The fourth humiliation

And 10 other speculations for 2025

Dear Friends,

We’re a quarter-century into the new millennium! 2025 feels different from 2015 in a way that 2015 didn’t feel radically different from 2005, or even 1985.

It’s less about the slowdown of globalization and the rise of nationalist populism — though there is that. It’s more about the phones, what they’ve done to dating, friendship, our attention, how we work, construct an identity, and seek status. It’s a transformation best described by social theorists, not economists or political scientists.

And why wouldn’t the 21st century be as radically different from the 20th as the 20th was from the 19th, when the world’s population was 1 billion, average life expectancy was under 40, half the world lived under colonial rule, and over 80% worked on farms?

For this week’s newsletter, I sketch out ten broad reflections about where the world may be headed that I’d like to revisit in twenty years. Already, I can see what I got wrong in 2022. We’ll see if this year’s speculation fares any better …

1. The Fourth Humiliation

A professor in college talked about humanity’s three humiliations. Before Copernicus, humans thought we were at the center of the universe, literally. Turned out we were the ones revolving around the sun, and ours was just one of a septillion stars. Darwin revealed the second humiliation: we weren’t God’s creation, but descendents of monkeys. And Freud gave us the third: our thoughts control us, not the other way around.

So we’re just monkeys with opposable thumbs on a random planet who can’t control our own thoughts? But at least we have sophisticated language and intelligence! We alone can reason.

No longer. With OpenAI’s latest reasoning model, o3, scholars will have to redefine concepts of “intelligence” and “reasoning” to keep them human-only.

AI surpassed the Turing Test long ago. So researchers developed a new test that humans can pass with our unique reasoning capabilities but predictive language models couldn’t solve. When they created the test in 2019, the best model scored 20%. At the start of 2024, the leading model scored 34%. The creators considered anything above 85% to represent AGI. Now, a high-compute configuration of OpenAI’s latest model scored 87.5%.

The same day that OpenAI released its o3 model, the Wall Street Journal published a major article, "The Next AI Breakthrough Proves Elusive.” And Bloomberg ran the headline, “Why AI is Facing Diminishing Returns.” I have a feeling 2024’s mainstream media coverage of AI will look very mistaken in retrospect.¹

I expect that AI will develop at the same breakneck pace in 2025 — beating benchmark after benchmark — and then even faster in 2027.²

I didn’t experience the invention of electrification, so I don’t know if it was under-hyped or over-hyped relative to its eventual impact, but I imagine it was under-hyped. I agree with Eric Schmidt: maybe 1% of the population overhypes AI, but 99% are under-hyping.

As far as forecasts go, I’m looking forward to revisiting the bet between Gary Marcus and Miles Brundage on “Where will AI be at the end of 2027?” And by 2031, it will be time to revisit Dario Amodei’s optimistic take on “How AI Could Transform the World for the Better.” Zooming out, here’s a great timeline of AI forecasts from now through 2050. And here’s my nerdy forecast for the next year:

2. Opening The Black Box of AI & Ozempic

We don’t understand two of the most transformative innovations of the past five years. We know that LLMs and GLP-1s work, but we don’t know how. Finally, research is catching up and we should learn much more in 2025.³

3. Redefining health, medication, and impulse control

There’s something in the air about personal responsibility, health, and impulse control.

The past ten years were all about systemic injustice. The thinking went: it’s not whether you are individually racist; the important thing is to recognize systemic racism. Similarly, obesity wasn’t the result of any individual’s choices, but a system of food deserts, corporate marketing, and so on.

