Latest update
Anthropic safeguards researcher says "pretty much all humans" will become permanently unemployable, puts broadly superhuman AI about 1.3 years out
First seen on X 11 hours ago@kimmonismus ♥ 1,41627 views
"It's very likely that pretty much all humans on Earth will become permanently unemployable and will need to strictly rely on government welfare." That's Tao Lin, who works on safeguards at Anthropic, blocking misuse of Claude, in an interview posted Nov 15. An X post sharing it under "broadly superhuman AI could be just 1.3 years away" passed 1,400 likes.
The interview was published on YouTube by the AI safety group Palisade Research. Lin said he has been at Anthropic about a year and a half, working on automated alignment research and AI control (checking that internally deployed models don't exfiltrate weights or deceive staff), and now on filtering Claude requests about bioweapons, hacking and military drones. "Superhuman AI is going to come here very quickly," he said. "I have timelines of something like 1.3 years," meaning around the end of 2027, for AI as good as or better than the best expert in almost every field, from programming to filmmaking to politics.
He put the chance of an AI takeover at "less than 50%" but called mass job loss "the most predictable effect": everyone losing their current jobs, moving briefly to other work, then becoming completely unemployable. Asked whether that happens within 15 months, he said no, citing "a large gap" between when AI can replace people and when it actually does, partly because "there aren't enough robots yet." Asked why he doesn't quit, he said he works on safety, not capabilities. He was not speaking for Anthropic.
Reactions split. YouTube comments ranged from "Well that was horrifying" to thanks for his candor. On X, one reply noted people said software jobs would vanish within a year two years ago; another joked the weekly usage cap should be raised first. Anthropic is set to list on Nasdaq on Nov 19.
“Now we are talking: An Anthropic researcher says broadly superhuman AI could be just 1.3 years away.”
