OpenAI works on GPT-5: What will new neural network be capable of?

OpenAI is working on GPT-5, the next generation of its artificial intelligence model. CEO Sam Altman officially confirmed this information in an interview with the Financial Times.

According to him, the developers train the new model not only on publicly available data from the Internet, but also on the internal databases of large IT companies collaborating with OpenAI. GPT-5 is expected to be significantly more advanced than its predecessor, GPT-4. The changes will be so significant that Altman does not even dare to predict what exactly the new neural network will be capable of. In the past, the head of OpenAI promised to create artificial intelligence that is not inferior to humans in communication and mental analysis. Perhaps the new GPT-5 will be just that.

GPT-5, according to Sam Altman, will at a minimum be a much more complex and versatile model compared to previous versions, with a range of new capabilities and skills.

“Until we start training this model, it's a fun guessing game for us. We're trying to get better because from a security perspective, it's important to predict what's possible. I can’t say what exactly GPT-5 will be able to do that GPT-4 didn’t do,” Altman said.

At the same time, Altman did not specify the specific timing of the release of GPT-5.

Let us recall that OpenAI presented the new GPT-4 model on March 14, 2023. And in June, a company executive said that work on GPT-5 training had not yet begun. Due to public concerns about the rapid progress of artificial intelligence, the company has decided to suspend work on a successor to GPT-4 indefinitely.

In August, however, OpenAI filed a trademark application for "GPT-5" in the US with support from Microsoft. And on November 7, the company introduced the GPT-4 Turbo model, which is a more powerful version of ChatGPT, the base of which was updated until April 2023 (while the ChatGPT base was until 2021).