WOLF(@WOLF_Financial):SAM ALTMAN SAYS OPENAI HELD BACK A NEW AI MODEL OVER SAFETY CONCERNS, BUT "THERE'S NOT A BIG SCARY THING IN THIS CASE" The model came in "a little bit worse on a few of the evals," he told CNBC today, and he called the decision an abundance of caution. OpenAI's rule, Altman says, is that alignment, safety, monitoring and security have to stay "way ahead of capabilities." The company has held models back before, fixed them and launched them later, and he put this one in that same category. He also says OpenAI is pacing its progress, which sometimes means not training a model at all. "Pacing doesn't always mean slowing," he said.

6hours ago

SAM ALTMAN SAYS OPENAI HELD BACK A NEW AI MODEL OVER SAFETY CONCERNS, BUT "THERE'S NOT A BIG SCARY THING IN THIS CASE" The model came in "a little bit worse on a few of the evals," he told CNBC today, and he called the decision an abundance of caution. OpenAI's rule, Altman says, is that alignment, safety, monitoring and security have to stay "way ahead of capabilities." The company has held models back before, fixed them and launched them later, and he put this one in that same category. He also says OpenAI is pacing its progress, which sometimes means not training a model at all. "Pacing doesn't always mean slowing," he said.