No Priors Ep. 40 | With Arthur Mensch, CEO Mistral AI
Ғылым және технология
Open Source fuels the engine of innovation, according to Arthur Mensch, CEO and co-founder of Mistral AI. Mistral is a French AI company which recently made a splash with releasing Mistral 7B, the most powerful language model for its size to date, and outperforming much larger models. Sarah Guo and Elad Gil sit down with Arthur to discuss why open source could win the AI wars, their $100M+ seed financing, the true nature of scaling laws, why he started his company in France, and what Mistral is building next.
Arthur Mensch is Chief Executive Officer and co-founder of Mistral AI. A graduate of École Polytechnique, Télécom Paris and holder of the Master Mathématiques Vision Apprentissage at Paris Saclay, he completed his thesis in machine learning for functional brain imaging at Inria (Parietal team). He spent two years as a post-doctoral fellow in the Applied Mathematics department at ENS Ulm, where he carried out work in mathematics for optimization and machine learning. In 2020, he joined DeepMind as a researcher, working on large language models, before leaving in 2023 to co-found Mistral AI with Guillaume Lample and Timothee Lacroix.
00:00 - Why he co-founded Mistral
04:22 - Chinchilla and Proportionality
06:16 - Mistral 7b
09:17 - Data and Annotations
10:33 - Open Source Ecosystem
17:36 - Proposed Compute and Scale Limits
19:58 - Threat of Bioweapons
23:08 - Guardrails and Safety
29:46 - Mistral Platform
31:31 - French and European AI Startups
Пікірлер: 18
Cet accent incroyable! Go Mistral you guys rule!
and also "food is better in europe" mic drop
Amazing podcast!!!
it was amazing! Thanks very much! keep up the good work in the old world/continent ;)
Great interview!
Can't help noticing that he was more tangible and open than Sutskever in his interview a few months ago
Its real good to hear what Arthur is talking about❤ . Wish I could share my ideas with him as well!🤷
French tech dudes always appear when Americans expect it the less. These guys are kind of the Daft Punk of adaptative LLM models.
quality data set at first was a smart choice...like cooking 3 stars . Good raw ingredients. Architecture come second.
@mikelastname
4 ай бұрын
Interesting point of view. Without good architecture, your good data gets abused. To stretch your analogy, cooking your fine ingredients on a candle is unlikely to yield Michelin star food. I feel like they are both similarly important and that your best bang for buck comes from incrementally advancing your architecture and your data quality in a known strategic direction.
The vocal fry is crazy. But I still love the podcast.
The large models show emergent behaviours, I think that's why they started closing things down. That by definition means it is hard/impossible to know how powerful they are and what they are capable of, which means leaving them in the wild could be considered irresponsible. If these tools are already having an impact for good (they seem to be) then saying that they can't be causing or used for bad is nonsense. Of course we all want the big models to be open source, I would love to dig around in there, but I do understand the caution.
@Bvic3
4 ай бұрын
No. They closed models because it became good enough to make commercial products instead of just being fun research. The same happened with image generation. There was insane Twitter+Colab sharing ... Until the images became so good that you had Midjourney making money. And the sharing stopped.
Bioweapons are such a topic because you can potentially kill extreme amount of people with small amount of resulting material itself. Logistics of usage are not like nuclear weapons where you need to carry tons of stuff to have big impact but result is still limited globally in the end. Here people are worried that one competent actor can impact whole world with easy endpoint usage logistics of bioweapons.
@zacboyles1396
5 ай бұрын
The problem with bio weapons isn’t the competent actors, it’s the thoroughly incompetent ones, so blind with making pharmaceutical profit on the only product that grants a person legal immunity around the globe if they so much as bump against a UPS box containing the product. Take a look at whose in charge of the next “response” and you will find all of the people who made the absolute worst decisions for citizens and at the same time those same decisions made them record profits. I think humanity has bigger issues than theoretical biohackers in a garage somewhere.
That vocal fry is incredibly infuriating to listen to