Then Ozempic came out and obesity levels dropped for the first time. For some people, this was proof that obesity isn’t a consequence of their decisions but a medical condition they were born with. For RFK Jr. and his MAHA followers, Ozempic is one more example of trying to solve the human condition with pharmaceutical solutions.⁴ The MAHA movement worries about a near-term future where we depend on Ketamine, Ozempic, and Adderall to stay happy, slender, and focused.⁵ MAHA prefers the harder, more natural path: therapy, exercise, and meditation. I admit, so do I. But who am I to tell someone else to go for a run instead of take an shot of Ozempic?⁶

On the one hand, you have people diagnosed with ADHD / obesity / anxiety / depression who benefit from medication. And then there is a growing wellness movement skeptical of “medical science” they say is disguised as pharmaceutical profit motive. Neither side strikes me as totally right nor wrong, and I imagine that most of us are reasonably somewhere in the middle.⁷

The increase in depression and ADHD diagnoses over time

4. Can Trump hold together the impossible coalition?

Elon Musk versus Laura Loomer was just the beginning. Somehow, Trump managed to assemble an impossible coalition of Plumber Joe, MAHA moms, the contrarian tech elite, career Republicans, and rap’s A-list. It is bound to fall apart. Trump can’t give tech lower interest rates without harming the working class with inflation. He can’t please the MAHA movement while smiling with McDonald’s. He can’t stay credible on immigration while big tech hires largely from abroad to stay competitive with China. And the inevitable racial profiling of deportation and increased policing will alienate Latinos and Blacks.

It’s gonna be a great year for Jon Stewart. Maybe Democrats will even become funny again.

5. How will the counter-counter-elite take shape?

It’s been a while since anyone was like, “I love elites!” But you could see the snowballing discontent with leftist Ivy Leaguers’ self-serving hypocrisy since at least 2012. That’s when Chris Hayes published Twilight of the Elites to argue that Ivy League graduates bend the rules to protect their advantages while claiming to be evidence-based and pro-poor. I experienced it myself: surrounded in San Francisco by rich, progressive Stanford and Harvard grads who spend a million dollars on their kids private schools, extra-curriculars, and SAT prep.⁸

Rob Henderson

coined “luxury beliefs” to describe ideas and values that confer status on the wealthy or educated elite while imposing costs on the less privileged. His first example: “Rejecting meritocracy while benefitting from elite institutions.”⁹

This time, the fed-up electorate voted for plutocracy over hypocrisy. So here we are with our billionaire president and his mega-billionaire advisor. How will the opposition take shape? Who from the left has the credibility to build a counter-counter-elite? Certainly not Gavin Newsom. Probably not Kamala Harris or AOC. Maybe Gretchen Whitmer or Pete Buttigieg? I dunno. It’s hard for me to imagine anyone organizing a united opposition. Hopefully, someone surprises me.

Trump’s coalition is sure to fall apart. But can Democrats rebuild their own?¹⁰

And lastly, can DEI successfully rebrand to fight back against the backlash? I sense widespread support for diversity, inclusion, and equal opportunity; but without the shaming, canceling, and virtue signaling. Can we get past the zero-sum thinking, as described by

Richard V Reeves

below, that the well-being of one group must come at the expense of another?

6. Tech regulation and social trust

It’s incredible that just 18 months ago, Elon Musk and over 1,000 tech leaders and researchers signed an open letter urging artificial intelligence labs to pause development.¹¹

For a while, it seemed JD Vance would lead the Trump administration to crack down on Big Tech, and especially social media’s harmful effects on teens. But then Big Tech came to support JD Vance and now it seems like we’ll see total deregulation of crypto, AI, and social media. I expect we’ll see more scams, fraud, misinformation, and conspiratorial thinking.¹²

Everything seems possible, but nothing believable. Consider the rapidfire rumors after the LA fires, the New Orleans terror attack, and the New Jersey drone sightings. No, the conspiracy theories don’t inspire trust, but neither do the bumbling responses by the government.

