Weekly Top Picks #113
As bad as the Industrial Revolution / Nature: AGI is here / Moltbook was fake (of course) / Don't obsess over what AI can or can't do: find your authentic self
Hey there, I’m Alberto! 👋 Each week, I publish long-form AI analysis covering culture, philosophy, and business for The Algorithmic Bridge. Paid subscribers also get Monday news commentary and Friday how-to guides. I publish occasional extra articles. If you’d like to become a paid subscriber, here’s a button for that:
THE WEEK IN AI AT A GLANCE
I’m keeping this week’s picks to four categories. I will do this as I see fit, depending on the importance of events. Just those four amount to ~3,700 words.
Work & Workers: Tech leaders love invoking the Industrial Revolution as reassurance; economic historians who actually study it would like a word.
Products & Capabilities: Nature says AGI is already here; DeepMind’s co-founder (and I) say no; the gap between them is itself the story.
Trust & Safety: Moltbook: 17,000 humans, 1.5 million “agents,” an exposed database, and the most predictable collapse in AI since the last one.
Culture & Society: The list of things AI “can’t” do only shrinks; start from what you want from life instead.
THE WEEK IN THE ALGORITHMIC BRIDGE
(PAID) Weekly Top Picks #112: Besides Moltbook: Can AI be profitable? / China is ahead is behind / Zero privacy / A new era of AI research / Anthropic’s conundrum / The useful power of power users.
(FREE) OpenAI Won the Consumer Mindshare—And Paid For It With Everything Else: I asked Claude and Gemini to investigate what’s going on at OpenAI. Things got out of hand (fictionalized, I wrote it myself!).
(PAID) The Stock Market Has No Idea What’s Coming: Investors are simultaneously terrified that AI works and that it doesn’t; both fears can’t be true, and yet both are rational.
(FREE) You Spent Your Whole Life Getting Good at the Wrong Thing: AI collapses the “how” of work into a wish. The hard part now is figuring out what to wish for.
WORK & WORKERS
The Industrial Revolution was not a vacation
Tech leaders love invoking the Industrial Revolution as a reassurance for what’s coming from AI. Satya Nadella recently told Dwarkesh he’d “love to compress what happened in 200 years of the industrial revolution into a 20-year period.” Sam Altman told Adam Grant that AI is “more like the industrial revolution than the internet revolution.” It sounds grand and comforting.
But what did the Industrial Revolution actually feel like for the people living at the time? A VoxDev podcast episode this week put economic historian Bruno Caprettini in front of a microphone, and the answer was: terrible, for a very long time.
As Caprettini explained, during the Napoleonic Wars, roughly 10% of Britain's prime-age male workforce was away fighting. With labor scarce, wages increased, and farmers turned to threshing machines to do the work. When the soldiers came home, Caprettini said, they found their jobs had been taken by machines. “The wages of these men very very quickly collapsed.”
What followed was “the 1810s and 1820s,” Caprettini explains, “years of intense social unrest”: first the Luddite revolts in the cities, then the Captain Swing riots in the countryside, the “largest episode of social disorder in English history.” Caprettini emphasizes how “frightening [the riots and riotters] seemed to elites.”
The parallel to AI is quite clear, given what AI leaders say. Caprettini and colleagues’ calculation: the threshing machines displaced “at least one third of the jobs of the people that were employed before the Napoleonic War” in the rural workforce. For comparison, they shared Dario Amodei’s infamous prediction on Axios: “AI could wipe out half of all entry-level white-collar jobs—and spike unemployment to 10-20% in the next one to five years.”
It was a manual labor bloodbath 200 years ago, and it will be a cognitive labor “bloodbath” now, to use Axios’ fitting expression.
Caprettini identified two factors that made the disruption explosive: “the scale and the speed of the shock.” The machines had been adopted while the men were at war, so “for them there was no transition,” Caprettini said. He sees both elements in AI. “Anyone who has played around with these models can see how vast their capabilities are and are becoming and how fast they improve,” he said.
But he flags a crucial difference. The industrial revolution was “deskilling” in the sense that it replaced artisans who had trained for “five to seven years.” AI, Caprettini argued, is deskilling “to the extreme”: coming for lawyers, accountants, and PhD researchers who have studied for “twelve to fifteen years.”
(I’ve written extensively about both things: 1) The importance of taking care of the transition much more than focusing on what dies and what emerges anew, and 2) the importance of assessing which is AI’s target workforce and how a combination of office + apprenticeship should be implemented across the cognitive economy to minimize the “vanishing threshold” effect. For reference: my proposal for the latter point, which went overlooked here on Substack, was shared by Johannes Kopf, CEO of the Austrian Public Employment Service (AMS), on his blog, Bluesky, and LinkedIn, which I kindly appreciate.)
One finding from Caprettini’s research offers a sliver of hope: “the blow of this new labor-saving technology was somewhat less big in villages that were in close proximity to an industrial town.” Displaced workers could leave for manufacturing jobs in the cities. The key variable was whether alternative employment existed nearby.
Alex Imas, a professor of economics and applied AI at Chicago Booth, put it sharply: “There was a generation + of people who had their lives made significantly, hellishly worse and it took decades for things to get better. . . We have more than a century of economics research since then for structuring policy that ensures that innovation/growth continues while keeping gains broadly distributed. We should use it.”
But Caprettini also said something that policy alone won’t fix: “It will not be sufficient to just give money to the losers. People who are losing jobs are losing much more than just a source of income. They’re losing what in many cases gives meaning to their lives.”
I also wrote about this, in October 2022, one and a half months before ChatGPT was released into the world, in one of my first “important” posts for this newsletter (please, take a moment to appreciate the missing quality of the cover image, done with Midjourney V3, which is itself a great illustrative example of how things have changed in three years). (By the way, my current attitude on this is also my last post):
That’s the part the Nadellas and Altmans of the world skip over. When they say “Industrial Revolution,” they mean the destination—the 14x increase in living standards over two centuries—but they wave away the generations who lived through the transition. As I wrote in The Human Toll of Waiting for AI to Take Over:
The transition is terrifying; the uncertainty, irritating; the pressure to adapt and reorient plans that took decades to build, burdensome. I’m safe today. My grandchildren will be even better off in 50 years. But what about the in-between?
Sources: VoxDev, The Algorithmic Bridge, Axios, TED, Dwarkesh Podcast
PRODUCTS & CAPABILITIES
Does AI already have human-level intelligence?
The most important thing this week for the product & capabilities section is the release of Codex GPT-5.3 and Claude 4.6 Opus (the same day!), but I already wrote my take on the short-term future of AI agents in You Spent Your Whole Life Getting Good at the Wrong Thing, so I will focus on something else. (You can read a real review of how one model fares against the other at Interconnects by Nathan Lambert. His TL;DR: “codex is a very useful coding tool, claude is the first agent.”)
Nature has a piece on AGI that asks whether it is already here. Four researchers—a philosopher, an ML scientist, a linguist, and a cognitive scientist—argued that, by reasonable standards, “the long-standing problem of creating AGI has been solved.”
