Pod Insights No Priors Mistral 7B and the Open Source Revolution With Arthur Mensch, CEO Mistral AI About this episode 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.
Show Links:
Arthur’s Linkedin
Mistral
Mistral 7b
Retro: Improving language models by retrieving from trillions of tokens
Chinchilla: Training Compute-Optimal Large Language Models
Sign up for new podcasts every week. Email feedback to show@no-priors.com
Follow us on Twitter: @NoPriorsPod | @Saranormous | @EladGil | @ArthurMensch
Show Notes:
(0:00) - Why he co-founded Mistral
(4:22) - Chinchilla and Proportionality
(6:16) - Mistral 7b
(9:17) - 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
Transcript Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.
0:06 Sarah Guo? Open source AI models have completely changed the landscape of technology over the past year. One tiny team of ex DeepMind and Meta researchers in France has made a huge splash recently, Mistral. This week, Alad and I are joined by Arthur Mensch, the CEO and co-founder of Mistral, who recently released Mistral 7B, an Apache 2 licensed open source model that has changed people's mental models about what can be done with small models. Arthur, welcome to No Priors.
0:33 Speaker 01 Thank you for inviting me. Very glad to be here.
0:36 Sarah Guo? Okay, so just six months ago, when we met, you were leaving DeepMind to start Mistral. It takes real guts to look at the scale of dollars and compute that OpenAI and Google and others have amassed and say, "We want to play in this game too, and it's important we do." Tell us about the inspiration to start.
0:55 Speaker 01 Me, Guo, and Timothée were, I guess, pretty early in the field. And it had been 10 years that we had been doing machine learning. And we did know where to start from and how to make a good model with a limited amount of compute and money. Well, not so limited, but at least more limited than where we were coming from. And so I think that's why we got started. The various companies we were in move in two directions that we hadn't anticipated when we joined the company. And we decided that there would be a very good opportunity for creating something that would be a standalone company in Europe focusing on making AI better,
1:36 Speaker 01 focusing on making frontier AI, and focusing on making open source AI as a core value.
1:41 Sarah Guo? Maybe we can talk about each of those pieces. So 10 years in machine learning before, you were a co-author on the Chinchilla Scaling Loss paper. You worked on... The sort of mixture of experts' ideas early. Can you talk a little bit about what your research directions were at DeepMind?
2:00 Speaker 01 Yeah, so I come from an optimization background. So my focus has always been for the last 10 years to make algorithms more efficient and to use better the data that we have to make models with good prediction performances. And so when I arrived at DeepMind, I joined the LLM team that was 10 people at the time. And very quickly, I started to work on retrieval augmented models. So with a paper called Retro that I collated with my friend Seb Borneaux, who is still at DeepMind. The point there was to use like very large databases during pre-training so that we didn't force knowledge into the model itself.
2:42 Speaker 01 And we would tell the model that it would have access to an external memory anyway. And so it was working quite well. We could actually lower the perplexity, let's say. That's what you work on when you make LLMs. There were some limitations that I think the community has started to address quite well. And that was at a time when retrieval methods weren't really mainstream. Now they've become completely mainstream. So that's the first project I did. I worked on a sparse mixture of experts, also quite quickly because that was related to my topic of postdoc, which was optimal transport. So optimal transport is a setting where you have, I guess, tokens. You need to assess them, assign them to devices,
3:26 Speaker 01 and you need to make sure that there's some good assignment in between the two of them so that the devices don't see too many tokens. And as it turns out, the way you do it is with optimal transport. Optimal transport is a mathematical framework to do it correctly. And so I started to work on introducing this to sparse mixture of experts. And very quickly, we started to move on to scaling loads. So how do you actually take the method that is working at a certain scale and try to predict how that will evolve with the scale, the number of experts, the amount of data you see?
4:02 Speaker 01 And so that's a work I've done with many colleagues as well. On how do you adapt the scaling loads for dense parameters, for dense models, to a setting where you want to predict the performance not only with relation to the size of the model, but also the number of experts? Because that was the second thing I worked on. And then connectedly worked on Chinchilla, which is, I think, a major paper in the history of LLM. Also with Seb, Jordan, Laurent, many other people. Basically, the story was that everybody was training models on too few tokens.
