing uh confusing, especially when you talk about consciousness, too. Like, is AI conscious? How would you think about that aspect of it? Uh, I don't think the AI systems we have today are conscious at all, but AI sounds conscious because it was trained on data from conscious people. So, if you trained AI on data from dolphins, it would mimic the noises that we hear from dolphins, right? So, we are conscious. How we use the words we use to describe our relationships to one another, this is what AI is trained on. So, but of course, it sounds like us, conscious people. Um, but under the hood, that's not It's just like self-aware, almost. The system? No, just conscious and and it's not self-aware.
It's No, it's not anything. It doesn't have any interest. Um, it's an a really impressive mathematical model cuz essentially, when people say, "Is AI conscious?" you're asking, "Is the data center in Memphis comp- conscious?" Uh, I would say it's not. Of course, you can go down the philosophical rabbit holes of, "Well, you can't really prove it is or it isn't." because we don't even know what consciousness is. We know what it is, we don't know how to relate it to matter. Therefore, we cannot say a silicon chip is or isn't conscious. Yeah. Um but I think we're safer saying that it's not conscious because there are actually a lot of uh consequences of throwing out the idea that AI is conscious. If you have a society that starts to speculate that AI might be more than it is, one, we're going to trust the system way more than we should. Um if you think it has feelings or sentience, you think it has your best interest when it has no interest.
Mhm. Uh and two, imagine a world in which people are starting to advocate for AI rights and AI, you know, voting voting rights in society. All of these things, like you could actually go down that rabbit hole. So, a conscious AI system is actually would be a really significant idea. I don't think these systems are conscious, but they sound really good at simulating us, conscious people, and that's what they're designed to do. They're trained on language. But it's so funny, nobody would say these systems are conscious. No one's going to say Dolly is conscious. So, if we started with image generators or video generators and not language, people most likely would not start to think that these systems are conscious. And it's usually the same amount of compute and the same computational exercise that goes into building these systems, but it's because language is so deeply connected to how we relate to one another. It's how we express most of how we are feeling in a in a in a verbal way. And now we're pass passing that capability to a system that [clears throat] can use that technology in ways that are indistinguishable from us. Mhm. So, of course we might think it's conscious because we're also biologically wired to associate language with consciousness. Right? So, we've been trained evolutionary 200,000 years to to get to this point and to use language to communicate what's in my brain to yours. And so, you know, I would be at risk as a species if I mistaken an inanimate object for an inanimate one, if I thought that that wolf I thought that rock was actually a wolf, well, I'm toast. So, when you chat with an AI system, of course you think that there's a person on the other side of that because you've been biologically wired to think that. Yeah, I think that I think some of the some of what you're saying right now is is part of the and I'm thinking about one friend in particular for why people say that these sort of doomsday scenarios will almost certainly never happen because unless cuz I think about intelligence and and consciousness in a way their only reference point for that really is like humans. And so, I mean, you could think about other mammals, but I mean, when it talks about when you think about something taking over the world or taking over all of these jobs and being able to replace you completely if AI in its current iteration is what you're saying and that that's how it works, then I think people will have a really hard time with believing any of these scenarios unless there's some sort of massive leap in intelligence or the technology just completely changes a later iteration of the technology, which allows some of the complex reasoning and self-awareness and even more unpredictability in the way that it would behave. So, do you think those do you think that line of reasoning that this is overhyped based on that present premise is accurate? Uh and you say that line of reasoning that these systems are still in there's they're bounded in how sophisticated they can get and how much reasoning that they can actually simulate, therefore, they're not truly a threat in many in many in
Threat to our existence and threat to our livelihood and our our jobs. I I I and I would say no, not necessarily because I think you can still use these tools as a threat multiplier. So, an actor uh that wants to commit harm in some way can now springboard uh using an AI system. So, we're going to see this kind of rebalancing of power. Uh So, you can still cause harm using current AI systems, which is quite scary. Um But, it's But, it's it's through human agency that that's coming to pass cuz what I'm really addressing is this alternative race doomsday scenario that is like the AI is going to be autonomous on its own and could in some way rival humans in terms of the control of the world and then of course taking over all the jobs because they could just they're smarter, they're more capable of doing it, and they could reason at complex levels.
Mhm. And I think that Well, those scenarios can happen in different ways. So, you could have a um AI that continues to advance over time uh and it impacts jobs more, but it doesn't necessarily also mean that we're we're headed towards an existential takeover by the technology.
Yeah. Um or you could also build AI in such a way that it is recursively self-improving and it still doesn't have inherent desires to do these things, but you've hooked it up to critical infrastructure uh and it's either made a mistake or it realizes, "Oh, I need to shut off this water plant because I need to achieve this goal." And so, it's not necessarily after people in the way that sometimes it's framed, but it's almost like a byproduct of what it's trying to do. Yeah. Um so, those are are possible. But, where I stand on on the kind of existential challenges of of artificial intelligence, I'm really happy that there are scientists, really smart scientists, working on that. Um and as a futurist and someone that does foresight, you always want to have a plan for the best and worst case scenarios. I don't know if it's if it if everybody should be, you know, at 6:00 p.m. going to to dinner with their family and solving the existential scenario and and mapping that out. I think that there are other things that people can be doing and in preparation for this technology. Um, but sure, it's not a a zero risk that AI just starts to behave in ways we don't understand, but I don't think it's in the ways that we tend to imagine and we tend to anthropomorphize. Um, but it doesn't mean that the technology in and of itself isn't risky. So, one of the other predictions that you made for 2026 was that LML LLMs, which is the, you know, the technology behind these chatbots, may be approaching a plateau in terms of their intelligence gains. So, that's highly relevant to sort of Mhm. what we're talking about. What kind of creates that ceiling in your mind? Yeah, so this transformer architecture, the the architecture that underpins LLMs, so they're the next token prediction, essentially. Um, one part of it is the we're seeing diminishing returns. So, you can throw a lot more data at these AI systems that we had described, they you know, read a bunch of text and spot patterns in that text, and a lot more computational power at them, but the output is starting to improve only marginally. So, it's starting to make less and less sense to continue to crank out these massive models and these massive um, advancements in next generation of these models because the return is diminishing over time. Um, so that's one aspect of it. And then there's these upper limits of how far can you really take a prediction machine? So, if it's still going to have, you know, even if the the the error is only 5% in most of the work that it does, well, if you're going to attach, if you're going to give that that system 20 tasks, that's, I don't know, what, a 40% now error rate over over that horizon. That's not going to be acceptable for a lot of workflows or a lot of industries. Uh so we might hit the limit there as well. Um and then I would say I mean by and large why would the AI journey suddenly stop? Right? So if if we've moved from, you know, deep learning and then we saw the rise of transformer architectures in in 2017, why does the journey suddenly end in progress? So I think we will start to to see companies. I think that they're actually already on working on this. Just last week Sam Altman stated, "We're going to need something past the transformer architecture that will be as radical as the as the transformers once were." And when I say transformers, I'm referring to the type of architecture underpinning ChatGPT for somebody that that doesn't know what I mean there. Um we saw seen Yann LeCun, the former head of AI at Meta, say something very similar. We've seen Demis Hassabis, the head of AI at Google, say something similar that we're going to need something different to take us over the curve so AI can be more reliable, more predictable. It can solve long horizon tasks and actually do that complex reasoning. I mean a LLM it doesn't actually understand the world or even simulate it. Yeah. So it doesn't actually know if you tip your glass, it could say that it's going to fall, but it can't actually simulate that water hits the ground and what all of that means. So if we're going to have systems in the world doing all this stuff, they're going to need to actually be able to simulate the world and LLMs don't do that. Yeah, I guess what the the example that comes up in my mind and this ties back to what I was saying about Bill Gurley and and the fear around AI and that these big picture narratives that people keep hearing in the news is that if I were to say to you we're going to cure cancer. And so the equivalent here would be like we're going to create a super intelligence that is is going to be able to just do everything that humans can. And and it's one thing if if I have a path to curing cancer and and it's something like tangible that I could show you. I'm like, we're going to go from A to B to C to D and then we're going to cure cancer. And so if people are hyping this LLM technology and they're like, we're going to get to super intelligence from the LLM technology, what we're saying here is that it might end up plateauing. Then the idea of super intelligence coming through LLMs is is probably not going to be the technology that gets us there. So then where that becomes relevant is that basically are we at like ground zero then? So then I'm telling you I'm going to cure cancer, but I don't know exactly how I'm going to do that, which is basically like anything else, right? So are we going to have to create a brand new technology that we don't even have a foundation for right now? And so that would radically change some of these future scenarios that we're talking about and our ability to even get there. Yeah, I'd say so one the systems we have today if all progress stopped and no one could do anything else, they are still radically disruptive.
