More or Less · Topics · Episode #166
An Ex-Anthropic Researcher Says There's a 10% Chance AI Kills Us All — Real Fear or Coordinated PSY-op?
Clip from the full recording, 32:34–40:03 · download mp4
A little-known former Anthropic researcher goes on an Altman-and-Taylor-Swift-scale media tour warning AI could end humanity by decade's end — and Dave lays out, half-jokingly, why the whole thing looks like a coordinated regulatory-capture play by the labs themselves.
A researcher who recently left Anthropic — with essentially no public profile before this — goes on what Jess calls a Taylor-Swift-level media tour (Anderson Cooper included) warning that AI carries something like a 10% chance of ending humanity, and that society has “until the end of the decade to survive.” Brit’s read is cynical about the media dynamics at play: “if it bleeds, it leads” — pointing out that actual AI adoption grew only 0.4% in August, essentially flat, even as doom coverage spikes, because bad news travels (“this is like the human experience… why everyone follows true crime”).
Dave lays out two competing theories. The first, which he frames explicitly as the conspiracy read: this is a coordinated PSY-op for regulatory capture — Anthropic’s messaging has been consistent for months, he’s heard for months that they’ve been quietly lobbying the White House to make open-source AI illegal, and an open letter weeks earlier (which Dave publicly dared signatories to back up by actually quitting) primed exactly this kind of dramatic “I quit and I’m scared” moment to force lawmakers to act.
Dave’s second, more sincere theory: he’d just spent the weekend with Astra and had his first real “marginally scary” experience with an AI model in a year or two — watching his 12-year-old son use it to 3D-model and print an object in about an hour, at a speed that, in the wrong hands, is “legitimately scary.” That’s what makes him take the researcher’s underlying point seriously even while doubting the packaging: the labs’ own capital and competitive dynamics mean nobody can actually afford to slow down no matter what anyone says publicly.
Key points
- An ex-Anthropic researcher with almost no prior public profile goes on a wall-to-wall media tour (Jess: "even Taylor Swift would be envious") claiming roughly a 10% chance AI ends humanity by the end of the decade.
- Brit's framing: actual AI adoption grew only 0.4% in August (essentially flat) even as doom coverage spikes — bad news travels regardless of real-world usage data.
- Dave's "conspiracy hat on" theory: coordinated PSY-op for regulatory capture, building on months of Anthropic quietly lobbying to restrict open-source AI, following an open letter weeks earlier that Dave publicly dared signatories to back with their careers.
- Dave gives the researcher credit for one thing regardless of motive: actually quitting and leaving Anthropic equity on the table, rather than just signing a letter and staying employed.
- Dave's "conspiracy hat off" theory: this week's hands-on time with Astra was the first time in a year or two a model felt genuinely scary to him — specifically watching his 12-year-old 3D-model and print an object within an hour.
- Dave's synthesis: whether or not the media tour is stage-managed, the underlying capability jump (and what it enables in the wrong hands) is real and worth taking seriously.
Where they landed
Quotes
“He thinks there's a ten percent chance AI's gonna kill us all... he's gone on a media tour that even Taylor Swift would be envious of.”— Jessica
“If it bleeds, it leads... AI adoption was 0.4% in the month of August, which is pretty terrible. It's like plateaued.”— Brit
“This was a PSY-op that has been planned for a long time to create regulatory capture... they've been trying to convince the White House and everybody under the sun to make open source illegal.”— Dave
Suggested tweets
A little-known ex-Anthropic researcher just went on a media tour warning there's a 10% chance AI kills us all. Jess Lessin: "He's gone on a media tour that even Taylor Swift would be envious of."
Tweet this →Brit Morin's response to the AI-doom media blitz: actual AI adoption grew 0.4% in August. "If it bleeds, it leads" — the panic and the usage data are telling two different stories.
Tweet this →Dave Morin's theory: the Anthropic doomsday researcher isn't a whistleblower, he's regulatory-capture theater. Months of quiet White House lobbying, an open letter daring people to quit — and then someone actually quit.
Tweet this →All topics · Full episode #166 + transcript · Subscribe on YouTube · Spotify