4:41 Speaker 01 Because of the paper from 2020 that happened to be not very well executed. And so what we observed is that you could actually correct that. And so instead of training very large models on very few tokens, you should actually grow the number of tokens as you grow the size of the model, which if you think about it, makes a lot of sense. Because you don't want to have infinite size model looking at a finite size, a finite number of tokens. And similarly, you don't want to have a finite size model looking at an infinite number of tokens.
5:18 Sarah Guo? There must be some proportionality, yeah.
5:20 Speaker 01 Yeah, exactly. And that's something that we showed empirically. And I think that's something that was adopted very fast because it was like a pure win. For the same amount of compute, you would get a model that would be better, but also a model that will be four times cheaper to serve. So that was definitely a gain. And as it turns out, we didn't go far enough. That's what we did at Mistral. We realized that there was also a lot of opportunity in actually compressing models more. I mean, we've seen with Llama that it was actually possible. What we showed with Mistral 7B is that we weren't, we were definitely far away from the limit of compression. We somehow corrected that by making a model very small, super cheap to serve, super fast,
6:03 Speaker 01 running on your MacBook Pro, but still good enough to be useful. And so that's one of the first achievements we made in the company.
6:11 Speaker 02 Yeah, I think a lot of people were really impressed when Mistral came out with the 7B model. Because A, you did it very quickly, you know, you did it in a matter of a few months. But B, if you look at the cost of actually running these models, obviously, there's, there's the training side of it in terms of actually building the model. But then there's inference and so much of the cost day to day. If you're a user of OpenAI or a user of Llama or a user of Mistral's, how much it costs to actually run the models, the inference, and that's also, that's often driven by the size of the model. And so I think by coming out with these small models that were very performant, it really made a huge change to how people thought about what was possible. Is there anything that you can share in terms of where you think this performance is going to go? Or, you know, how you think about the sizes of models, both in terms of smaller, more performant models,
6:55 Speaker 02 as well as do you folks plan to build something very large, more like a GPT-4 or GPT-5 equivalent over time?
7:03 Speaker 01 Yeah, sure. So I think what we've seen, like in the, in 2022, 2021, you had a few companies that were really focusing on pushing the performance of models. And if you want to push the performance, the pure performance of models, you don't care about inference, because you're, well, you are not going to use the model. You're just going to see whether they're good or not, and that's really for scientific purposes. But then when you start thinking about deployment and enabling downstream applications, then you need to think about what it is going to cost in runtime. So you're not only worried about the upfront payment you need to make to get the model, but you're also worried about the runtime. And so I think the coefficient that you put in between the inference cost and the training cost really is really business dependent.
7:47 Speaker 01 And as a company that intends to have a valid business model, we think a lot about inference cost. We think that it's super important to get to a regime where inference is super cheap and so that you can run agents, you can run, you can basically use AI LLMs everywhere for all of your use cases and you're not blocked by cost, which is the case for the largest model currently. So that's definitely something we had in mind, and we knew that we could make a model, a 7 billion parameters of a model, very good. That's for sure. This is definitely not the end of the story. Now, the question is, do we, do we train bigger models? And the answer is obviously yes. There's still a limit to what a certain model size can do.
8:32 Speaker 01 This limit was, I think, underestimated. But if you want to get to more reasoning capabilities, you do need to move into larger models. The other thing about moving into larger models is that it enables you to train smaller models that are better, which is through variety of techniques like distillation or synthetic data generation. So these two things are quite related. If you want to make very strong, small models, you do need to have bigger models. And we are R and D training larger models for sure.
9:03 Sarah Guo? Can you tell us about your approach to data and annotations? Because we kind of talked about the other two dimensions.
9:10 Speaker 01 Yeah, so we've talked about compute, and obviously data is super critical. So we work from the open web. Well, we do a lot of work. I think we do a good job at getting some good data. The data quality is really what makes the model good. I mean, data algorithms, obviously, but that is super important. We put a lot of focus on that. And I think we do have a very good data sets, that's for sure. Data annotation is, I guess, another topic. It's not related to pre-training. When you pre-train a model, you really want to have the purest knowledge, the purest quality of data. When you want to align your model, instruct it, ask it to follow instructions, which is useful for many use cases because
9:55 Speaker 01 it makes it steerable, you do need to have a certain amount of either human-produced annotation or potentially machine-produced annotations. And so that's something that we start working on. We're not the top experts in the world in making good instruction fine-tuned models. We're definitely ramping up, and the team is getting better and better at that.
