At this point, unless you are very talented and/or working at Anthropic/OpenAI/Deepmind, I dont see much reason to avoid working in AI. The timeline is already burnt. The people who burnt it, often in the name of altruism, should be ashamed. But at some point the benefits of trying to do good things with a dangerous technology outweigh the downsides of accelerating progress. Prior to ~now it was quite bad to work on AI in more or less any capacity. But the train is leaving the station anyway. Marginal impacts are now smaller than the plausible positive impact of using the tech for good. Accelerating AI was an incredibly dumb strategy but at this point might as well play to the out where alignment isn’t that hard.