Sam Altman disputes reports that GPT-5 is being trained

OpenAI is not training the fifth version of the GPT large language model. Instead, developers are focusing on expanding GPT-4’s capabilities, said Sam Altman, the startup’s CEO.

The OpenAI CEO spoke at an event at the Massachusetts Institute of Technology. One of the attendees asked Altman about a letter published by the Future of Life Institute urging a pause in AI research.

Altman said the initiative lacks a technical nuance about where to pause. He also rejected reports that OpenAI is allegedly training GPT-5.

“In an earlier version of the letter it said that we are training GPT-5 right now. We are not doing this and do not plan to for some time. … We are building things on top of GPT-4, which, in my view, have all sorts of safety issues that we need to address,” he said.

However, Altman lent support to several lines of the open letter, noting the importance of ensuring the safety of AI models and their alignment with human values.

In March, OpenAI presented the multimodal GPT-4 model. Unlike its predecessors, the new model analyses not only text but also images.

In the same month, more than a thousand industry experts called for a pause in the development of language models more powerful than GPT-4.