News
OpenAI Concedes It Does Not Know How to Safely Reach Full Recursive Self-Improvement; xAI Targets 2027
Fortune reports OpenAI stated this month it does not yet know how to 'safely get all the way to aligned, full RSI' and 'cannot assume that progress in alignment and safety will keep pace,' while still targeting an autonomous AI researcher by March 2028 and an automated 'research intern' before that. Musk said for xAI's Grok models 'every successive model is built by the one before it,' with full automation possibly by the end of this year and no later than 2027. Future of Life Institute's Anthony Aguirre calls full autonomy 'probably the worst idea in the history of humanity,' making this the rare story where the labs' own written hedge is the news.
Source
↳ Follow the thread