Frédéric JacobsASML rumored to be investing €1.3bn into Mistral's €1.7bn Series C.
Really curious about the long game here. Bring computational lithography aspects, currently provided by NVIDIA in-house/in partnership with Mistral?
👏 “ASML and Mistral announced a strategic partnership based on a long-term collaboration agreement to explore the use of AI models across ASML’s product portfolio as well as research, development and operations, to benefit ASML customers with faster time to market and higher performance holistic lithography systems.”
https://www.asml.com/en/news/press-releases/2025/ASML-Mistral-AI-enter-strategic-partnership
The ASML and Mistral partnership is going to cover topics like using AI to better control the plasmas used for EUV.
I hope they will publish some details about these efforts 🤞
https://www.tf1.fr/tmc/quotidien-avec-yann-barthes/videos/invite-arthur-mensch-le-francais-qui-fait-trembler-les-geants-americains-de-lia-73247323.html
I hope they will publish some details about these efforts 🤞
https://www.tf1.fr/tmc/quotidien-avec-yann-barthes/videos/invite-arthur-mensch-le-francais-qui-fait-trembler-les-geants-americains-de-lia-73247323.html
Invité : Arthur Mensch, le Français qui fait trembler les géants américains de l'IA - Quotidien | TMC
Arthur Mensch, co-fondateur et CEO de Mistral AI, est l'invité de Quotidien. À 33 ans, il a créé la première "désacorne" française de l'intelligence artificielle avec une levée de fonds record de 1,7 milliard d'euros. Il a expliqué sa philosophie "open source" et pourquoi la France peut concurrencer les géants américains comme OpenAI.
A lot of LLM experimentation is mostly hype, and unlocks very little practical value. And then there's stuff like this.
At a time where chip production is experiencing throughput issues, the Mistral x ASML partnership is delivering with an AI-powered diagnostic tool that identifies lithography EUV scanner issues 120x faster.
pettter@fj Really hard to tell what the actual "120 times faster" thing refers to here: The system being described seems to me to be about detecting potential issues in submitted code, which admittedly is probably a much better use of LLMs in programming than most, since it's about applying _more scrutiny_ to _existing code_, rather than producing _more code_ with little or no scrutiny.
Cassandrich@fj "AI powered" here is equating two completely different things, the useful one of which is not "AI", for marketing purposes. 🙄
Witton says FUCK ICE@fj I don't know much about AI, but this I think is a good example of how tailored applications can be very powerful and useful, a totally different bucket from LLMs