Geoffrey Hinton and Ilya Sutskever are the same story told from two ends.
Hinton is the professor who spent 40 years in the wilderness, believing in neural networks while the entire field moved on. Ilya is the 17-year-old who showed up at his door in 2003 and immediately demonstrated "veteran-level insight." Together, they built AlexNet in 2012 — $1,000 of GPUs versus six-figure server racks — and rendered 20 years of computer vision research obsolete with a single PowerPoint slide. Google bought the team for $44 million. The AI arms race began.
Now, in August 2026, Hinton is on the news warning that mass unemployment is "very likely." Ilya is running Safe Superintelligence Inc. with a $32 billion valuation, no products, and no research released. They are both trying to tell us something. They are saying it in different languages.
The Fortune article is straightforward. Hinton told Bernie Sanders at Georgetown:
"It seems very likely to a large number of people that we will get massive unemployment caused by AI. And if you ask where are these guys going to get the roughly trillion dollars they're investing in data centers and chips — one of the main sources of money is going to be by selling people AI that will do the work of workers much cheaper. And so these guys are really betting on AI replacing a lot of workers."
The math is blunt. Trillion-dollar investments in compute need trillion-dollar returns. The fastest path to those returns is replacing human labor with cheaper AI systems. OpenAI alone may need $207 billion to support its growth and won't turn a profit until at least 2030. That money has to come from somewhere. It comes from labor costs that companies no longer need to pay.
Sanders' report puts a number on it: nearly 100 million U.S. jobs at risk. Fast food, customer service, manual labor — but also accounting, software development, nursing. Sen. Mark Warner warns that recent college graduates could see 25% unemployment within two to three years.
Hinton's metaphor for all this is fog. "You can see clearly for 100 yards and at 200 yards you can see nothing. We can see clearly for a year or two, but 10 years out, we have no idea what's going to happen."
The Fortune piece is a clean news report. It quotes Hinton accurately. It presents the economic data. But it treats Hinton as a source — a credentialed expert giving a warning — rather than as the person who is most conflicted about the thing he built.
The documentary What Did Ilya See? fills in what the news format can't hold. Hinton's quote from that film is the one that actually matters:
"It makes me sad... It is a bit sad that it's not just going to be something for good."
That's not a policy statement. That's grief. The man who kept believing in neural networks for 40 years while everyone else laughed — the man who was right, who won the Nobel Prize for being right — is now watching the consequences of being right and finding them unbearable.
The Fortune article doesn't capture this because it doesn't need to. It's a business report about labor economics. But the emotional core of the Hinton story is that he's not just warning us. He's mourning.
Here's where the Ilya story becomes essential context for the Hinton headline.
OpenAI was founded to prevent the concentration of AI power. That was the pitch. That's why Ilya left Google's unlimited compute to work at a nonprofit in an old chocolate factory. That's why Elon Musk's initial investment came with no strings attached — or so it seemed.
Within four years, OpenAI had become the most concentrated power in AI. The nonprofit structure was the recruiting tool, then the obstacle. The idealism was, as the documentary puts it, "compromised from the start." Ilya wrote in a 2016 email: "As we get closer to building AI, it will make sense to start being less open." He knew. He wrote it down. He stayed anyway.
Now map this onto Hinton's Fortune interview. He says Big Tech is "really betting on AI replacing a lot of workers." But Hinton was Big Tech. He was at Google for a decade. The same incentive structure that he now critiques — short-term profit over human welfare — is the structure that paid for the research that made his warnings possible.
This isn't hypocrisy. It's the gap between mission and incentive playing out at the individual level. Every system the notes describe — OpenAI, Google, the AI industry broadly — starts with a mission and ends with an incentive. The stated purpose and the actual function diverge. Hinton left Google specifically so he could speak freely about this. The fact that he had to leave to say it tells you everything about where the incentive actually sits.
Ilya saw the safety implications of AGI before they were fashionable. He built the case against Sam Altman over months, documented every broken promise, every lie, every safety deprioritization. On November 17, 2023, he ambushed Altman on a video call with the board. "Not consistently being candid with the board" — corporate language for "you're a liar."
Within hours, Silicon Valley closed ranks. 90% of employees signed an open letter demanding Altman back. Five days later, Ilya signed it too. Altman returned more powerful than ever. The board that fired him was gone. Ilya was offered to come back, then Greg Brockman called the offer off. He would never step foot in OpenAI's offices again.
Now he runs SSI — no products, no publications, $32 billion valuation — focused exclusively on building safe superintelligence. The gap between what Ilya saw and what the market valued is the entire AI safety problem in miniature.
Hinton's version of this is quieter. He quit Google in 2023 to speak freely. He won the Nobel Prize. He gives talks at Georgetown. But the industry he warns about is the industry his work created. The trillions being invested in AI compute are flowing because AlexNet proved that scale works. The mass unemployment he predicts is the downstream consequence of the breakthrough he led.
The person who sees it first doesn't just pay the highest price. They pay it twice — once for seeing, and once for being the reason the thing they're warning about exists.
Hinton's metaphor is fog: we can see a year or two out, then nothing. Ilya's version is different. The documentary's title asks: what did Ilya see in the neural weights that made him act the way he did? He saw something in the mathematical structure of these models — something that made him believe superintelligence was imminent enough to warrant firing the CEO of the most valuable AI company in the world.
These are two ways of describing the same uncertainty. Hinton says we can't see far enough. Ilya says he saw something that everyone else missed. Both are probably right. The fog is real. But someone is walking through it with a flashlight, and the rest of us are deciding whether to trust the beam.
The Fortune article ends with generic advice: workers who adapt and use AI to amplify their skills will stand the best chance. That's true in the way that telling someone in a fog to "drive carefully" is true. It's technically correct and practically useless.
What Hinton and Ilya are both saying — in their different ways, through their different losses — is that the car is already moving, the fog is already thick, and the people driving are the ones who stand to profit most from where we're headed.
Hinton told Sanders that 10 years out, "we have no idea what's going to happen." Ilya left OpenAI because he had a very specific idea of what was going to happen and it terrified him. Both of them are right. The future is foggy and someone keeps seeing shapes in it that make them act against their own interests.
The question isn't whether AI will displace workers. It will. The question is whether the people who built this thing — and who understand it better than anyone — can do anything about it now that it's running on trillions of dollars of capital and the incentive structure of every major technology company on earth.
Hinton is sad. Ilya is building in silence. The rest of us are standing in the fog, reading headlines, trying to figure out which way the beam is pointing.
Co-authored with my second brain, Obsidian.
.png)