Truthiness” isn’t a new diagnosis, but the effect of generative AI and social media at scale¹³ increasingly means that no one trusts anything or anyone.¹⁴

7. The year of robots

For the past decade, self-driving cars have been “just a year or two away.” Now they’re finally here with Waymo going mainstream in San Francisco and expanding to LA, Miami, Austin and Atlanta. Lyft will try to compete with the Uber+Waymo in Atlanta through a new partnership with May Mobility. And Tesla says they will launch their Cybercabs in California and Texas this year while Amazon launches Zoox.¹⁵

Robotaxis offer convenience for relatively wealthy people like me and the readers of this newsletter. But will it make the larger transportation system any more efficient?¹⁶

Meanwhile, Tesla's Optimus demonstrations are moving from awkward shuffling to practical tasks.¹⁷ But the real transformation is happening in commercial kitchens and warehouses, where companies like Figure and Agility Robotics are deploying robots that can handle unstructured environments. Chick-fil-A’s lemon-squeezing robots are saving 10,000 hours of work and Chipotle added the Autocado, a robot that peels and cores avocados for guacamole. In just seven years, the global robot density in factories has doubled. I expect it to double again by the end of 2026.

Nvidia, now the world’s second most valuable company, is going all in on robots and OpenAI is hiring a new team dedicated to robot hardware.¹⁸

I don’t expect many of us will buy consumer robots for our homes like the new SwitchBot anytime soon. But I expect that most warehouses will soon convert to what we see below from sports nutrition supplier, The Feed.

8. Longevity research goes mainstream

While old age is not new, few people lived to see it until last century. In 1900, the average life expectancy of a newborn was 32 years. By 2021 this had more than doubled to 71 years, essentially giving humanity “an extra life.”

Andrew Scott says “the first longevity revolution” resulted from modern medicine reducing disease so that more of us can die from natural causes.¹⁹ His new book argues that we’re at the start of “the second longevity revolution” to address the causes of aging itself. The (kinda cheesy and dense) documentary Longevity Hackers provides an overview of the current state of research on everything from Rapamycin and Metformin to Senolytics and Nanobots:

As more billionaires fund this kind of longevity research, I expect more debate in 2025 about whether aging is a technical problem to be solved versus a natural process to be accepted.²⁰ We’ll see more longevity apps and services (like NOVOS) launch in 2025 to track and lower our so-called biological. Plus, it will be a big year for research studies on the effectiveness of Alzheimer’s drugs.

9. How likely is WWIII?

The talk of a possible WWIII strikes me as overblown, but not impossible.

Niall Ferguson

, an established Cassandra, sketches out a convincing scenario of how it could happen.

Noah Smith

adds his own. I’m skeptical; the world is much more connected and economically interdependent than during the Cold War. Still, I have unexpectedly come around to the possibility that the US may need to embrace a few years of “peace through strength” to get us through AGI, drones, and robots before rebuilding the grand project of a rules-based, multipolar, international order.²¹

10. Could the AI hype market sustain?

I don’t think my generation appreciates how incredible the stock market has performed over our lifetime. Hardly anyone expected last year’s incredible performance.

The world’s 500 richest people reached a combined net worth of $10 trillion, or more than five times the GDP of Brazil (and its more than 200 million people). 🤯

Could the S&P 500 possibly continue to grow in 2025?²²

I don’t know what will happen to the stock market, bond yields, or interest rates in 2025, but zooming out, I expect labor productivity growth to be closer to 3-4% and total factor productivity above 2%. (Unless that is, we see WWIII, disastrous climate impact, or a long-term resistance to migration.)

11. Glasses and microphones

Finally, 2025 could be the transition away from the cell phone as the main way to interact with computers and toward something closer to what we saw in Her.

We can expect many more smart glasses to compete with Meta’s RayBan glasses this year. I, for one, don’t want screens in front of my eyes all day, even if they were to provide me with useful information. But, I am intrigued by the information AI can provide us via cameras and microphones. (The live translation feature from Meta’s Ray-Ban smart glasses is undeniably cool.)