10:19 Sarah Guo? One of the things we haven't talked about is the fact that you guys are an open source company, which is very, very different from the other sort of labs working at the state of the art today. Like, why is that important?
10:33 Speaker 01 So if you look at, if you look back at the history of machine learning in the last ten years, it went very fast. I mean, we went from a poor cat-dog detector to something that basically looks human intelligent. And it's useful to remember how that happened. It happened because you had many academic labs, you had many industrial labs actually spending more money on different problems. And there was like full communication, almost full transparency until 2020. Like whatever was done in whatever lab, even in computing lab, was actually published at NERIP, was published at ICML. And every, every six months,
11:15 Speaker 01 we will gather and we'll get new ideas. Ideas will circulate and everybody will build on top of the work of others. And that's the way we went from something, well, potentially interesting to something very interesting. But then the issue is that around 2020, some companies started to be quite ahead on some field and realized that some value could be accrued. And then at that point, opacity made it back to, to the field. And I think that's, that's a cycle we've observed in software already, the cycle between openness and closeness. We are observing it again. We think that it's too early and we think
11:57 Speaker 01 it's, it's really damaging for the science to actually move into such an opaque regime where you have a few companies basically doing the same thing, just not communicating about it, spending billions of compute, doing exactly the same thing. And where really the technology we're looking at is not working completely yet. So still, it doesn't resonate well. Memory mechanisms are not very well understood. Causality mechanisms are not well understood. It's not super steerable. There's a lot of biases. I mean, it's, it's incomplete. There's many things to be done. We still need to invent new techniques. And how are you going to invent new techniques if nobody is speaking about it when?
12:39 Speaker 01 In order to invent new techniques, you need to still spend some some large amount of money to actually try things at scale. And the few companies that have the money to spend have not refused to communicate. That's something that we deeply regretted and and that's something that we are trying to change because we do have some substantial amount of money to actually spend on compute. We do have some good ideas. We know that there's a big community that is awaiting for AI players, for open source AI players to appear. And we're very grateful that Meta is moving into that direction. By doing what we do, by being much more open about the technology we create, we want to steer the community into a regime where things just work better, where things are safer because put under more scrutiny.
13:22 Speaker 01 And it's really our intention there is to, well, to to take that position and and to, well, to change the rules of the game because we don't think that this is moving into a proper direction.
13:34 Speaker 02 Yeah, it's very interesting because if you look at the the current discourse, the really big tech companies are claiming that open source AI is dangerous. And it feels like really a form of regulatory capture, right? They want regulators to step in so that they can constrain innovation and kind of control an industry. And, you know, the reality is, if you look at things like global health, global equity, open source is one of the biggest potential ways for all of humanity to benefit from this technology in a way that's transparent and open and people can really understand and and see. How how do you approach safety and policy and thinking about, you know, the right ways to think about safety in the context of open source?
14:13 Speaker 01 So I think we approach it from a very pragmatic point of view. So the question is, is open sourcing today a model that we do? Is it a dangerous thing? Is it actually enabling bad actors to to do to misuse the model? Is it giving them marginal capacity, like extra marginal capacities in in pursuing their bad endeavors? I think the answer to this question is is no. That's the that's my conclusion. We've seen, well, a lot of ideas around bioweapons, around, I don't know, nuclear terrorism and the like. And it's it's very interesting because if you actually assume good faith of these arguments,
14:59 Speaker 01 and I think for many, in many cases, people are in good faith, and that's an assumption that we always make. And if you try to go down their arguments, I mean, we realized, so that's what we did. And we really realized that there was really nothing to it. So nothing is showing that LLM is actually marginally better than a search engine to find knowledge on topics that would enable bad use. And the other thing is that it's not even proven, and it's certainly very, very likely not to be the case, that knowledge is not the bottleneck for the actual misuse that we're talking about. So we have two things. In order to demonstrate that open sourcing large language models is actually unsafe,
15:41 Speaker 01 you need to demonstrate that they have marginal improvement over Web Engine and that knowledge is the bottleneck for creation. So, and in the two cases, the answer to these two questions is no. And so that's, that means that we believe that we can open source models today and that actually it's the best way of putting things under high scrutiny so that we are ready for potential new generations of models that could be super intelligent. And in that case, I think we can rehave, we can rediscuss these premises. But today, we're really talking about the compression of knowledge that is widely available on the web.
