OpenAI commence training on GPT-5

Estimated reading time: 2 minutes

OpenAI also forms a safety committee after two senior execs from its ‘superalignment team’ resign over concerns that OpenAI isn’t taking AI safety seriously.

OpenAI commence training a new advanced AI model that promises to replace its current flagship product, GPT-4o.

“OpenAI has recently begun training its next frontier model and we anticipate the resulting systems to bring us to the next level of capabilities on our path to AGI,” or artificial general intelligence, the company says.

OpenAI executives have been teasing the next generation ChatGPT as a major advancement, but for now, the company is declining to reveal more, including what the AI model will be called and whether it’s the long-awaited GPT-5.

Its announcement mainly concerns a new “Safety and Security Committee” meant to rein in future AI projects. That comes shortly after the leaders of OpenAI’s Superalignment team resigned. One of them, Jan Leike, said he left because OpenAI isn’t taking AI safety seriously.

• OpenAI introduces tool for Identifying images generated by DALL-E 3
• 5 compelling reasons to dress your age
• How can students’ safety be ensured beyond campus?
• Is AI all Sunshine and Rainbows? Exploring the potential downsides
• Impact of Artificial intelligence on life and jobs

In response, OpenAI CEO Sam Altman acknowledged that Leike is “right [that] we have a lot more to do [and] we are committed to doing it.” Today he then announced the Safety and Security Committee, which Altman will lead alongside OpenAI Board Chairman Bret Taylor, co-creator of Google Maps, and Nicole Seligman, general counsel for Sony. 

The committee will also include at least five technical and policy experts and tap outside experts, including former NSA Cybersecurity Director Rob Joyce and former US Assistant Attorney General for National Security John Carlin, for input.

“While we are proud to build and release models that are industry-leading on both capabilities and safety, we welcome a robust debate at this important moment,” the company says. 

Still, it doesn’t look like the debate will be open to the public. Instead, the committee’s first task “will be to evaluate and further develop OpenAI’s processes and safeguards over the next 90 days,” the company says. OpenAI will then publicly share an update on the “adopted recommendations” following a board review.

The timeline suggests OpenAI could release its next-generation model in the fall.