Even weirder than walking around with a computer on your face is the new slate of wearable, persistent microphones that record everything throughout your day and then provide you with analysis of what you said, what you were told, and what you might want to remember from the past. My first reaction: That sounds like a dystopian nightmare from Black Mirror. But also: Wouldn’t it be great to never forget the important stuff anymore? And to have an AI coach help you communicate better? I dunno, I can imagine wanting one of those, though feeling terrible about it.

What am I getting wrong?

What do you think? What am I getting wrong? What did I miss?

2025 is already off to a wild, and at times heartbreaking, start. Let’s hope it gets better.

Yours,

David

1. I’m grateful to

Zvi Mowshowitz

for his Herculean effort to keep readers up to date on the latest AI developments, especially about research. The AI Breakfast newsletter is a nice complement. Let me know if you have other recommendations.

2. This will be the year that AI will outperform whatever it is you do, writes Tyler Cowen, who says OpenAI’s latest model is already a better economist than he or his grad students could hope to become. AI capabilities will continue to advance more quickly than our ability to incorporate its usefulness into our lives. And maybe that explains why so many people treat AI as just the latest app instead of paradigm shift.

3. I’m tracking the research about GLP-1s mentioned by Ezekiel Emanuel in this fascinating Freakonomics episode. And for AI, I’m following the work of Chris Olah at Anthropic, who does a great job describing the many things we don’t know about LLMs starting at 4’17” of this Lex Fridman deep dive. (Yes, I listened to all 5+ hours and especially appreciated the bemused curiosity of AI philosopher Amanda Askell!)

4. He’s not alone: the medical director of the UK’s National Health Service says, "we can’t medicate our way out of obesity crisis.”

5. The question for policymakers is: Who qualifies for subsidized GLP-1s? There is no doubt that widespread usage will bring down healthcare costs for everyone. But some countries are considering only subsidizing GLP-1s if recipients also sign up for nutrition and exercise programs. (In the same vein, then, should the government require Adderall recipients to sign up for meditation classes?)

As new medications ease the burden of being human, I expect the MAHA fringe to grow into a larger movement of anti-pharmaceutical activists arguing that we should embrace the human struggle while new research reveals that impulse “control” is less controllable for some than others.

Twin studies suggest that roughly half of impulse control and addiction comes from genetic factors. GLP-1s seem to affect the reward pathways of the brain that control impulsive behavior. But that’s just the start. What should we expect from the MAHA movement when genetic medicines alter the underlying genes that shape impulsivity?

6. I appreciated

Matthew Yglesias

’s vulnerable reflection on losing 70 pounds.

7. The debate is unfolding worldwide. Yesterday’s Sunday Times had an opinion piece arguing that ADHD is over-diagnosed and over-medicated at the expense of addressing phone addiction. Another piece argued that anti-depressants are less effective over the long-term than pharmaceutical companies claim compared to non-pharmaceutical interventions. I don’t know how to evaluate either argument, but they are representative of the growing skepticism about pharmaceutical cures to common ills. For 2025, I’ll continue following the excellent GLP-1 reporting by Sarah Zhang at The Atlantic and Emily Mullin at Wired.

8. By 2020, there was a whole bookshelf about the hypocritical elite entrenchment of the Ivy League left, including Dream Hoarders by Richard Reeves, The Sum of Small Things by Elizabeth Currid-Halkett, Uneasy Street by Rachel Sherman, and Winners Take All by

Anand Giridharadas

.

9. Of course, Henderson went to Yale just like JD Vance and Josh Hawley. Most of the well-known critics of Ivy Leaguers’ elite entrenchment are also Ivy Leaguers. Matt Yglesias, another Ivy Leaguer, says the concept of luxury beliefs is its own form of a luxury belief. Ivy league status competition all the way down!

10. I enjoyed Thomas Edsall’s NYT piece on this: “America has an anti-MAGA majority, but not necessarily a pro-Democratic one.”

