Why Did Kevin Roose Lower His Odds of AI Doom?

Kevin Roose's probability of doom has gone down in the last few weeks because everyone is freaking out about AI risk, a shift he links, with some hesitation, to feeling paradoxically less worried.

Before Jacob Coxon resigned from Anthropic, and before Hugging Face, almost no one thought about AI risk or existential threats. The only people thinking about it were weirdos in the Bay Area. Now people on the street ask about it, and people at dinner tables talk about it. High school classmates he has not spoken with in 15 years text to ask whether everyone is going to die.

He frames the new attention as a possible movie moment, humanity banding together to save itself from an asteroid, and he stops short of treating the rescue as certain. A lot still has to go right. Waking up to the risk was the first hurdle, and the next task is figuring out what to do. Optimism rests on slowing development and buying more time for better ways of controlling the systems. He wants AI built to cure cancer and disease and to help people live longer, healthier lives, and he does not want to file expenses by hand. He wants the technology in daily life. Holding safety and control together is what would make the result amazing.