“We are a few months away” from catching up with the American artificial intelligence giants, assures Arthur Mensch, CEO of the French company Mistral

The American OpenAI has published nearly 400 new mathematical results on complex problems in a few weeks. “We are a few months away” from catching up with the American artificial intelligence giants, assured Thursday October 8 on France Inter Arthur Mensch, CEO of the French company Mistral.

“In the coming months, we will become much better at mathematics,” adds Arthur Mensch, thanks to its new fundraising of three billion euros announced in September, but also thanks to its new model, Mistral Large 4, unveiled on Tuesday. A “technological leap”, he believes, “in the sense that we have trained it on much more computing capacity, and therefore […] it can run for hours and hours and carry out tasks autonomously”.

The CEO of Mistral explains that “effectively, once you have unlocked the agentic [AI system that produces tasks] and autonomous capabilities” of artificial intelligence, “you can put the models on the mathematics and it starts to solve them for you quite efficiently”. The new Mistral model, ML4, is precisely capable of carrying out "very long" tasks with "tangible results", details Arthur Mensch.

New capabilities which are “indispensable for all areas of fundamental research”, according to the CEO of Mistral. “We see it in mathematics, where it goes quickly because there is no physical constraint, but that will be the case in biology, in physics,” he lists. Hence the importance, according to him, of being able "in Europe, in particular, to produce this technology, because if we do not produce it, we will have to import it", particularly from the United States, where companies "like OpenAI, choose to keep their model in-house and to do their mathematical research in a completely opaque manner and without going through the community".

“If we want fundamental research in Europe, we absolutely must produce artificial intelligence”

Arthur Mensch

on France Inter

Asked about the likelihood that artificial intelligence could lead to the destruction of humanity, Arthur Mensch replied that "it doesn't mean anything." The alert was launched this summer by a former Anthropic researcher, and supported by a security manager of this company, who estimates the risk that AI will “kill all humans” within the decade at “more than 10%”. “This is absolutely not the major risk that we see today,” according to him.

The boss of Mistral rather denounces the risks of "concentration", in the field of AI, "from the fact that you have two or three companies in the world which will export a technology which creates most of the growth in the world". “There has been so much expense to be able to build the models that you have companies who have an interest in there not being too much competition, and in saying 'It's so dangerous that I would like it to be us who could deploy it throughout the world'”, worries Arthur Mensch. “That is a real problem since you then have a problem of dependence which will be created between Europe and the United States,” he adds.