Meta executive Nick Clegg calls AI language models 'quite stupid' as the company announces plans to give away its AI tech

• Nick Clegg, Meta's president of global affairs, says AI language models are still "quite stupid."
• Clegg said some of the hype around AI overestimated its abilities.

Meta executive and former UK deputy prime minister Nick Clegg isn't sure how intelligent AI really is.

In an interview with the BBC's Today Programme, Clegg, who is Meta's president of global affairs, said some of the hype around AI overestimated the tech's abilities.

"Hype has somewhat run ahead of the technology," he said, referring to AI language models. "In many ways, they're quite stupid."

Clegg said AI models such as OpenAI's GPT-4 or Google's PaLM 2 fall "far short" of a level of intelligence that can think autonomously, despite some experts' concerns.

Representatives for Meta did not immediately respond to Insider's request for comment, made outside normal working hours.

The remarks came shortly after Meta announced it was open-sourcing its own AI model, LLaMA 2. The move will allow the large language model to be used for free by researchers or commercial entities.

The announcement is part of a partnership with Microsoft, which is seen by some as an attempt to try and best OpenAI, a company that has benefitted from a series of major Microsoft investments.

Meta announced the news on Wednesday as part of Microsoft's Inspire event. In a news release, Microsoft called the relationship with Meta a "growing partnership."

Meta, which has been keen not to be left out of the AI boom, has been making the technology a priority in recent months. The company initially released a version of LLaMA exclusively to researchers in February and has touted other powerful AI products that have not been released to the public.

Meta's foray into AI has been somewhat stalled by the company's lack of consumer-facing AI products, an aspect that appeared to make it less visible in the space.