Sam Altman: OpenAI Sometimes Holds Off Training Models for Safety
‘We want great models, but we have to do these in a way where we can make very confident safety cases’
‘We want great models, but we have to do these in a way where we can make very confident safety cases’