How France wants to play a leading role in AI
It was a soap opera that kept investors and developers around the globe on the edge for several days. With its constant stream of new developments, OpenAI somewhat overshadowed Kyutai. Kyutai is a new research laboratory for artificial intelligence (AI), representing France's latest attempt to develop sovereign AI technology and maintain Europe's independence from the United States and China.
Behind Kyutai are three billionaires: Xavier Niel (Iliad), Rodolphe Saadé (CMA CGM), and former Google CEO Eric Schmidt. Niel and Saadé each invest 100 million euros in the non-profit research lab, which also receives financial support from Schmidt. Kyutai has a total budget of 300 million euros. This initiative is not the only one in France within the AI field, as several startups have emerged in recent months, putting Paris on the international map as a rising hub for AI.
"I don't want us to sit at the back of the AI train"
"I don't want us to sit at the back of the AI train," said Niel in September. At that time, he announced his intention to invest 200 million euros in the sector to create an ecosystem for AI in France. It is similar to what he has done with Station F, the world's largest startup campus, and the computer programming school Ecole 42 in Paris.
Niel, Saadé, and Schmidt are also investors in Mistral AI, a startup for generative artificial intelligence founded earlier in the year by Timothée Lacroix, Guillaume Lample, and Arthur Mensch. In September, Mistral released its first significant language model and could become a unicorn after the next funding round. It aims to launch the first models for businesses early next year.
Emergence of national AI champions
The two French billionaires have also invested in Poolside AI, which, like Mistral and Light On, builds AI language models parallel to OpenAI, Meta, and others. Poolside is backed by American entrepreneurs Eiso Kant and Jason Warner, who have made Paris their base. Industry experts believe that their decision to settle in the French capital has a signaling effect for others.
France aims to support the emergence of national AI champions, stated Digital Minister Jean-Noël Barrot at Kyutai's launch. "My goal is to maximize AI innovations to compete with the world's greatest forces so that France doesn't miss the next big revolution," he told the business newspaper "Les Echos."
To achieve this, France has implemented various plans in the AI sector since 2017, investing €2.5 billion in research, education, and business support. The Jean Zay supercomputer, which contributed to the development of the Bloom AI language model, is set to receive an additional €50 million from the government.
Bringing back researchers
"Sam Altman, his team, and their talents are welcome in France if they want," Barrot wrote on Twitter after the removal of the OpenAI CEO. Bringing back AI researchers and developers to Europe is also the stated goal of Niel and Saadé. According to Niel, the best talents from France end up at OpenAI, Microsoft, and Google, coming from prestigious institutions like École Polytechnique or École Normale Supérieure in Cachan. "They are good, but many have gone elsewhere. The idea is to bring them back and preserve our talents."
The Kyutai team consists of six researchers, all of whom have previously worked for major US technology companies such as Apple, Google, Meta, or Microsoft. Édouard Grave, Hervé Jegou, and Alexandre Défossez were with Meta, and Laurent Mazaré and Neil Zeghidour with Deep Mind. Patrick Pérez recently led the AI laboratory at automotive supplier Valeo. Kyutai's scientific advisory board includes Bernhard Schölkopf from the Max Planck Institute in Tübingen, Meta's Chief AI Scientist Yann LeCun, and Yejin Choi from the University of Washington.
Europe significantly lags behind the US and China
Jegou most recently headed Meta's AI research lab, Fair, which was founded in Paris in 2015. Besides Meta, Deep Mind, the AI startup acquired by Google in 2014, also operates a research lab in the French capital. As do parent company Alphabet, IBM, Fujitsu, and Samsung.
"We don't want our children to depend on algorithms invented with different rules outside the borders of Europe," notes Niel. "The idea is to create an environment for researchers in France, in Europe, that complies with our rules, our standards."
However, Europe still significantly lags behind the US and China in AI. Lorenzo Ancona and Niccolo Bianchini from the Robert Schuman Foundation think tank argue that, in terms of AI development, the European Union is presently, at best, a secondary participant. This is attributed to challenges such as inadequate investments, an unfinished unified European market, limited appeal for European talents, data shortages, and a convoluted regulatory landscape.
Insufficient access to investment capital
Ancona and Bianchini express concern over the insufficient access to investment capital, especially when comparing it to other regions. They highlight that venture capital investments in AI in the US from 2010 to 2020 were ten times greater than those in the Eurozone. The duo attributes the investment shortfall to the strict regulations of the European banking system, which restricts risk investments, and the absence of prominent European AI companies.
In a study released this spring, the liberal think tank Institut Montaigne suggests investing 1 billion euros to position Europe as a pivotal player in AI and narrow the gap with the US and China. Industry observers note that venture capital funds have increasingly focused on Paris as a hub for AI. British and American players based in London reportedly spend two days a week in Paris, often meeting at the popular startup hotel Hoxton.
The Parisian AI scene can still only dream of the resources of tech giants in the US. Mistral, France's highest-valued AI startup so far, just raised 105 million euros in a funding round in June, while Aleph Alpha from Germany secured over 500 million dollar. Mistral is now expected to hope for 400 million euros in a new funding round.