Yes, I agree.
And I think people need to still understand that. Most companies have not even absorbed what that's what a present-day LLM means for the relevance of their business model if it's going to be relevant at all. Um so that there still could be a lot of disruption we haven't absorbed and diffuse these AI systems to their potential.
we were talking about in the beginning of the conversation, yeah.
But then if like you're going to say, okay, the system's going to be able to do anything that a human can do in one shot, um it can you know, a robot could walk into a room and it could pick up a bunch of things it hasn't seen before, it could simulate all that stuff. Is this architecture going to take us there? Probably not, but I don't think we start at zero. I think the stack will likely continue to build so it could be an LLM with, you know, a world model, so more of a world simulator on top of it or like a large action model on top of it. And then there's also neuro-symbolic models. So these are models that uh you could give them more rules and constraints to follow. They're not just predicting uh the next word. So, companies, academics are working on um architectures that can complement large language models. So, I don't think we start from zero. And then, of course, you can't ever predict a breakthrough. So, there sure, I'm sure there's a lab somewhere working on some architecture that may seem irrelevant um or unrelated to transformers. Uh so, that's probably happening, too. But sure, if someone was to say AGI or ASI by 2030, no matter what, you can't make those types of predictions because yes, we don't necessarily have the path. And those are the but those are the predictions that like people hear Mhm. in the news over the past year. Yeah, so I think yeah, those are the predictions. You can't really tie a date to it, and that's why the date keeps changing. But you can say over the longer term, these systems are going to get more impressive and be able to complete more non-cognitive tasks. Now, is the technology behind the robots and so like I was just in LA, you see the the the Waymos, uh the self-driving cars going around, which on its own is just mind-blowing to me that like the amount of things you're thinking about as a human when you're driving a car and the things that you need to be aware of and the 10,000 different scenarios of like that person's getting close to me right now. Uh I should probably like slow down, and it's happening in like a like a fraction of a second. So, to think about a self-driving car even in its current iteration is just wild. And then, you think about what Elon is is building with the Optimus robots, and and then you go into China, and the they seem like even more wild what they're doing, picking up things or whatever. That's not LLM LLM technology, is it? Or is It could be a mixture. So, there's a lot of different types of AI, too, right? There's computer vision, there's deep learning and machine learning. So, all of these different types of AI systems, depending on what the task is at hand. So, if we're driverless vehicles, no, it's not just going to be a next token prediction. When you interact with your test verbally, it is. Um but there's all different types of AI. It's actually a very broad field. Most people think of ChatGPT
Okay. when they think of AI, but there's all different types. And that's also what's happening in science. Uh so there's more nar- narrow AI systems that are being used to specifically work on Alzheimer's or medical problems, and they're much more appropriate to biology than just ChatGPT is. So all of these things are also happening in parallel. They just don't get as much um visibility because they're maybe not as interesting to people. What they're working on is interesting, but they don't come with these kind of grand statements and gestures. So what sort of So the big thing that has sort of been hyped is is agents. And so if people are listening and they're not familiar with that terms, I mean an AI that could do a task autonomously and do it well and maybe even have the capability of humans. What sort of technology would be required to actually develop those sorts of capabilities? Yeah, and we're we're getting closer. So Clawbot, Mont Bot, um some people may have heard of that. Uh it's almost an AI. It is an AI agent that can do things like send emails and stuff, just not necessarily reliable reliably. Uh but sure, in a world where AI can hold a state consistently over many many different tasks, and it can do 25 things. Um a large action model could be something more applicable for agents or a neuro-symbolic model. So the AI actually has constraints um on top though. Like you'd want the language models so you can interact with it and just give it um an instruction in human language. Uh but you're also going to want to be able to hook it up to tools. You're going to want it to be able to predict and simulate what action should come next, not just what word should come next, and be able to do that over 20 different um 20 different tasks connected to one another. Companies are working on that and there is elements of agents where you don't necessarily need all of that reliability for it to be useful and productive. Um, but we're we're we're not as close to reliable agents as people think, but we are still use they're still productive agents that exist right now. And whether you want to call them agents or not or quasi agents.
