Europe needs to maintain the risk of open source or delaying AI models, Meta’s chief AI scientist said. It has advantages as it allows for faster progress.
According to Meta’s lead AI scientist, Europe needs to maintain open source or risk lagging behind artificial intelligence models.
Yann Lecun said Europe and some countries are trying to make the open source model illegal because they want to go ahead of their political rivals. This is a “big mistake.”
He made comments during a presentation at the AI Action Summit in Paris on Monday.
“When you study in secret, you fall behind,” said the French-American computer scientist. “The rest of the world will be open source and overtake you. That’s what’s happening right now.”
The open source AI model allows software to be shared free and openly for anyone for any purpose.
Lecun is a powerful advocate for large-scale language models of open source, reiterating that these systems should not be controlled by a small number of people or companies. He said that open source models can benefit everyone because progress is faster.
“We cannot come from a few companies from the US or China’s west coast,” Lecun said Monday.
His comments follow the release of AI models from Deepseek, a Chinese AI startup that rattles US technology and AI companies from late January. Third-party testing showed that the model was better than its Openai, Meta and other top developers, and the company said it was built for less money.
Deepseek’s R1 model is open source and can be downloaded and built by others on it.
“Deepseek benefits from open research and open source (such as Meta’s Pytorch and Llama),” the chief scientist wrote in a January thread post. “They came up with new ideas and built them on top of other people’s work.”
The meta AI model called llamas is mostly open source and is proposed by Lecun in the company. Originally founded as an open source AI company, Openai has recently moved to a closed model.
European AI companies using open source models include French startup Mistral and German Aleph Alpha.
The companies have criticized Europe’s proposal to regulate basic model manufacturers. Both its own and Italian legislators are pushing a framework that allows model makers to be self-adjusted and compete with US tech giants.
The European Union’s artificial intelligence law, approved in 2024, aims to combat the risks associated with powerful AI technologies. Recently, there has been debate on how to regulate basic models, such as large-scale language models, under the law.
