OpenAI safety employee quits, says ‘time for trial and error is over’

David Robinson warns that AI capabilities were advancing faster than researchers' understanding of alignment