In a recent interview, OpenAI CEO Sam Altman offered a refreshingly candid perspective on the state of generative AI. While acknowledging the significant progress made with GPT-4, Altman said kinda sucks compared to OpenAI’s vision for the future.
“I think it kind of sucks, relative to where we need to get to and where I believe we will get to,” Altman stated in a conversation with Lex Fridman. This blunt assessment highlights OpenAI’s focus on continuous improvement to make current advancements stepping stones for future breakthroughs.
Rapid Progress Expected in GPT-5
Altman draws a parallel between the advancements from GPT-3 to GPT-4 and the anticipated leap from GPT-4 to GPT-5. He suggests that within a short timeframe, we’ll likely view GPT-4 with the same limitations we now associate with its predecessor.
Altman’s statement reinforces the notion that the field of generative AI is on an exponential curve, with rapid progress expected in the coming years.
Despite his reservations, Altman acknowledges the value of GPT-4, particularly as a brainstorming partner. He highlights its potential to assist with complex tasks by breaking them down into smaller, more manageable steps.
However, Altman still believes that the functionality remains somewhat inconsistent. “When that works, which is not very often, it’s very magical,” he remarks. This suggests a need for further refinement in GPT-4’s ability to provide consistently reliable assistance.
OpenAI Confirms Development of GPT-5
OpenAI has confirmed the development of GPT-5 but has yet to announce a specific release date. There is no detailed information available on what GPT-5 can do. However, the general consensus is that it will be a major leap from the older models.
During an interview with Bill Gates in January, Altman mentioned that the team will focus on enhancing reasoning abilities and incorporating video processing capabilities for the upcoming models.