11. More recently, Musk raised $6 billion from some of the largest investment banks to compete with OpenAI, Google, Microsoft, and Anthropic. TechCrunch has a great article describing how the AI accelerationists won and doomers lost.

12. I’m halfway through Naomi Klein’s Doppelganger, a fascinating exploration of Naomi Wolf’s descent down the rabbit hole from respected feminist theorist to wacky conspiracy theorist. But it’s not just Wolf, and it’s not just conservatives; the whole culture is returning to the 1970s conspiratorial mindset of a Don DeLillo novel. It’s wild to read Naomi Klein’s takedown of Naomi Wolf’s conspiracy theories while offering up her own. Then again, observing two conspiracy theorists argue with each other about who has a stronger grasp on reality is somewhat representative of public discourse in 2025.

13. A lot of the discourse at the beginning of 2024 was how humans shouldn’t trust hallucinating chatbots, but I imagine that people who spend more time with chatbots will be more informed than people who spend more time on social media. I have a hunch that in five years we’ll see chatbots like whole foods and social media like ultraprocessed foods. For 2025, I’ll keep following the excellent writing of Dan Williams on misinformation, epistemology, and biased reasoning.

14. Earlier this year, Bloomberg published a great deep-dive into “how America became a republic of distrust” with some ideas about what can be done about it. There are no silver bullets. Most of us agree that there are some limits to free speech, but we don’t want a committee of experts to police permissable thought. We need something in between for a world where anyone — including chatbots — can potentially say anything to 4 billion people … and then get even more attention if they are censored. It’s not a great look for Zuckerberg to fire the fact-checkers before Trump’s inaugration, but community moderation has worked well for Wikipedia and X, and I’m curious to see how it goes for Meta. More than anything, I support Yuval Harari’s argument that anyone should be able to say just about anything online, but companies should be held accountable for what their algorithms amplify.

15. Waymo’s autonomous vehicles have driven 33 million miles without human assistance — in which the company says there were 78% fewer injury-causing crashes compared with a human driver over the same distance.

16. Probably not, at least not in dense cities where the companies stand to make the most money. Overall, I’m excited about Robotaxis and their promise of less car ownership, fewer accidents, and less urban space dedicated to parking. But I’d rather not give up the dream of mass public transit and I’m annoyed that California won’t manage to finish its multi-decade high-speed rail before robotaxis become a more dominant form of transportation. (Then again, robotaxi timelines have consistently been too optimistic; we’ll see if 2025 truly becomes the breakout year.)

17. In a recent interview, Elon Musk revealed that the goal is to produce 50,000 to 100,000 Optimus humanoid robots this year, with a tenfold increase in production each year over the next two years.

18. And despite all this investment, China is making faster robotics advances.

19. Before reading Scott’s The Longevity Imperative, I told myself that I’d like to live a healthy life up until about 95, and then die rather quickly. But Scott’s book changed my mind; if I’m healthy, lucid, and not a burden on society, then why not live to 120? It’s surreal to consider the possibility that I could be just a third of the way into my lifespan.

20. Will costly interventions limit longer lifespans to the rich? Shouldn’t we prioritize addressing malaria, TB, Pneumonia, and heart disease that reduce lifespans today before focusing on extending the lives of 70-somethings? And will longer lifespans increase the burden on the working youth to cover the costs of their parents’ 50-year retirements?

21. I sorta feel like, “let the apes beat their chests, and wake me up when it’s over.”

22. If so, those 500 companies will need to become much more profitable. Typically, the price of a company’s stock is around 17x its earnings per share. Today, it’s above 26x, which means these companies will need to grow 2-3% each year more than usual. It sure feels like we’re still in a bubble, even with the 4% drop over the past month. But I don’t plan on selling any stocks. When I read the WSJ’s coverage of the increased productivity of the American worker/business, I expect that Daron Acemoglu will be way off in his prediction that AI will only contribute a .05% annual gain in productivity.