16:22 Speaker 01 And so marginally speaking, we're not creating anything that is more dangerous than before. So, so I think that's, that's really there's a trade off, that's, there's a dynamic conversation to be had. That's what we discussed at the AI Safety Summit. For sure, this needs to be revisited as we, as model capacities build on. But today, going banning open source, preventing it from happening is really a way, well, to enforce regulatory capture, even though the actors that would benefit from it don't want it to happen. But by design, if you actually ban small actors from doing things in the most efficient
17:03 Speaker 01 way, which is open source, you do facilitate the life of the larger incumbents. And that's something that would be, I guess, detrimental to Mr. Life for sure.
17:13 Sarah Guo? What do you make of the arbitrary sort of compute and scale limits proposed?
17:19 Speaker 01 That's interesting. I don't exactly know how they came out, they, they came up with this threshold. It's a high threshold by any by any standard because if you compute like if you if you make the bad faith assumption that this is the float sixty four, it's actually gives you it's approximately three hundred million last run compute. So that's that's high. That's not something we can even afford and that we won't be able to afford for for the coming years. So it's it's high. It's very arbitrary because who tells you that beyond that ten to per twenty six, you end up with bad capacities. Models start to have to see the emergence of bad behaviors. That's definitely not proven.
18:05 Speaker 01 Relating capabilities to scale is also very approximate in the sense that it really depends on the data. The data set is super important. If you train your model on generating, I mean, there's a focus on bioweapons. So let's say if you if you want to if if if we want to prevent models from generating chemical compounds because we think it's an enabler of bad behaviors, which I have said, we don't think it is the case. But if you do want if you want to prevent that, well, you do need to to adapt your compute flop budget to to the data set in that we are working on. So as it turns out, that's what they did because they they actually made a specific flop
18:46 Speaker 01 budget for for biology, I think. So you can see the bioweapon narrative building on. But this is completely this, we should really focus on capabilities and not pre market conditions. And I think that's, I mean, to some extent, there's there's still consensus around that. So everybody knows that it's imperfect. It's a proxy, which is maybe fairly correlated. But definitely, we need to come up with agreeing on how we measure capabilities, agreeing on what capabilities we deem dangerous. And I think we don't agree with one another on that topic.
19:25 Speaker 01 But but these are really the should be the judge and not obviously pre market condition or the number of flops that you do.
19:32 Speaker 02 Do you know why there's such a focus on bioweapons? I I asked this as a I I worked for almost a decade as a biologist. And you know, when I look at how some of the complexity of building viruses or some of the complexity of actually doing these things, I'm surprised that there's so much focus in the community on that specific example. Do you have any sense of the origins of why people keep bringing that up? Is the because it's actually hard to translate. It's not some digital thing that you manipulate.
19:57 Speaker 01 I think it's a it's a very interesting question. Honestly, I don't have the answer. It's it's almost epistemology at that point. So how did this idea appear? And how did it get amplified by the policy people? And how did it actually ended up being heard by national security? And I think it started somehow by with GPT four and X D or something like on page forty six. Not exactly the same, these these numbers, but where they generated some chemical compound, which and then they said a small remark saying that, OK, maybe that's that's not the direction we want to take. We don't want to have a model that reasons about chemical compounds. The chemical compounds in question wasn't dangerous, but
20:42 Speaker 01 but then it it made this just observation, which is definitely expected. If you train on articles on biology, you're definitely going to be able to produce some chemical compounds. So expected observation. And then I don't somehow people build things on top of it. Nothing was observed. No scientific studies in proper form was published. But then policy papers started to to cite non scientific papers arguing that these were scientific evidences that the bioweapon narrative was actually true. And then policy papers started to cite the other policy papers. And all of a sudden, you end up with like fifty papers saying that for sure,
21:27 Speaker 01 bioweapon is going to blow us up. And and that this is what the policymakers read at the end. I think that's how we ended up where we are today. So that's some of the construction to be made. Unfortunately, I think the the open source community hasn't been vocal enough because they didn't see it coming. But right now, this is changing and and I'm very glad that it is.
