One thing I have changed my mind on a lot over the past year is open source AI models. AI capabilities are looking more sigmoidal, making alignment risks seem less dangerous, while despot control of AI has become far more of a risk with the election of Trump, and the control of XAI, Deepseek and
OpenAI seeming worse than what one would expect from an AGI with ethics more or less randomly selected from the corpus of human writing. I still hold to my view that if someone is choosing between becoming a doctor and becoming an AI researcher, X risks should lean one further towards becoming a
DOCTOR, contra the famous advice from 80k. For most people at least, I think.