Arthur Mensch, tall and thin with a shock of unkempt hair, arrived at a speech last month at a sprawling tech center in Paris dressed in jeans and a bicycle helmet. He had a modest outlook for a person European officials are counting on to help propel the region into a high-stakes showdown with the United States and China over artificial intelligence.
Mensch, 31, is the CEO and founder of ai/” title=”” rel=”noopener noreferrer” target=”_blank”>Mistral, considered by many to be one of the most promising rivals to OpenAI and Google. “You have become the model of ai in France,” Matt Clifford, a British investor, told him on stage.
A lot depends on Mensch, whose company rose to prominence just a year after he founded it in Paris with two college friends. As Europe struggles to gain a foothold in the ai revolution, the French government has pointed to Mistral as its best hope for creating a standard-bearer and has pressed European Union policymakers to help ensure the company's success.
artificial intelligence will rapidly enter the global economy in the next decade, and policymakers and business leaders in Europe fear that growth and competitiveness will suffer if the region does not keep pace. Behind their concerns is a conviction that ai should not be dominated by tech giants, such as Microsoft and Google, that could forge global standards at odds with the culture and politics of other countries. At stake is the larger question of what models of artificial intelligence will end up influencing the world and how they should be regulated.
“The problem with not having a European champion is that the roadmap is set by the United States,” said Mensch, who just 18 months ago was working as an engineer at Google's DeepMind lab in Paris, building ai models. His co-founders, Timothée Lacroix and Guillaume Lample, also in their 30s, held similar positions at Meta.
In an interview at Mistral's spartan, whitewashed offices facing the Canal Saint-Martin in Paris, Mensch said it was “not safe to trust” American tech giants to set ground rules for a powerful new technology that would affect millions of lives. .
“We cannot have a strategic dependency,” he said. “That's why we want to make a European champion.”
Europe has struggled to produce significant technology companies since the dotcom boom. While the United States produced Google, Meta and amazon, and China produced Alibaba, Huawei and ByteDance, owner of TikTok, the European digital economy did not achieve results, according to a report in the French newspaper artificial intelligence Commission. The 15-member committee, which includes Mensch, warned that Europe was lagging behind on ai but said it had the potential to take the lead.
Mistral's generative ai technology enables companies to launch chatbots, search functions and other ai-based products. It has surprised many by building a model that rivals the technology developed at OpenAI, the American startup that kicked off the ai boom in 2022 with the ChatGPT chatbot. Mistral, named after a strong wind in France, has quickly gained ground by developing a more flexible and cost-effective machine learning tool. Some large European companies are starting to use its technology, including Renault, the French automobile giant, and BNP Paribas, the financial services company.
The French government fully supports Mistral. President Emmanuel Macron called the company an example of “French genius” and invited Mensch to dinner at the Elysee presidential palace. Bruno Le Maire, the country's finance minister, frequently praises the company, while Cédric O, France's former digital minister, is an adviser to Mistral and owns shares in the new company.
The French government's backing is a sign of the growing importance of ai. The United States, France, Britain, China, Saudi Arabia and many other countries are trying to strengthen their domestic capabilities, triggering a technological arms race that is influencing trade and foreign policy as well as global supply chains.
Mistral has become the strongest European contender in the global battle. However, many question whether the company will be able to keep up with large American and Chinese competitors and develop a sustainable business model. In addition to the considerable technological challenges involved in building a successful ai company, the computing power required is staggeringly expensive. (France says its cheap nuclear power can meet energy demand.)
OpenAI has raised $13 billion and Anthropic, another San Francisco company, has raised more than $7.3 billion. Mistral has so far raised approximately 500 million euros, or $540 million, and earns “several million” in recurring revenue, Mensch said. But in a sign of Mistral's promise, Microsoft took a small stake in February, and Salesforce and chipmaker Nvidia backed the new company.
“This could be one of the best opportunities we have in Europe,” said Jeannette zu Fürstenberg, CEO of General Catalyst and founding partner of La Famiglia, two venture capital firms that invested in Mistral. “Basically, you have a very powerful technology that will generate value.”
Mistral shares the view that ai software should be open source, meaning that the programming codes should be available for anyone to copy, modify or reuse. Supporters say allowing other researchers to see the code will make the systems more secure and boost economic growth by accelerating their use among businesses and governments for applications such as accounting, customer service and database searches. This week, Mistral released the latest version of its model online for anyone to download.
OpenAI and Anthropic, on the other hand, keep their platforms closed. They argue that open source is dangerous because it has the potential to be co-opted for bad purposes, such as spreading disinformation or even creating destructive ai-powered weapons.
Mensch dismissed such concerns as the narrative of “a fear-mongering lobby” that includes Google, Microsoft and amazon, which he said sought to cement their dominance by persuading policymakers to enact rules that would crush their rivals.
The biggest risk of ai, Mensch added, is that it drives a revolution in the workplace, eliminating some jobs and creating new ones that will require retraining. “It's coming faster than previous revolutions,” he said, “not in 10 years, but rather in two.”
Mensch, who grew up in a family of scientists, said he was fascinated by computers from an early age and learned to program when he was 11. He played video games avidly until age 15, when he decided he could “do better things.” with my time.” After graduating from two elite French universities, the École Polytechnique and the École Normale Supérieure, he became an academic researcher in 2020 at France's prestigious National Center for Scientific Research. But he soon turned to DeepMind, an artificial intelligence lab acquired by Google, to learn about the industry and become an entrepreneur.
When ChatGPT burst onto the scene in 2022, Mensch teamed up with his college friends, who decided they could do the same or better in France. In the company's spacious workspace, a group of sneaker-wearing scientists and programmers now busily tap away at keyboards, encoding and feeding digital text scraped from the Internet, as well as reams of 19th-century French literature, no longer subject to copyright. From author. law, in the great linguistic model of the company.
Mensch said he was uncomfortable with Silicon Valley's “very religious” fascination with the concept of artificial general intelligence, the point at which, technology leaders such as Elon Musk and Sam Altman believe, computers will surpass the cognitive ability of humans. , with potentially dire consequences. .
“All of AGI's rhetoric is about God's creation,” he said. “I do not believe in God. I am a strong atheist. So I don't believe in AGI.”
A more imminent threat, he said, is what American ai giants pose to cultures around the world.
“These models are producing content and shaping our cultural understanding of the world,” Mensch said. “And it turns out that French values and American values differ in subtle but important ways.”
With his growing influence, Mensch has intensified his calls for lighter regulation, warning that restrictions will harm innovation. Last fall, France successfully lobbied in Brussels to limit regulation of open source ai systems in the European Union's new artificial intelligence Law, a victory that helps Mistral maintain a rapid pace of development.
“If Mistral becomes a major technical power,” said O, the former digital minister who led the lobbying effort, “it will be beneficial for all of Europe.”