Cuz in some ways agents can sound like a nightmare to the average worker working at a Fortune 500 company, but on the flip side to somebody who's like an entrepreneur, it feels it could feel like a dream of like rather than hiring 50 different people with money that I don't have, uh, you end up employing these 25 to 30 agents that are doing all these tasks and you're setting them up in different departments of your company and they're all operating in and carrying out these tasks at a very good to elite level. I mean, that would be like hugely valuable and and those sorts of scenarios I thought were fascinating of like could there be startups and people in the future who basically are doing the workload of like 300 people and there's like five of them with like 200 agents. And that's already happening. So, AI first companies where AI is the native software that they're building with from scratch. And so, yes, so these kind of quasi agents where it's an LLM that has access to different tools. So, it it has access to access to your email, it has access to a bunch of financial reports or whatever it is that you need as an entrepreneur. You don't have the money to hire financial analyst or someone to get the latest headlines about something say in geopolitics, but that would be a a helpful hire for you. You now, maybe for 20 bucks a month, can install an agent to keep a tabs on headlines, write the draft of the newspaper, uh, or your your newsletter. Have another agent that files in all of the invoices and just sends them to one person instead of 10 people in payroll to approve, that's a game-changer for for entrepreneurs and startups. And those all exist. So, I think this next chapter what we'll see in entrepreneurship is going to be incredible. And think of all the people that have ideas, it's just way too much effort to go and start them or rate way too much risk. Now, you can start to stream these systems and and see, you know, even with the the kind of basic coding tools that we have today, you can get not an MVP that's maybe cyber secure yet or reliable, but you can start to see the architecture of an app for 100 bucks. These are kind of game-changing new paradigms that haven't fully We haven't fully seen the outcome and the output of what this will mean at scale, but it's really exciting. It's really cool. Or someone with an idea that just is terrible at writing newsletters and terrible at copy or terrible or would never have hired somebody to do graphics for their website. So, but now they have AI to do it and they just would have never hired anybody. No. That's really cool. Yeah, I guess the gap right now it would be well, not just capability, but like the actual handing over of trust, especially when it involves interactions with like financial things, uh large data sets, um interacting with other people, obviously on on your behalf and so like mistakes that are being made, but yeah, I think the prospect of that for an entrepreneur is incredible. So. Yeah, that part is really exciting. And even I mean, I had a nightmare scenario where somebody created a digital twin of me that could you could call it, you could do all this stuff. Without your permission?
Yeah, I had no idea who this person was. And they had all this stuff about it. You can call it, you can learn, you can ask her questions, she'll tell you about AI and all this stuff. And I can't even remember how I got notified about it. But I didn't really want to hire a lawyer to deal with that and pay all that money. So, I just used ChatGPT to draft a cease and desist and I sent it to the person and they took it down. Uh and so, that's a scenario as someone who runs their own company, it would be nightmare for me to bring on legal a legal to do engage in a legal dispute over that and it worked. Um so, those are all and you can imagine that type of scenario at scale. It's pretty cool. Yeah, that is really cool. I mean, the the access to information and things is just is just mind-blowing and and sort of these gatekept type professions and communities and things that you could never access on your own now suddenly become Of course, it hallucinates sometimes, but it becomes accessible to the average person. Yeah, that part is really cool. And we'll see more of that. I mean, there [snorts] is of course a world where we kind of splinter and we live in these kind of AI-generated realities, but there's also a world in which finally, most people aren't going to tune in and watch the Fed chair speech, but what is being said could actually matter for how you're going to do your family planning or which if you should take that job or not. Most people just don't have the tools. They haven't studied finance. That speech is irrelevant to them. It doesn't make any sense. But now you have a system that could summarize and customize the content just for you. So, we will see this kind of opening of gated intellectual domains for the average person, which is really cool and will lead to all sorts of new unexpected unlocks the same way the smartphone and decreasing cost of communication and photography led to all sorts of new ideas and companies and ways of working. So, a few I think a couple months ago, Geoffrey Hinton who people are not aware is like called the godfather of of AI and it seems like when you look at his scientific work, it seems like basically all of his technology is being built on some of the things that like he developed back in like 2012. Um so he said that AI capability is compounding at an extraordinary rate. So, in some domains effectively doubling within months. Um, and so when an expert makes a claim like that, what exactly do you think they're measuring that it's rapidly accelerating? So, it is we are seeing really intense acceleration and exciting acceleration for better for worse on thing in things like software or maybe areas of mathematics. So, not necessarily in things that AI is being trained and optimized to do. So, not necessarily tasks to the average person that are relevant always to the average person, but AI is improving in certain specific domains at rates that past technologies didn't improve at. So, in some in some aspects some of that is is correct. I think people then extrapolate and it's improving everywhere at a 10x rate and that's not necessarily true. But, if you're to zoom down in math, it's it's been pretty fascinating to watch it in software which isn't surprising, it's made of software. Some pretty exciting gains as well. And then of course anything in language. If you were to look at GPT-2, so this was OpenAI's kind of first crack at the can. And I think it came out in 2019. And then GPT-4 just, you know, four to five year three to four years later. That's a pretty stunning jump. It was not very coherent, GPT-2. And then GPT-4 in 2022 could write a high school essay better than most people. It's pretty interesting. That's a pretty big leap. Is it a leap? Yeah, is it a leap that, you know, means that no jobs are going to exist by the end of the 18 months? No. But, when you measure it in isolation, it is quite impressive. Yeah, I mean when you think about those sorts of scenarios ago and you think about 10 years ago that that would even be possible, I think people would think that that was crazy. Mhm. And so the rate of change feels exponential in some regards. And And yeah, if you look at past technologies, like the rate of change just feels like in some in some regards, like whatever one year was before, it feels like 10 years of of change or a technological advancements happening so quickly.
And it And it it that like these these numbers of improvement and these metrics are they are really impressive. And it is interesting how easily AI has just filtered into people's workflows, and they just don't think of it. Just Oh, just ask ChatGPT. Uh the amount of people that turn to that system over the course of two years for most people. Yeah. That is quite stunning if you think about how long it took people to get on email, and then to, you know, build a website and have a a presence on the internet, to how long it took AI to become in people's day-to-day workflows and work lives. It's pretty striking. Mhm. So, we address we address like the claims, some of the scenarios, the technology. I mean, if you were to address people's cultural mood right now, and then getting into the practicals of like the skills that they might need. So, you know, there's a noticeable undercurrent of powerlessness that people feel as a result of these headlines. And And so people feel like this is happening to them, but it's not happening with them. And let me just ask you this directly. Do you feel like this trajectory that we're on is like inevitable? N- Nothing about the future is inevitable. AI is a general-purpose technology, so it's going to not the technology literally itself behave like electricity, but it will become much more foundational like electricity has or like the internet has. Um so that much I mean it already is um, moving in that direction. So, AI becoming a part of our lives, it already is. So, is that inevitable? Is it not? It's already here in that way. the the the outcomes, where do we go from here? None of that is inevitable, right? The future is the combination of the decisions that we make right now. Decisions are still being made today, and they're going to be made tomorrow. And all of those decisions will steer us in different directions. So, I understand the feeling of powerlessness, and there's areas that I feel it, too. And the power asymmetry is real. But, those are also designed spaces, and decisions can be made that move us in a different direction. And that's not inevitable. That's still up for grabs. So, I wouldn't I I can I do empathize with feeling powerless, and I think it is important that we pay attention to these power dynamics. They're extraordinary. And we haven't seen power asymmetry like this, private sector over public sector, since 1700s. So, this is really significant, um, but we aren't doomed. And where all of this nets out, that is still a open design question. It's not done. It's not solved. Even though it's going to be pitched as solved, especially if you have to raise a bunch of funds for it to turn out that way. Um, but it's absolutely not inevitable. Yeah, so I think it becomes a matter of where do where can people most find their agency in this conversation? Because some of the powerlessness feels like, okay, well, people didn't explicitly ask for AI like 10 years ago. It wasn't like everybody did a mass survey, and everybody's like, yes, let's create artificial intelligence. And so, I mean, you mentioned it earlier, there's about seven companies that are really like shaping this technology. And then you look at the valuations of those companies. Now you have I don't know, there's like Yeah, there's like five or six now that are worth