OpenAI's CEO Sam Altman revealed on Thursday that the company was unable to release new products at the expected frequency due to computing limitations. He explained that as AI models become more complex, the company struggles to manage multiple projects, especially in terms of computing resource allocation.
It is understood that the Microsoft-backed company is currently collaborating with Broadcom and TSMC to design its own chips to address the computing capacity issue. Altman confirmed that OpenAI will not release its next large AI model GPT-5 this year, but plans to release some "very good versions" later.
OpenAI's chief product officer Kevin Weil emphasized that the company needs to perfect the model, ensure security, accuracy in simulation and other aspects, and expand computing scale before pushing the next major upgrade of its video model Sora.
Moreover, Altman added that the company's image model DALL-E currently has no release plan, and the next model will be "worth waiting for". OpenAI also denied the plan to release its AI model Orion this year.
OpenAI strengthened its competition with Alphabet's Google by launching a "Ask Me Anything" session on Reddit and a new search feature for ChatGPT.
Regarding the recent series of executive departures, Srinivas Narayanan, the company's engineering vice president, said that although he was sorry to lose some partners, the company still has a talented team and has welcomed many new members.