Daniel Kokotajlo

That's a documentary, the way a press release is a novel.

It features Clement DeLange from Hugging Face, Ryan Greenblatt, Daniel Kokotajlo, and Natasha Tiku from The Washington Post.

The criticism is part of why it's interesting, though.

CyberNews says it leaves out independent cybersecurity experts and goes easy on accountability.

This was written by a former OpenAI researcher and other super forecasters with a pretty impressive track record.

Also, as you may remember, Daniel Kokotajlo put up an impressive stand against OpenAI on their non-disparagement clause.

He was essentially willing to forfeit millions of dollars, and yes, you can make that much as a safety researcher at OpenAI.

He was willing to forfeit that so he wouldn't have to sign the non-disparagement clause.

Then apparently in late 2026, the US Department of Defense will quietly begin contracting OpenAI or Google directly for cyber data analysis and R&D.

But I'm kind of confused because already, for at least a year, OpenAI have been working directly with the Pentagon.

Yes, before you guys tell me, I'm aware that Daniel Kokotajlo, who is the main author of this paper, did make some amazing predictions back in 2021 about the progress of AI.

I can link to that in the description, but that doesn't mean he's gonna be always right going forward.

Mm.

So, uh, if we take the prediction of, uh, Daniel Kokotajlo, uh, for, in, in his paper, AI 2027-

Mm

...

Now, I got to give a big shout to AI Village, which is part of AI Digest.

So the AI Village was actually an original idea proposed by Daniel Kokotajlo, where he proposed giving one hundred agents their own computer and let each pursue their goal in their own way according to their own vision while streaming the entire process.

So Daniel Kokotajlo is, of course, an ex-OpenAI researcher.

He got a lot of attention when he was kind of the whistleblower against OpenAI, pointing out areas where he disagreed with them and how they were doing AI research.

Point being, he's a machine learning AI researcher on the very inside of a frontier lab, and by all accounts, a highly ethical person.

Their case for apocalypse rests on two core ideas, machines pursuing goals their human creators never intended, and second, where a system gets smart enough to rewrite its own code, accelerating far beyond our ability to control it.

Just recently, former Anthropic researcher Jacob Coxon quit publicly, declaring that tech companies were, quote, "racing straight to super intelligence and gambling with our lives," a post that racked up 153 million views in 36 hours.

Anthropic's own alignment science lead went further, putting the odds of AI wiping out humanity this decade at over 10%.

And it's not just the whistleblowers.

So in fact, OpenAI had claimed that GPT-2 was too powerful and too unsafe to release.

Um, and so this, um, kind of theater of, you know, "Our models are too powerful, we're so scared" has been happening for such a long time.

So then I saw philosophers and other, other, uh, researchers pontificating about whether, uh, the GPT model can be ethical, whether it can be artistic, um, and things like this, and I was so irritated [laughs] beyond, beyond words because, um, a- as I say in my upcoming book, that's like saying, you know, can a bridge be ethical? Can a bridge decide to collapse? No.

Um, you think about the people who built the bridge, and you think about what, um, tests they did or didn't do, uh, rather than the, an actual artifact having some kind of emotions or agency of its own.

If the Hugging Face hacking swarm had been significantly more intelligent but similarly misbehaved and misaligned, it might have cost billions of dollars of damage or even cost lives.

And the other is Daniel Kokotajlo.

He was also a researcher for OpenAI.

Our ability to even notice misalignment problems is already quite poor and is set to get much worse in the near future.

He said China is not the only adversary because the industry may be building one of its own in misaligned AI.

Daniel Kokotajlo, formerly of OpenAI, said the labs are getting worse at spotting misaligned models.

Coxon and Turner both presented their numbers as personal estimates.

The journalist Taylor Lorenz has called Coxon's resignation post sanctimonious doomer posting.

The top three most viewed episodes from last month starts off with Bishop Robert Barron on This Past Weekend with Theo Von, bringing in 2.1 million views.

In second place we have Daniel Kokotajlo on The Joe Rogan Experience with two million views, and the third most viewed podcast episode from September 2026 was Neil deGrasse Tyson on This Past Weekend, totaling 1.7 million views.

The top three most viewed episodes of Kill Tony from September starts with episode 783 featuring Gary Owen and Brian Simpson at 3.9 million views.

In second place we got Yaakov Smirnoff and Ari Matty with 3.7 million views, and in third place is the Mystery Boys, Duncan Trussell and Kurt Metzger, clocking in at 3.2 million views.