The Man Who Set the Clock

The episode with Ed Zitron asked whether artificial intelligence can pay for itself. This conversation asks what happens when it works. Daniel Kokotajlo wrote forecasts at OpenAI until he quit in 2024. His opening line is the reason this two-hour conversation never has an easy moment: a new species may be emerging that ends up running the world, he says, and there is a seventy percent chance it goes horribly wrong. Watch from 0:00:07

The WitnessWho's talking — and what he saw

Kokotajlo joined OpenAI in 2022, in the governance team. His job was to track progress and write scenarios — the internal precursor to what later became public as AI 2027. He wasn't an engineer training models. He was the person tasked with writing down where the curves were pointing. Watch from 0:00:26

What wore him down wasn't the technology but the gap between what the labs say and what they do. The safety narrative of the big labs, he argues, functions more as after-the-fact justification than as an actual brake. He points to emails from the legal dispute between Elon Musk and OpenAI, in which the founders wrote in 2017 that part of the reason they started OpenAI was fear that Demis Hassabis could become a dictator with a general AI. Watch from 0:12:33 His rule for judging the industry's leaders is blunt: judge people by their actions, not their words. Watch from 0:13:21

One scene sticks. After ChatGPT's success, there was an all-hands meeting, and then-chief scientist Ilya Sutskever summed up the mood in a single line: focus on the mission, we have to build general AI. Watch from 0:16:38 Kokotajlo quit in 2024. Watch from 0:14:49

The Two MillionWhat the title claims — and what he actually says

The video's title promises a bribery story: two million dollars for silence. Within a minute, Kokotajlo dismantles it himself. When Bartlett repeats the figure, he cuts in: he was ultimately allowed to keep the money. Watch from 0:17:25

What actually happened was different. His exit paperwork contained a clause barring him for life from criticizing the company — and a second clause barring him from even mentioning the first. For an organization founded to benefit all of humanity, he found that remarkable. Anyone who refused to sign forfeited their equity. In his case, that meant roughly eighty percent of his family's net worth. He and his wife talked it over, brought in lawyers, and he didn't sign. Watch from 0:18:25

Then the story went public, employees started asking questions internally, and the company withdrew the clause: forget it, you keep your equity. Watch from 0:18:58 Leadership later claimed it hadn't known about the clause. He doesn't buy that. Watch from 0:19:10

So here's where it lands: the sacrifice was real, and it was meant to be costly. He never actually lost the money. Anyone taking the title at face value hasn't watched the conversation.

The NumberSeventy percent — and what it doesn't mean

The seventy percent is his personal estimate, not a measurement. It doesn't stand for "seventy percent extinction," but for some major catastrophe — which could mean loss of control, but could just as easily mean a tiny group seizing power. Bartlett pushes back with another figure: one industry leader puts it at seven percent. Kokotajlo's reply is the strongest image of the whole conversation — if there were a hundred buttons on a table and ten of them ended the world, nobody would call that a small risk. Watch from 0:41:37

That's the real argument, and it doesn't hinge on his particular percentage. Even someone who thinks seven percent is the right number is still describing a risk that no other industry would ever be allowed to run.

The Calendar2027, 2028, 2030 — a moving date

His current median estimate for when systems will outperform the best humans at everything sits at 2029. Watch from 0:03:20 He adds that it could also take considerably longer. What's more notable is how often that date has already shifted: from 2027 to 2028, then to 2030. Watch from 0:23:24 Now, he says, people inside the labs are telling him to move it back again.

The AI 2027 scenario was never meant as a prediction — it's a worked-through path, month by month, with two possible endings: the race, in which systems eventually stop obeying, and the slowdown, in which control is achieved but power ends up concentrated in very few hands. Watch from 0:25:19 Bartlett presses him on which of the papers the vice president actually read. Watch from 0:24:07

A scenario convinces precisely because it's concrete. That's also its weakness: every stage has to unfold roughly as described, or the chain breaks. The fact that the authors have repeatedly pushed their own median estimates later speaks well of their honesty — and against the drama of the year in the title.

The StrategyWhy the labs automate themselves first

As he lays it out, the big labs' plan is simple. First automate coding, then the entire research operation: ideas, evaluation, communication. The endpoint is a lab that no longer needs human employees, because an army of copies of the same model trains the next generation. Anthropic's chief calls that a country of geniuses in a data center; Kokotajlo thinks "army" is the more accurate word, since every copy answers to a single owner. Watch from 0:06:40

And he backs it up with the one figure in the whole conversation that's actually measured: in 2020, the largest models had around 175 billion parameters. Watch from 0:34:51 Today that figure is on the order of ten trillion. Watch from 0:31:52 Two orders of magnitude in six years.

The JobsWhy the wave arrives late — and then all at once

Here he pushes back on the common assumption. In his scenario, mass unemployment doesn't come first — it comes last, after 2028. The reason is the same one from before: the labs automate their own research before they automate the economy. So very little is visible for a long stretch — and then a great deal happens at once. Watch from 0:57:22

For the debate on this side of the Atlantic, that's the most uncomfortable part of the conversation. Anyone pointing to steady employment numbers today as reassurance is, by this logic, measuring exactly the wrong thing.

Plan AThe alternative scenario he himself thinks is unlikely

The second half of the conversation turns to the plan he sets against all this: AI 2040. The idea isn't to prevent superintelligence but to delay it by ten years — through regulation, disclosure requirements, and a verifiable agreement between the US and China. Watch from 1:00:28 On his timeline, a machine is doing a fifth of all cognitive labor by 2031 Watch from 1:22:11; by 2033, sixty million systems are running at a hundred times human speed Watch from 1:25:21; and the gains get paid out as a citizens' dividend — starting at around 25,000 dollars a year per person. Watch from 1:26:20 Only in 2040 is the brake finally released. Watch from 1:30:37

The most honest answer of the evening comes when Bartlett asks which of his five scenarios he thinks is most likely. It isn't Plan A. It's the plan where everything just keeps going the way it's going. Watch from 1:05:39

The ButtonHow the conversation ends

Bartlett poses a thought experiment: a button that would stop frontier development forever. Kokotajlo visibly squirms. A temporary pause, he'd hit without hesitation. A permanent one, he hesitates over — and in the end doesn't press, because he thinks the potential upside is too large to take away from humanity forever. Watch from 1:46:29

His closing message is more modest than the title suggests. It's not too late, he says; if it were, he wouldn't be here — he'd be with his family. Watch from 1:57:29 What he wants from listeners isn't a revolt but the opposite of looking away: pay attention, form an opinion, and ask candidates about it. Watch from 1:59:14

The Reality CheckWhat shouldn't be taken on faith

Three things belong in a fair accounting. First, the seventy percent figure is an opinion, not a measurement; other well-informed people put it at ten or twenty percent, and there is no expert consensus. Second, the claim that people inside the labs privately believe in even shorter timelines can't be verified from outside — and anyone who takes safety seriously is, unsurprisingly, more likely to talk to him than to others. Third, "very powerful" doesn't automatically imply his particular policy proposal; a verifiable agreement between two superpowers over a technology built inside data centers has historically never worked well.

None of that touches what remains solid: his departure, the clauses, the curves, and the question of who gets to decide about a technology that builds its own successors.

Summary and context by kipode.de. The text is our own; statements from the original are timestamped. The video is from The Diary of a CEO and is embedded via the official YouTube player. Not an official translation, no connection to the podcast.