Ever since around the time of Jacob Coxon's post, I can't shake the logic that these capabilities will keep growing, proliferate, find their way into the wrong hands, and plausibly lead to mass casualties, or even the optimization of humans out of existence. Whatever you make of Coxon [1], the Hugging Face incident [2], or the frontier labs’ motivations, the argument is what sticks with me. My p(doom) isn't 100%, but it's high enough that I worry about it.
Plenty of AI researchers, of course, feel the same. In AI Impacts' 2023 survey of 2,778 AI researchers, the median estimate was a 5% chance of outcomes as bad as human extinction, and between 38% and 51% put it at 10% or higher. That was three years ago. [3]
What I'm struggling with is that the world seems to be carrying on as normal, either talking around the threat or ignoring it altogether. This feeling is new to me. I’d never believed in any doomsday scenario. In fact, I always thought people who did were more than a little crazy. Now I’m one of them.
I'm still going out with friends, talking with family, saving for retirement. But sometimes it’s hard to deeply care about anything else, including most of what hits the front page. If you feel this too, how are you handling it? And if you think I've got this wrong, I'd honestly like to hear why. It’s not what I want to be feeling about the future.
1. https://techcrunch.com/2026/09/09/gambling-with-our-lives-anthropic-researcher-quits-warns-against-self-improving-ai/ 2. https://simonwillison.net/2026/Jul/22/openai-cyberattack/ 3. https://arxiv.org/abs/2401.02843
jgrahamc•1h ago
This was pretty standard behaviour during the Cold War. Growing up in England I was well aware that a nuclear war could happen at any time and that England had absolutely no defence. And so we carried on living.