over a trillion dollars, which if you were to think back in 2011 even, that's like just unfathomable that they they would have these valuations. So, people feel like people are deciding their life for them in some sense. What are the things that you feel like people could be confident that they have like agency over and power or would you direct them to? I mean, how you spend your time a big act of agency. So, from how we spend our even, you know, social media as one massive platform with artificial intelligence powering one a company that's trying to drive the future. Um you can direct your time in different ways. Even the different AI systems that you choose to use, if you feel like a company aligned with a future that's more in line with what you think is a better outcome for for the rest of humanity, then that's agency you can use and just in the various tools. Um there are I mean, I think AI needs to be much more of a voting issue than it is. Uh and so, there are things that you can demand from elected officials. There are things that you can ask of elected officials in this moment. Uh how are you using AI even in your job? I mean, how is your company deploying these systems? Can you get into those meetings? Are there surveillance policies that your work is going to use or not use? Uh these are all things that we can ask uh and we can engage in. And I also think the more and this is counterintuitive because I get that there is also a resistance to use the to using these technologies. The last thing you want to do, if you think somebody is getting more powerful, is give them more power. But the more you do engage with AI, the better feedback you have about it, about who it works for and who it doesn't work for. Uh so, that can make some of your feedback uh just a bit more informed and a bit more targeted. So, I think we aren't as we aren't entirely helpless. And And collectively, too, there are a lot of people that seem that feel the same way. So, the collective outcome of that is kind of steering us in different directions. I mean, the in the different elected officials over the the next couple years are going to be really significant in how this technology gets built. Um in I think it was Singapore just this week there was an announcement of helping people become AI fully AI literate. So, there'll be 100,000 people that get this training because you are a specialist. And this is another thing too, you are a specialist in your domain. So, how understanding how AI does and doesn't relate to the work that you do or the field that you're in. And so, Singapore is going to sponsor 100,000 people to get the expertise that they need in this world. Their whole policy and campaign around AI is we can't this isn't happening without people. We need to bring everybody along in this ride. That's an entirely different paradigm for AI. So, we aren't stuck uh in a world where we feel like leaders are just going to drive in certain way and we have political leaders that aren't necessarily going to have our backs. Um there are things we can buy and not buy. There are people we can vote for not vote for that lead to different outcomes and we're seeing different outcomes around the world. Yeah, cuz it seems like there are there are a few camps of of responses. And and so, the the Tristan Harris response is you know, we need to like halt basically in some ways the just progress of this um and that just like on the reckless development of this without even like thinking through. And so, the energy of a social movement to the point of maybe protest or whatever, people being like, "Hey, this is going too fast and like we need to like slow down and seriously like take a look at what we're building rather than just like gung-ho, let's just continue." I mean, that's one route. And then you have people in the government obviously that may or may not be on board with regulating it. And so, so that's like a whole entire conversation of maybe don't regulate it. Uh like we just need to keep pushing. And so, people hear all these things, and so it's like to devote your attention to that sort of thing or the very like day-to-day practical thing is like just figure out how to use it in your own life and accept that like this is happening. How would you think about those different scenarios? Um I mean, I don't know if those are all the binaries that we have to exist with. I mean, I think you can engage with AI, understand how it what it means for your job in the shorter term, how well these systems do or don't work for you or your community, while also thinking, "Okay, you know what? I actually don't like the speed that these systems are moving at. I think I do want to raise that to elected officials." Like those things aren't mutually exclusive. I think you can exist in both of those domains. Uh the idea that you might be in a country where a government feels like they just don't want to regulate this technology. That is tricky and that is real. Um but for most of those scenarios uh the government or the person in power isn't there in perpetuity. Um and there have been scenarios where people have been quite vocal and policy has gone in different ways as a result. I mean, I think So, I think there are little acts of agency. I mean, and especially in America money talks. So, if we can direct our dollars in a different way, and I mean, I'm Canadian, but I do live here now. But if people in a country feel like they want things to go in a different way, I mean uh the the the markets seem to speak really loud here. Uh so, those are other acts that people can can do, but I would say the biggest fear for me, it actually isn't the existential scenarios that some people paint about the future. Um it isn't this AI race winning, losing. It's that we actually feel hopeless. Because a hopeless society is a disempowered society. And a disempowered society doesn't do anything. We [clears throat] feel like it's over. We unsubscribe from the moment and we let other people do it and other people build it. And I think that's the the only way to guarantee an outcome that doesn't work for most people. So, it's not that somebody It's not that no dreams are going to be built. They just wouldn't be ours. So, I think we still have to the small ways we can exert agency still matter in summation. I'm I would be much more concerned if everyone said it's over. Someone else build it. Forget about it. Yeah, I mean I think I think some some people might So, some of these scenarios might feel unfamiliar or calls to action even approaching a local official or talking about these things. I don't know what percentage of people have actually done that before versus or even unsure of even how to do that. And so, those little moments of of agency like people might listen to something and be like, okay, well, what do I do? What do I do now?
I mean, and the irony is sure use ChatGPT to write a really compelling really compelling email to your local elected official that this is something that matters to me and my community. And there are I have started to see as much as we can trust anything on X, but I have started to see different I will say people aspiring to step into politics or that are running where AI is starting to be a bit more of part of their campaign. So, I think that there's that. There's other countries that are building it differently that show that it actually is possible to have outcomes that aren't zero-sum for the people and for the private sector. So, this isn't kind of you you know, the only game in town that's happening. There are countries that are also coming together to build competitors to the reality that we're in. So, there are all of these different outcomes and kind of signals in the data that show that this is this zero-sum future isn't the one that we have to all be headed for. But I think the worst thing we can do is fully subscribe from the moment. And that doesn't mean everybody has to be a part of every debate everywhere. I'm at the existential campaign, I'm at the Workers Rights Act. You don't have to be a part of everything, but just the small ways you can exercise agency if you feel like you have it, I think are really important. What I love about your work is you I feel like you can you help instill a hope in people in very practical ways. And so whether your podcast and Substack, a lot of the things that you've been writing and talking about, I think it really helps people like prepare for that transition. So like of of what the next 10 years are going to look like. And so if you were to focus you you called this the dawn of the um independent
Mhm. era. And so can you explain a little bit about that and what you feel like is coming? Yeah. Um and of course nobody can make predictions about the future, especially not a futurist, but what we can start to see in the data the rise of more contract or independent based work as the dominant form of work in the market. So if you think about this from the perspective of a company, and we've even talked about it today, uh a financial analyst job today may look very different in 18 months. You still might need somebody to do different things, maybe direct a bunch of AI agents, but the skills are going to change. And maybe that workflow is entirely different again in 48 months. So I'm less likely if I'm the CEO of a of a big company to hire for a that role as full-time. I'm going to opt much more for a shorter-term contract, a year, 18 months. So when you think about that times a lot of the jobs in Well, we can you know, focus on the knowledge economy for now. Um we'll start to see the rise of much more independent based work in in the workforce. And that's kind of new for a lot of people. We do see the rise of much more I mean, you and I are both technically independent workers as it is. We do our own thing. Uh but the dominant form of labor may start to look more like that. Where instead of working for one company doing one thing, you work for a few companies uh doing the thing the skills that you are endowed with and the skills that you have um in the field that you work in. And that's a very different type of workforce because it means you become your own CEO, right? You are an organization of one. And you apply your skills to a variety of different companies or projects. And so that's I'd say the rise of the independent era. And it's a very different future. Uh it has many implications for things like health insurance and social security. Not everybody wants to be an independent worker. There is a lot of comfort for some people. Some people it's their nightmare, but for others being able to have consistency 9:00 to 5:00, but we should start to see the idea of a 9:00 to 5:00 job for one company will be a chapter in human history. And that chapter is closing for sure. That could I mean that could bring a lot of anxiety for for some
of anxiety.
