OpenAI’s Altman Slams Brakes on o3 Hype After Weeks of Teasing AGI Breakthrough

OpenAI CEO Sam Altman wants everyone to calm down about artificial general intelligence. Again.

“Twitter hype is out of control again,” Altman posted on X Monday, pulling back from weeks of cryptic teasers and bold predictions about the company’s latest AI model, o3. “We are not gonna deploy AGI next month, nor have we built it.”

“[Please] chill and cut your expectations 100x!” he wrote.

Such statements seem like a sharp U-turn from Altman’s recent blog post where he claimed OpenAI knew how to build AGI and predicted people would be working hand in hand with AI agents this year. He even suggested superintelligence could emerge “in a few thousand days.”

“We are now confident we know how to build AGI as we have traditionally understood it,” Altman wrote on Jan. 5. “We believe that, in 2025, we may see the first AI agents ‘join the workforce’ and materially change the output of companies.”

That kind of talk sent the AI community into a frenzy. But while Altman even theorized about the advent of superintelligent tools, other experts were more skeptical about the hype of using vague terms or grasping for mushy, imprecise definitions rather than standards.

“When considering what Altman said about AGI-level AI agents, it’s important to focus on how the definition of AGI has been evolving,” Humayun Sheikh, CEO of Fetch.ai and Chairman of the ASI Alliance, told Decrypt. “AGI has not yet reached a level of true sentience, and I don’t believe it will for quite some time.”

This isn’t the first time Altman has had to walk back his own hype. After hyping up GPT-4o’s capabilities, he later dismissed his own claims and called ChatGPT “mildly embarrassing at best.”

Do you see a pattern? Some people do:

Altman himself has reflected on how people debate the potential impact of new AI models, only to end up demanding more capabilities as models fail to meet their expectations.

As usual, Altman sprinkled in some scintillating promises to adorn his otherwise disappointing tweet. The OpenAI CEO said his company has some “cool stuff” in the pipeline, just don’t expect it to pass the Turing test anytime soon. He did confirm in an earlier X post that OpenAI o3 mini, a weaker version of the full o3 reasoning model, is just a few weeks away.

OpenAI did not respond to requests for comment about specific capabilities of the o3 model or timelines for its release.