OpenAI expert: AI hype on social media is more complicated than you think!

On social media, Noam Brown, an expert in artificial intelligence reasoning at OpenAI, said that there is currently a lot of "vague hype" about artificial intelligence, which has attracted his attention. While he acknowledged that progress in the field of artificial intelligence brings optimistic prospects, he also pointed out that there are still many research problems that need to be solved and emphasized that OpenAI has not yet achieved so-called "superintelligence."

Brown's statement is in sharp contrast to some recent views within OpenAI. In January, OpenAI researcher Stephen McAleer said: "I kind of miss the time in AI research when we didn't know how to create superintelligence." This suggests that OpenAI may have found a way to A clear path to artificial superintelligence (ASI).

Before joining OpenAI, Brown worked at Facebook Artificial Intelligence Research (FAIR), where his research focused on developing artificial intelligence systems that can beat human players in complex games such as poker and diplomacy. His research on the poker AI "Libratus" explored the concept of "test-time compute" and showed that increasing AI calculation time can improve game strategies.

Brown later introduced these concepts into OpenAI and applied them to the development of language models. The company's latest o1 model is based on this idea, opening up a new development direction by increasing "thinking time" rather than just improving training performance. He believes this alternative scaling approach will enable the emergence of new AI capabilities, noting: "Scaling in this area is still in its early stages."

Brown’s perspective reminds us that although the prospects of artificial intelligence are exciting, we cannot ignore the limitations and challenges that still exist in the development of current technology.