A group of AI researchers has released a collection of videos warning that superintelligence could wipe out humanity, with one expert saying the chance of extinction is about a coin flip. The videos come from current and former employees at OpenAI, Google DeepMind, and Anthropic.
Palisade Research, which describes itself as a non-profit studying AIs’ capabilities and motivations, gathered the interviews and posted them on frominside.ai. The project includes a dozen separate conversations with people working inside the field.
Geoffrey Irving’s Warning
Geoffrey Irving, a former OpenAI and Google DeepMind employee, opened the series with a blunt assessment. “The chance of human extinction is about a coin flip, in my view,” he said in a new interview.
Neel Nanda’s Odds
Neel Nanda, a Google DeepMind research scientist, offered a slightly more optimistic but still alarming number. “There’s at least a 10 percent chance that it causes human extinction, and that is ridiculously high,” he said.
Nanda’s 10 percent figure is lower than Irving’s coin flip, but he frames it as absurdly high. Both researchers are warning about the same thing: the potential for AI to cause humanity’s end.
Daniel Kokotajilo’s Dire Claim
Daniel Kokotajilo, a former OpenAI researcher, went further. He described superintelligent AI as god-like in power. “Unfortunately we don’t know how to control them at all,” he said. “And so probably nobody would control them. This is exactly as dangerous as it sounds and must not be allowed to happen.”
Kokotajilo’s warning is stark.
Why They Keep Working
The videos also address a practical question: why would anyone work on something this dangerous? Some answers were direct.
- Google’s Mary Phuong acknowledged the skepticism her position invites. “I think you absolutely should be suspicious of what I’m saying because I am being paid by the lab,” she said.
- Neel Nanda offered a personal commitment. “If I did not believe that my work was directly reducing these existential risks, I would quit… These companies are not going to stop making these systems just because I quit.”
Both statements suggest a tension: these researchers see the danger clearly, but they believe their work is the best way to reduce it.
What the Videos Don’t Offer
The collection includes segments from multiple speakers on similar topics, including whether the concern is hype or marketing. The site also asks why researchers would work on something they believe is this dangerous.
“This is exactly as dangerous as it sounds and must not be allowed to happen.”
The videos do not propose a single solution.
Reading Between the Lines
The videos show a split in how researchers present themselves. Some, like Phuong, acknowledge the suspicion their positions invite. Others, like Kokotajilo, frame the work as a moral obligation.
The project raises a question about trust. When the people building AI are openly worried, what does that mean for everyone else?
This is a striking admission from the very people most capable of shaping the future. The videos do not offer comfort, and they do not offer answers. They offer a warning, and they ask the viewer to take it seriously.
Source material: “AI researchers put out videos saying superintelligence is ‘exactly as dangerous as it sounds’,” The Verge.
Get the Notebook.
The day's best stories and every fresh verdict, in plain English, in your inbox by seven. One email a day, no more.