In an interview with computer scientist and podcaster Lex Fridman, Sam Altman had some surprising words about GPT-4, the latest AI model from his company, when asked about its most impressive capabilities.

“I think it kind of sucks, relative to where we need to get to and where I believe we will get to,” Altman said. “I expect that the delta between [GPT-]5 and 4 will be the same that was between 4 and 3 and I think it is our job to live a few years in the future and remember that the tools we have now are going to kind of suck, looking backward at them, and that’s how we make sure the future is better.”

GPT-4, Altman noted, was a huge advancement over GPT-3. But people who have used GPT-4 now scoff at its predecessor, which at the time seemed revolutionary. That’s where he stands now, as GPT-5 gets closer to release. (OpenAI has said its next-generation model will release in the next year, but has not been more specific than that.)

“Look, I don’t want to downplay the accomplishment of GPT-4, but I don’t want to overstate it either,” said Altman. “And I think this point that we are on an exponential curve, we’ll look back relatively soon at GPT-4 like we look back at GPT-3 now.”

Altman said he uses GPT-4 as “sort of like a brainstorming partner” and notes it can sometimes help on longer-horizon tasks, breaking them into smaller steps and assisting with those.

“When that works, which is not very often, it’s very magical,” he said.