for for people of like I mean I know what it's like to like run run a run a company and be an entrepreneur. I mean we've been doing this for like 8 years, but there's a lot of stuff that that goes into that. Um and so it I wouldn't say it it it does sort of require a particular type of person or at least a particular type of mindset to be able to endure because there's so much uncertainty. There's so much like like if you're working for five different companies and you're a freelancer and one of those things drops, then 20% of your income like it it goes out the window versus I'm they might be bored at work, but the safety net of uh I have like $80,000 a year coming in.
You can plan. Right? It's it becomes impossible to not know what is my income next year or the year after, but remember um things the variables don't change in isolation. So when the fabric of the workforce starts to shift, new business models, new types of platforms and infrastructure start to come into play that allow that to make more sense. So, right now what we're probably doing is extrapolating, okay, everything stays the same. The only thing that changes about the world around me is that I now hold three different jobs with three different companies, and there are these massive gaps between new projects that I'm going to work on and existing ones fading away. Other variables will will start to change. But, yeah, I mean, it is a very unsettling future for many that take that enjoy the security of a stable job. Uh and I mean, it is something that I do try to call attention to on my platforms. If we can see this transition happening, uh something like social security and something like healthcare is going to become really important for people. And even just having the ability to reach up and grab new skills or to be able to continuously pivot, um that's an entirely different market, but it's one that we're starting to move towards. So, my advice for people in this moment is don't think of your job in terms of the title that you hold. Think of the skills that you have under it. So, be industry agnostic, be job title agnostic. What are the types of skills that you're performing? Right? Okay, so I exercise judgment when I do this. I use creative intelligence because I'm the person that always comes up with the ideas. Whatever it is that that are the kind of skills that you that you occupy, that you use every single day, those are what you'll continue to apply uh it's just in different ways. And so, maybe you'll be directing AI systems to do those tasks, but applying the expertise that you have in the role that you've, you know, been maybe holding for the last 5 years, and that's how you start to look at it. So, it's going to change, of course. I mean, and work has always changed. Even though even the idea that a lot of people work from home, like that was so radical even a decade ago, that a significant portion of people would opt for virtual work or at least partial virtual work. So, I think we can adjust to it, uh but do we have the infrastructure in play, the policy support, um the entire different again I've mentioned social security cuz I think it's really important and we don't have those safety nets. But yeah, thinking of yourself as an organization that offers a bundle of skills to a variety of different projects, that's that's one way to think about the future and to think about skills over thinking about job titles, much more important. Now, I don't know if that in terms of thinking um about skills invalidates this entire question, but the the idea being a synthesizer um versus and like a generalist versus like a specialist in in one sort of area. How would you think about that? Cuz like for example, if you have been working for the past 10 or 15 years and you've been doing this one sort of thing, right? Let's just say marketing. You're doing you're doing email marketing um and there's someone who might be a generalist. They might not be in like an expert in like email marketing, but they might be doing seven different things across domains and marketing. So, they have like a way uh wider expertise and also in some ways I would I would probably say their risk is lower because they can pipe in and out of different things because they are like general and synthesizer versus everything isn't relying on them being an email marketing and them going, "Oh, like I don't know how to do anything else." Do you feel like that aspect of the conversation is going to be relevant with
today
Mhm, like who who's better positioned? The experts and the people with domain expertise or the generalists? I don't think that we've netted out on on on if there's if who wins what where because there's so let's take that example. You're working in email marketing and I think we could probably safely say that is something that AI is already doing. Um so, going forward that's you know, if it hasn't already impacted your work, it's probably going to. Okay, so does that mean that the journalist comes in and scoops up all of those email marketing jobs? Hmm. If you look at the skills under email marketing, why is it that somebody clicked your emails over the other companies? And how did you think about adjusting those campaigns when it wasn't working? Right, what was the psychology that went into how you structured those titles and where you placed images that allowed somebody to respond differently? How does that type of thinking happen in a world where AI is marketing the product and you're giving AI the framing of your brand, the culture of the company, whatever it is that you're selling. So, what were the underlying skills that went into you structuring that? And that's why it's so foreign for us to think about how we were thinking or how we are thinking. We just don't think about it that way. We just do the task. But what was the thinking that happened before that task and how you evaluated whether that was successful? That is what you are taking with you to the next thing.
And now you're saying that's what's going to be valuable to a company potentially.
