I lean skeptical here, so I’ll play devil’s advocate. As much as I would like to see a more multipolar AI landscape, I don’t see how Mistral can realistically come to parity with American or Chinese competitors. America is established enough that many of the best minds in Europe are willing to go there in spite of political turmoil, and has a strong culture of young people eager to prove themselves who went to college surrounded by very valuable resources that don’t exist elsewhere at comparable scale. Both France and America are playing catch-up industrially, having to reshore after a long period in which offshoring was believed to be the future, amidst a regulatory environment that can be a major obstacle, but in America, all of the financial power that depends on their superpower status is willing to push for success in that difficult undertaking. The EU, as far as I’ve seen, treats AI as more of a novelty, and is starting with a much shorter stock of resources and experienced personnel on top of that. All of these advantages compound, and are supplemented further by the fact that anyone with something major to contribute and broadly Western interests in mind is likely to consider supplementing America’s present advantage to be the stronger play.
China, meanwhile, has scale of its own, and the implicit support of everyone that would prefer to see America unseated as the sole superpower. They also have a large population, a rapidly-scaling higher education system that is increasingly competitive internationally, and a willingness to approach America’s present advantages pragmatically, leveraging an international presence in the U.S. university system well to close gaps where they exist. They also have a much safer supply chain, less dependent on foreign trade, which could rapidly become decisive if the world’s political situation deteriorates.
I realize the LLM leaderboard isn’t the be-all-end-all, but it does generally line up with how different LLMs have performed when I’ve used them, or stress-tested them when writing a paper. Of the 25 top models, 16 are from a variety of American companies (8 for OpenAI, 3 for XAI and Anthropic, 2 for Google), and the remaining 9 are from a wide array of Chinese startups. Mistral’s first appearance is in the 27th position, with a score of 52 on their evaluation, relative to 68 for OpenAI’s top model, 67 for the top Chinese model, and 65 for XAI’s top model, with Anthropic receiving a score of 63. Gaps like these are self-reinforcing, especially without China’s vast population of students to draw on (with a natural degree of exclusivity) or America’s frighteningly-efficient talent pipeline of schools, startups, VCs, and headline-making compensation packages.
Yeah, agree with most of this. I added a not saying that it’s the narrative that I think is shaping France actions, rather than my view of the situation.
I agree that Mistral cannot come to parity on technical prowess. I think that from the sovereignty angle, it still can be a success if they make models that are actually useful for industry and are trusted by French national security agencies, which seems to be more like Mistral’s goal lately.
I lean skeptical here, so I’ll play devil’s advocate. As much as I would like to see a more multipolar AI landscape, I don’t see how Mistral can realistically come to parity with American or Chinese competitors. America is established enough that many of the best minds in Europe are willing to go there in spite of political turmoil, and has a strong culture of young people eager to prove themselves who went to college surrounded by very valuable resources that don’t exist elsewhere at comparable scale. Both France and America are playing catch-up industrially, having to reshore after a long period in which offshoring was believed to be the future, amidst a regulatory environment that can be a major obstacle, but in America, all of the financial power that depends on their superpower status is willing to push for success in that difficult undertaking. The EU, as far as I’ve seen, treats AI as more of a novelty, and is starting with a much shorter stock of resources and experienced personnel on top of that. All of these advantages compound, and are supplemented further by the fact that anyone with something major to contribute and broadly Western interests in mind is likely to consider supplementing America’s present advantage to be the stronger play.
China, meanwhile, has scale of its own, and the implicit support of everyone that would prefer to see America unseated as the sole superpower. They also have a large population, a rapidly-scaling higher education system that is increasingly competitive internationally, and a willingness to approach America’s present advantages pragmatically, leveraging an international presence in the U.S. university system well to close gaps where they exist. They also have a much safer supply chain, less dependent on foreign trade, which could rapidly become decisive if the world’s political situation deteriorates.
I realize the LLM leaderboard isn’t the be-all-end-all, but it does generally line up with how different LLMs have performed when I’ve used them, or stress-tested them when writing a paper. Of the 25 top models, 16 are from a variety of American companies (8 for OpenAI, 3 for XAI and Anthropic, 2 for Google), and the remaining 9 are from a wide array of Chinese startups. Mistral’s first appearance is in the 27th position, with a score of 52 on their evaluation, relative to 68 for OpenAI’s top model, 67 for the top Chinese model, and 65 for XAI’s top model, with Anthropic receiving a score of 63. Gaps like these are self-reinforcing, especially without China’s vast population of students to draw on (with a natural degree of exclusivity) or America’s frighteningly-efficient talent pipeline of schools, startups, VCs, and headline-making compensation packages.
Yeah, agree with most of this. I added a not saying that it’s the narrative that I think is shaping France actions, rather than my view of the situation.
I agree that Mistral cannot come to parity on technical prowess. I think that from the sovereignty angle, it still can be a success if they make models that are actually useful for industry and are trusted by French national security agencies, which seems to be more like Mistral’s goal lately.