Sam Altman has confirmed that GPT-5 will not be released this year. The OpenAI CEO participated in a Reddit Ask Me Anything (AMA) on Thursday and was clear when asked about his company’s next AI model.
Please follow us on Facebook and Twitter.
When asked about the launch date for “ChatGPT-5 or its equivalent” and its possible features, Altman avoided raising false expectations. “We have very good launches planned for the end of the year, although nothing that we are going to call GPT-5,” he said.
The entrepreneur’s response is quite clever. He does not deny that OpenAI will launch a new cutting-edge AI before the end of 2024, but he dismisses the idea that it could be called GPT-5. However, this doesn’t completely rule out the possibility of a model with that name in the future.
In fact, when asked why GPT-5 was taking so long, he explained that they were prioritizing the deployment of O1 and its successors. “All of these models have become quite complex, and we can’t launch as many things in parallel as we would like. We also face many constraints and tough decisions about how to allocate our computing power to a lot of big ideas,” he said.
The startup’s leader has talked several times in the past about the technology that will succeed GPT-4, even unofficially referring to it as GPT-5. However, he has also acknowledged that he does not know what its final name will be.
OpenAI has several announcements planned for 2024, but none of them will be about GPT-5.
Months ago, Sam Altman stated that GPT-5—or whatever the next cutting-edge model is called—would put GPT-4 to shame, saying that GPT-4 would be remembered as “the dumbest AI we’ve ever used.” Since then, OpenAI has introduced technologies like GPT-4o and O1, but it seems they have a few more surprises in store.
Days ago, The Verge reported that OpenAI was planning to launch a new language model called Orion in December. According to the outlet, this AI would be the direct successor to GPT-4 and would be 100 times more powerful. It was also mentioned that, unlike previous technologies, it would not be included in ChatGPT from day one, and it had not yet been decided whether its final name would be GPT-5.
The report quickly reached Sam Altman, who strongly debunked it on X (Twitter). The OpenAI leader called it “fake news out of control” and added, “We have a lot of cool stuff coming. But I’m offended by how the media is willing to publish random fantasies.”
Beyond the reference to GPT-5, Altman’s AMA introduced other interesting ideas. For instance, he expressed his belief that it is possible to develop artificial general intelligence (AGI) using existing hardware.
He also addressed the delays with Sora.
He also discussed why the debut of Sora, the text-to-video generator, is taking so long. He explained that this is due to the significant time and resources needed for the inference process, as well as security concerns. “We have to refine the model. We need to ensure security, personifications, and other factors, and we have to scale up computing power,” he said.
The repeated mentions of the need for more computing power to advance the development of GPT-5, Sora, and other cutting-edge AI are not a coincidence. A recent report revealed growing discontent from OpenAI with Microsoft over delays in providing additional computing capabilities through Azure, as well as challenges in obtaining solutions from other providers.