under it. Yeah, so maybe you're not writing emails anymore cuz AI writes it better, but you are bringing the skill set in that people don't like to hear the number first. When you start with the price, even though it's an amazing sale, they don't want to hear it. Like you're still bringing in those types of things. And this is why or if we can you'll have better ideas perhaps if you're thinking about it that way than somebody that's just never in their life dealt with marketing. They can do some stuff, but you could outcompete them in certain areas. Like you're thinking about how you're thinking. And that's of course not just like a universal thing that's going to apply across the board, but it can apply in many cases. So, does that mean that someone wouldn't have to know email marketing at all and that like because of AI and that is that it's helping with that someone would be able to pipe into temporarily doing email marketing because they carry a set of skills? Wait, so I'm not sure I understand what you mean. So, if someone's carrying the set of skills across like domains and across uh tasks and they're working with AI to accomplish those things, does that mean that their ability to know the ins and outs of email marketing is less relevant? Okay, so I love that you asked that because what you were essentially saying is you could have skills in one area and they actually become more applicable in another area. So, the person that had amazing judgment skills, let's say in finance or human resources, may actually be the best person to run marketing in the AI-first world because the types of skills that they were executing may actually work well in the types of decisions that they're making where AI is the dominant platform. So, yes, it is possible that somebody that the skills and this is the you know, to quote one of my previous professors, the skills that made you dominant before AI may not be the same skills after AI, right? So, the email marketing person, based on the types of decisions that they were making, if they can think about how they were thinking, they may be the best person to make judgment calls on hiring. Right? So, we're going to start to do different things. So, that's why it's so much more important to think about the underlying skills of how you made decisions, for not just the simple task that you did, but how did you make the decision before that task? And we're not taught to think about our thinking on that meta level, but it's really important. And that also expands the type of work you think you're capable of because you might be limiting yourself to email marketing, but you could actually be the best person to make judgment calls on finance decisions because the AI is crunching all the numbers. You're making the judgment call. Do we go with market X or market Y? And market X is these types of people and this type of this type of group is very sensitive to price every time we engage with them on email, but this type of group never cared about the emails about price and we're trying to target luxury. Let's go after them. And that could be the email marketer. Wow. Yeah, [snorts] I mean, I think it it in terms of further cuz we said in the beginning of the the conversation the idea of envisioning a future world that that doesn't exist, it almost feels like Chinese to a lot of people. So, if we were further paint that world, you you've talked about like career ladders shrinking, um, and that the shelf life of skills will shorten. So, the very idea of like a career might evolve. And so, like what is that What does that mean exactly? Yeah, so I think we have spent the last few decades building up the idea of work where you learn, you work, and then you retire. And that worked for a certain type of economy where skills and tasks were cumulative and pretty predictable. But in an era where AI and different technologies will continue to change the types of skills that we're going to need to bring to the table and how different types of work gets done, in which types of products are interesting and our buying behavior, that consistency of working vertically up a ladder starts to make less sense. Because in in an era of skills over, say, job titles and sometimes even over experience, it doesn't necessarily matter if someone has worked 5 years or 15, if the person who's worked 5 years is continuing they can learn different skills or how they apply their skills, um, is more advantageous. So, that career ladder starts to make less sense and we're already starting to see some of it not entirely get pulled out, but junior hires not necessarily making us not We're seeing lower numbers of junior hires. Um, and that I think is actually going to be temporary. I think we're going to start to see them funnel into different types of roles like directing AI systems and the AI agents. Um, and then that becomes And that's actually quite interesting, right? If you take a legal firm, a partner is probably best to make the final judgment calls like, "Okay, we've been in this type of court scenario before. This is how this judge behaves." But the junior or the younger person that now is no longer on that paralegal ladder or whatever came first, but they're on the AI agent director ladder, they're I don't know who's they're pretty equally important in that firm because that legal company can no longer keep up if they aren't diffusing agents the way their competitors are. So, that idea of the autonomy ladder and all of that that starts to make less and less sense in a skills over job title era. What is like a resume like if you apply for a position, what is a resume even Oh my goodness. What is resumes for who? Written by AI for AI. Resumes I don't We're going to have to figure out different ways for how we signal our skills because now it is a complete it's complete deception. Nobody's real reading it and writing it. It's me much of these resumes are optimized by AI so we're going to have to show different ways a lot more real-time tasks tests of of getting work done. But resumes also I mean they are a reflection of usually a hierarchy of experience. And so in an era of skills, it's a little bit different, right? It's I just need this outcome done and [clears throat] do you have the skills to do it and it it's probably shorter term so the investment is a little bit versus if you're hiring someone and you're like I'm I'm hoping to bring you up the career ladder we're making an investment in you versus you come with your bundle of skills that probably also going to change hiring dynamics but we're again in this strange in between times. We're hiring I don't know how people are making decisions right now. Who wrote that? What wrote that? Which model? And who's reading that What's reading that? Hiring is kind a little bit of a nightmare right now. Mhm. Yeah, I mean even even if you get a letter from from someone like that wants to work at your company or something. I mean you know, it's impossible to know like does that actually which I think you've talked about one of the skills which I want to get into is like if you're using AI then you have to be able to explain and and articulate yourself based on whatever it put out there. So, it's like an extension of you in some sort of way. Can you take me through that? Yeah, I mean, um being able to explain the reasoning behind why you made decisions and what those judgment calls led to, that's a skill that's interesting to me. So, if I if somebody did send me a resume uh and I'm reading the resume and I'm reading their cover letter and then they'll say, "Yeah, I'm very, you know, AI savvy. I've built a bunch of agents." That's also interesting, too. I think people that can uh build different build different agent workflows or or optimize AI in certain ways that are helpful to complete tasks more quickly, that's an a new area I'd pay attention to, especially if I'm younger out of college. Uh but if someone was to say, "I use AI in X, Y, and Z ways, and when it performs this task and gives me this output, here is the the tradeoff tree that I used to make decisions for my company about whether this is acceptable or whether it's not. And when I built this, this led to this type of performance as a result. Or we increased gains by X or by Y." And always use measurable numbers on your resume. then that's something that's interesting to me. So, people who can explain and show that they have judgment skills and they have deep problem-solving skills. And that's also another thing. If you were applying for a job and you were applying for a job in something that seems unrelated, but I could see that you could solve problems. You're explaining the problems that you solved in your company and at your company and how you did it, that's a really important skill. It doesn't actually matter that they're logistical problems and you're trying to now move into to marketing or something else. These are type the types of skills or that are very transferable. And again, it's means taking a step back and thinking, "How do I solve problems? What skills am I even using every day?" Yeah, because I think at some level I'd imagine some people have shame or they don't want to reveal in some way that they are using AI to do certain things cuz they don't want to seem inauthentic. I do imagine a world where like everybody's assuming that everybody is using AI and it's out in the open and so that sort of shame mentality starts to go away. Eventually, it's it's also really funny because being in these in-between times, we hope that employees and people are building up AI skills. But then for some employers, and I had a there was a scenario this week, if you were to tell them, "Oh, I used AI for this," they would still maybe be upset. Like, "Oh, use it for everything else, but except for the work you do with me." But I think eventually, we will by eventually, probably in the next year. I mean, I also know employees on the other employers on the other side where they already expect you to come to the table with AI skills. It's just Please let that be the default. You should be using it in everything you do. Um but eventually, it will just be like the computer. I don't know the last time somebody was in a job interview and it was like, "Can you operate this?" Or somebody was, you know, I proficient in PowerPoint. I mean, we're just not doing those things anymore. And AI will be the same. It's not going to be a tool that people grab. It's going to be just the thing that we're on. Yeah, I think it's the language thing that you said earlier. You were talking about how much language is associated with consciousness and then like human intelligence. And so I think the difference of using a computer is like if you're using AI to communicate with this person over there, they might feel like they're talking to a robot and not actually talking to you. And I guess that's where like the ownership of the insides of whatever you're putting in front of them becomes an authentic extension of you versus it's just something that like ChatGPT spit out and then I gave it to you. Yeah, if you're continually if you're using AI in very generic ways, um and then that's where the story ends. And then that's the text, that's the love note, that's the propo- uh I don't think it's going to work out well. Maybe right now you have some arbitrage opportunity in a sense amazing that somebody's now magically a poet. But in time we're going to start to expect that AI is the default interface and it played a role in drafting and so where you is uniquely you in that I think will be important. But yes, eventually and that is also a concern right on the flip side. What does it mean when AI is always between you and I? Always. Right the way your iPad is right now. What does it mean for that iPad to be talking back? Yeah. In this moment, listening, observing, interpreting, studying us, extracting data from us. That's what's also coming. And that's really interesting. For there there's really scary parts of that and interesting parts of that. But yeah, it will it will change how humans relate to each other. We want to be steering that in a way that we don't continue to have end up in environment which more and more challenging for us to relate to each other. But yeah, what does it mean for AI to be between most people? Yeah, because there's like a there's a processing of information that happens very quickly. And and you could you could relate to this even to romantic relationships as So let's just say you're in a fight with your partner and then rather than saying something that's extremely emotionally unhealthy, utilize AI to like come up with a response. But at some level that's probably better than the explosion and saying something that's very unhealthy, but unless you've like internalized the ideas that AI came up with in your relationship of why that is a better response and then trying to embody that in some way, then it's just like a piece of paper that that you're sending. It's not really you and I guess in a sense. Yeah, and that's some of the studies that have alluded to cognitive atrophy or this cognitive offloading. Some of the experiments that they've done to reach those conclusions show that when students use an essay or use AI to write an essay, 5 days later they didn't even know what was in the essay. They didn't feel like they could take ownership for those words. Like kind of can kind of the idea was mine, but I don't know what actually happened between the first and second paragraph. Imagine that, right, in relationships. Like, "Well, this is what you told me and this is why you said we would never experience this again." Did I? Oh, what I didn't know that that was exactly in those words.
