In the past few days, people noticed that some OpenAI engineers seemed more excited and mysterious than ever on social networks. They interpreted the teasers as a sign that OpenAI was on the verge of a big AI breakthrough. Also, as Sam Altman teased what the near future might bring, the speculation soon shifted towards AGI and ASI.
AGI will give AI like ChatGPT the ability to tackle any task just like a human would, only the AI would be much faster and have access to far more knowledge. ASI will then exceed what the human mind can comprehend.
“We are now confident we know how to build AGI as we have traditionally understood it,” Altman said in a blog post a few days ago, before the current hype about AGI and ASI. He also said that OpenAI was already starting to think about developing superintelligence.
Fast-forward to early Monday and the OpenAI CEO took to X to clarify that OpenAI isn’t about to deliver any sort of AGI upgrade to ChatGPT. At the same time, he did mention the next big model coming to the ChatGPT experience. His remarks come at a time when reports say OpenAI might hold a mysterious meeting with the US government about “PH.D-level super-agents.”
Super-agent refers to the evolution of AI agents that are starting to arrive. These are AI programs that can handle complex tasks for you. ChatGPT has a Tasks mode that lets you set specific instructions for ChatGPT, including the ability to set reminders for you. Also, the ChatGPT macOS app can connect to certain code and productivity apps to assist users with code and text generation.
Then, assistants like Siri and Gemini will be able to perform more on-device actions on smartphones, including accessing and controlling apps on behalf of the user.
According to Axios, one of the leading AI firms will announce a “next-level breakthrough that unleashes Ph.D.-level super-agents to do complex human tasks.”
The report mentions that Sam Altman has a closed-door briefing with US government officials in Washington on January 30th. It’s unclear whether the topic of super-agents is on the table for the purported meeting, but Axios notes that whatever the ChatGPT creator might be working on, it must be significant.
But sources say this coming advancement is significant. Several OpenAI staff have been telling friends they are both jazzed and spooked by recent progress.
Against this backdrop, Altman made the statement above a few hours ago, saying that “Twitter hype is out of control again.”
“We are not gonna deploy AGI next month, nor have we built it,” Alman said, not that I’d expect the CEO to confirm the existence of AGI on X before OpenAI is ready to announce it.
“We have some very cool stuff for you but pls chill and cut your expectations 100x!” Altman continued.
What is OpenAI about to release? It can’t be ChatGPT GPT-5, though that upgrade can never be ruled out. Reports say OpenAI has been working on the next-gen ChatGPT model for a while. The company could also unveil better AI agents, but not the super-agents Axios talks about.
More likely is for OpenAI to release the o3 reasoning model for ChatGPT, which the company announced before Christmas. Altman on Friday posted on X the tweet above, thanking safety researchers for the o3-mini tests.
“We have now finalized a version and are beginning the release process; planning to ship in ~a couple of weeks,” Altman said. A couple of weeks means late January or late February, which would be right on schedule. OpenAI said last month that the ChatGPT o3 models should be ready by the end of January.
