Beating AI News Flash: OpenAI CEO Altman spoke at the Semafor conference about RSI (recursive self-improvement), which means letting AI participate in developing stronger next-generation AI and continuously improving its own capabilities.
The host asked him whether OpenAI has already achieved this capability.
Altman replied that, like AGI (artificial general intelligence), AI self-improvement is also hard to draw a clear line around. You cannot simply say that there was no AGI before and then suddenly there is.
As a result, this statement was taken out of context and spread as if Altman had personally announced that OpenAI had already achieved AGI.
In fact, he subsequently clearly explained OpenAI's progress: the company is indeed already using its own AI technology to accelerate next-generation model development. But if AI were to autonomously design the next-generation model without human oversight and continuously improve itself, his answer was "absolutely not."
Altman also said that OpenAI could have gone further on RSI, but currently cannot guarantee the safety of doing so, so it has chosen to slow down.