like a conduit of information that went from AI through you to someone else instead of like something that Yeah.
you sat with.
And that's going to I mean, and there's many different ways to look at this. On the one hand, we could put some of the responsibility onto AI companies designing systems that don't ever push back on us, that don't ever ask clarifying questions to make sure that that thing resonates with us. Or if we're in a very heightened emotional state, you know what? I I'm an AI, don't forget that. I think you should go seek a friend or somebody that's actually, you know, living with a heartbeat. Um those are design decisions, and then there's also the AI literacy component of that. And having agency, how much are you offloading everything? And And would you feel good taking everything to Reddit and getting feedback on it? Maybe in some scenarios, but maybe it's not always appropriate. And then, of course, there will be this change where a new technology is changing. I mean, if we were to actually map out how text has text messages have impacted the relationships people have, that text has gotten shorter, brief, but continuous. That's probably had a profound impact on relationships, and we just haven't really thought about it or haven't really measured it.
It's true. And we Some people seem okay. I don't know how everyone's doing, but, you know, some of us have made it on the other side of text. So, then there's also the scenarios where we can actually see and maybe it does strengthen. I mean, for some people long-distance relationships have become better because of FaceTime and you're not actually fully separated all the time. Maybe there is a world, and there actually is a world, right, that could strengthen our relationships. But again, depends on how it's designed, how we use it, and all of that is still design space. Mhm. So, I mean, that's the ability to understand and explain the ideas that are coming from AI. In terms of like a non-negotiable becoming AI literate to to begin with, does that really start with knowing how to prompt? How would you coach people through becoming AI literate in this kind of new world? Yeah, um I would say it starts with probably knowing what AI is and what AI isn't. So, AI isn't your friend. It's not a system that has your best interest. Uh it's a system that makes really strong predictions with words. Uh and that is exceptionally helpful in many scenarios. Uh but that's what AI is. It isn't some oracle. Um it doesn't have all of the answers. It doesn't have an It shouldn't have this aura of authority. Uh and then knowing that AI is only going to be as helpful as the thinking you've done and the questions that you're asking it. So, what you bring AI, it's going to meet you at that bar. Uh and the output is going to meet you. It's not going to magically raise the bar on what you suddenly bring to the table. So, the thinking you do before AI still really matters. Um and then, yes, you can start to think about the idea of prompting. So, um am I giving AI a role? If you're asking it to help you write a stronger essay or newsletter or op-ed, uh giving it, you know, you are an editor at, you know, publication that you're trying you're aspiring to write for, uh and giving it some boundary, giving it some identity that it can inhabit as it's trying to help you with the problem that you're solving. So, there are some nuances to prompting. Prompting's going to be this evolving um conveyor belt, though, as AI becomes more integrated into our world. We're not going to necessarily need to give it all this context ahead of time. And then, like, it's just going to be a bit more continuous. So, the prompting itself will change. But, yes, the the thinking that goes into using AI really matters. Um it is absolutely not magic. It has nobody's in best interest. It has no interest at all. Um and to still know that your time is valuable. So, spending just unlimited time engaging with an AI system has an opportunity cost. You're you're spending that time not doing something else. So, to also remember that, right? This is a technology people building it have certain incentives. Um so, to still have ownership over your time. What are the cognitive things that you feel like people should not be giving up to AI, and then how do you retain those things, and then bring them bring that to the table when you're using AI? The cognitive things people shouldn't offload to AI. Uh I think your judgment. So, AI shouldn't be making final judgment calls. if the if the if the question is help me with this email, okay, fine. I'll offload that if it's not really an important email. But if the person on the other side of that email and is a partner and you're at the edge of a relationship, don't offload the the drafting of that email to AI. Maybe make sure, you know, do I come off not narcissistic? Do I come off Sure, you can but don't offload that judgment call to AI and an AI system. Um if you are making a big life decision about something, retain your judgment, retain your discernment of of making that problem on your own. Um I think we shouldn't offload any of that. I think problem-solving, you still need to know how to how to solve problems, and that is a really employable skill in the future. And if you offload that, uh that's going to have economic impacts, uh and then also behavioral and your ability to navigate really relational impacts, your communication skills, I think we shouldn't offload. And that one seems paradoxical, right? Because if you're going to use AI, you're not inherently [clears throat] communicating. But how how communicate to the AI also actually matters. The more you can articulate your problem or the thing that you want it to do, the better it's going to work for you. Uh and the the more you're going to need to learn how to work with people. So, if we're in this system in this world where AIs are everywhere and we're streaming them, the people we choose to build our teams with, that's going to I'm going to choose the person that is enjoyable to work with and can share their ideas with me because now we have all of these kind of R2-D2s around us. So, those personal emotional intelligence skills also still matter. yeah, I think that if those are those are some of the cognitive skills, your judgment, your discernment, your problem-solving, critical thinking, emotional intelligence, communication, and yeah, I mean, you can think of AIs in some ways maybe as a bit of an extension of us the same way if you're playing to golf and you and you left your smartphone in the lounge, it would be panic uh panic at the disco. It will probably eventually be like that with AI. Um but yeah, this this isn't a system that's alive, it's a system to support. So, I did but the would you do you consider those to be the skills that then would transcend fields essentially, judgment, communication, or are there other ones that you would an understanding of an ability to articulate ideas? Like are those the ones that are going to transcend or would you is there additional ones that you would place on top of that as you kind of
So, those ones would be foundationally really important for the future. If you cannot make any judgment calls and I mean, if we're going to be moving towards a world where we start to direct AI systems, we do that with judgment. What do you ask it and was the answer that it gave you good and sufficient? So, I think that's really important um contextual awareness is important. Having some understanding of the data how AI is largely a reflection of the data that it's trained on. So, if it wasn't trained on data that's going to encompass anything that you're going to need, it's probably not going to work well for you. Uh Uh so I think that that's probably a skill that that matters, some aspects of of data literacy, and then domain-specific skills still matter. So sometimes you hear people say it's totally irrelevant. Not entirely true, right? Somebody with domain expertise in biology is going to do much better at applying AI in the synthetic biology company of the future um than someone that's maybe walking in with expertise in geography. And somebody that's applying AI to geospatial intelligence, and they've worked in space or those domain expertise I think are still very relevant.
is still relevant then, in a sense. School so yeah.
Education, school, and the ability to even build up a domain of knowledge of these specific domains and fields.
Yeah, if you're working in a field where domain expertise are are important, then yeah, I think school that's still really relevant. I mean, if you're going to work in robotics or you have aspirations to work in gene editing, and going to school and get building those domain expertise I think is still very valuable. Education, when we zoom out though, what skills is it giving people right now? I think we do need a probably a mass updating of that and to be optimizing for skills over memorization and and and word recall and that sort of thing. But theoretically, if we can do that, then I think education I mean, and and school isn't just for those types of skills, right? You you need to learn how to interact with people in the world. You need to learn how to how to work with groups, how to meet people that are different from you. Um school is also important for for those reasons as well. But we do need some updating to [clears throat] curriculums. I remember like 10 years ago I was working at a teaching class at this boot camp called General Assembly in uh in New York here. And at that point, I mean, it was for marketing. And and so you we actually had people that were like in their 50s coming into the course and they were off the heels of like print marketing and like magazines and stuff like that. So the age of social media was like this huge adjustment. And it felt like in some ways they they felt like they were like behind the curve of like they settled into like these this is the way the marketing world works and now I have to have this huge adjustment. And that was over like a many year period that they found themselves in that position but it feels like with the coming age of AI people are going to find themselves in that position in like one two years of of constantly having to learn new things which could feel overwhelming for people in some ways but it seems like it's going to be like the baseline of like the new of constantly learning you Yeah, yeah, yeah. That and those would be other skills that we could add adaptability learning how to learn most of us don't know how do we learn the same way most of us don't step back and think what were the skills I just applied to make that decision. So learning how to learn and being able to adapt to new situations new context is really important but then you also have to remember once those become the foundational skills that we have and that we've built it becomes easier to engage in those pivots and to be in a workforce or world where we do have to continue to learn new things over time. So starting from where we are right now it's really overwhelming but eventually the same way most people are quite fluid with using a computer and now we do all of these different things in the computer than we did in 2001. Eventually when you're you're more adept at adaptability and you're more resilient to it it stop stops feeling like such a big radical change. So if you fast forward five years and someone embrace these principles of like the independent era and what's coming uh how how do you think that they'll not only survive but maybe thrive through like the vol- of what's coming. Yeah, I mean I think I mean if you have adaptability skills, that's pretty key, right? So how well are you adapting to this moment right now? And if you can work on that, um that's going to give you an element of resilience and allow you to spot new opportunities. If you've built skills like judgment and you built skills like continuous learning, there are a lot of opportunities that are open that are up for grabs right now. I mean most companies are trying to figure out how do I do a lot of this with AI agents. And if you have the domain expertise of marketing or of finance or of HR and you're able to help you structure those workflows, that's very very valuable. Being able to have some flexibility with your time, I mean if we can get this right, um a world where we have a little bit more flexibility and freedom could work out well. Again, that's going to that's a big F and that's not just going to be an inevitability. Um and I think
[sighs and gasps]
I mean the second part of your question, how does this end up for them in 5 years? Yeah, how do you think that that person will navigate the coming age versus someone who was resistant or didn't put these things into practice? I think one person will be building the skills to participate in the job openings and and not even just the job openings of the future, right? There's a lot of ways that these tools can help you in your own life. These technologies can help you in what you're doing in your day, on your weekends and optimize the things that you're doing, give you time back and you might be missing that if you're just saying I'm not doing any of this. and I mean I just because a few people say that building one version of the future, it doesn't mean that that's going to happen. And so the person who is engaging with this technology, who's spotting the gaps, I mean and one thing we didn't talk about a lot of the companies that we see today are also not guarantees in the future. Yeah. Most of the companies we think of in our day-to-day lives were born of the internet era. We are in the next chapter. Who's building what's coming? Those are all up for grabs, too. So, this is a really interesting moment. And if we can have different voices close some of those gaps and reach for some of those opportunity that that's going to be really really interesting. That's an important point. Yeah, I mean, a lot of these companies, Apple, Meta, I mean, they're even having to adapt to this coming age. I mean, they're thinking about the smartphone being a predominant fixture of society in the past 17, 18 years. And possibilities being floated in the future of like, is the smartphone going to be the predominant thing that we're carrying around?
Absolutely not. Yeah, I mean, so those are those are conversations that are happening and and like Apple has been the dominant Samsung and some of these other players have been the dominant people. So, I think that's interesting to think
If you want to talk about inevitability, no company that exists today has is inevitably going to be a part of the future when it's AI first. Not one of them. Mhm. And most companies know that and that's why we're seeing a lot of panic buttons and a lot of different code reds and all of these things. Everybody's adapting to this moment. Of course, there's different resources, different power, different capital. Um but by and large, there's no future is guaranteed for any of these companies, either. Uh so, I think those are pretty interesting opportunities and design spaces and different business models waiting to either be disrupted or or to be seized, and that's really cool. And I think I mean, yeah, there is a world and it's not a far away world where we get this right, where people are protected, where we can control AI in ways that um we want it in to be in control of it. And this can go well. Um there's of course ways where it doesn't go well. Uh, but neither of those scenarios are inevitable. And the decisions, you mean the future hasn't happened yet. We're making the decisions about it today. And my last my last question is, you know, if you were to go back to the beginning of the conversation and the subject Sarah that we talked about, um, you know, the one who started to believe that AI is overhyped or confused about this moment, what do you feel like someone like her most needs to understand about the moment that we're living in? Sarah, what is what does she need to understand about the moment that we're living in?
That we are in an industrial revolution moment. Uh, and Sarah can reflect on how much things did change. Timelines are always a bit wonky. Um, what comes out on the other side of that is always a bit gray. You can see some of it, you can't see all of it. Um, but how we live will change over time and that is also not new. I mean, a lot of the jobs that we hear about and the companies didn't even exist 25 years ago. Um, so expect life to be quite different in some ways, better in some, just strange in others. And I think the more you can lean into the future and the more you can start to try to take part in your own life. I mean, you're the ex- you're the expert of your own life in the domain that you work in, uh, the more in control you're going to be of this moment. Uh, and you can probably pay less attention to the headlines uh, and more attention to this technology's impact on your own day-to-day life and your own day-to-day workflow. Beautiful. Well, thank you so much for being here today.
me. Yeah.