21:50 Sarah Guo? I think it's mimetic. You have the factor of like, the world just goes through the COVID-19 pandemic.
21:56 Speaker 01 Yeah, I mean, the COVID trauma for sure played a role in that narrative. I mean, that's definitely a traumatism. Thirty million people died. That's that's definitely something we don't want to reoccur again. I don't think AI is going to be the one triggering the next pandemic. It's always going to be climate change. That's the that's that's the way they it was. And and that's probably where the focus should be. Instead of focusing on hypothetical, non-proven biological risk induced by token generators.
22:30 Sarah Guo? If bioweapons is not a like pragmatic concern in the, you know, visible future, there are real concerns around guardrails about what we want our AI models to actually generate. Like, how do you how do you think about that?
22:48 Speaker 01 Yeah, so I think this is a very valid concern. Models can output any kind of of text. And in many cases, you don't want it to output any kind of text. So when you build an application, you need to think on the guardrails you need you want to put on the model output and potentially also on the input. So you do need to have a system that filters input that are not valid, that you deem illegal, and output that are not valid or that you deem illegal. So the way you do it, in our mind, is that you do create the modular architecture that the application maker can use, which means you provide the raw model, so the model that hasn't been altered, to ban some of its output space.
23:32 Speaker 01 And then you propose new filters on top of that that can detect the output that that we don't want. So it can be pornography, it can be hateful speech. These things you want to ban when you have a chatbot, for instance. But these things you don't want to ban from the raw model. Because if you want to use the raw model to do moderation, for instance, you want your model to know about this stuff. So really.
23:59 Speaker 01 Assuming that the model should be well behaved is, I think, a wrong assumption. You need to make the assumption that the model should know everything. And then on top of that, have some modules that moderate and guardrail the model. So that's the way we approach it. And it's a way of empowering the application maker in making a well-guarded application. And it's, we think that it's our responsibility to make very good modules that allow guardrailing the model correctly. It's part of the platform. And we think it's, it's the way of, uh, we, there should be some, uh, health, healthy competition on that domain of different startups working on guardrailing the models.
24:40 Speaker 01 And the way you make this healthy competition is not by trusting a couple of companies to do their own safety. It's rather for, it's rather, the, the way you do it is to ask application makers to comply with some rules. So chatbots should not output hateful speech. And so that means that now the application makers need to find a good guardrailing solution. And now you have a competition where you have the, where there's some economic interest in providing the best guardrailing solution. And so that's the, that's the way we think the, the ecosystem should work. And that's the way we position ourselves. That's the way we build the platform with modular filters and modular mechanisms to,
25:24 Speaker 01 uh, to, to control the model well.
25:26 Speaker 02 It's great that you folks are being so thoughtful about that. I think when people talk about safety, they really talk about three topics and sometimes they talk past each other. One is this sort of moderation or textual based risk. And so that's risk of hateful content, legal content, bias, et cetera. There's a second class, which we talked about already, which is, uh, physical risk. It's things like bio weaponry or the ability of AI to help derail a train or interfere somehow. And then third is like existential or species risk. And that's when people start talking about AGI and new forms of life and, uh, you know.
26:00 Sarah Guo? Resource competition.
26:02 Speaker 02 Resource competition or aggregation or things like that.
26:05 Speaker 01 So first of all, it's, I think it's very important to address these three things separately and to hammer home that, oh, solutions exist for the first one, that the second one.
26:17 Speaker 01 There's no evidence that it actually exists at that point and no evidence that it will exist in the near future. The third point, that's, I think that's very philosophical. Obviously, if you can make a system of arbitrary complexity, it can start doing anything that you don't want it to do. We are not at a stage where the model has arbitrary complexity. And so this is very abstract to me.
26:42 Speaker 01 I think there's still, I mean, we'll move on to a world with agents and AI interacting with one another. And that will create a lot of complexity. The anticipating that complexity will necessarily yield to a collapse. We call it to collapse in machine learning when suddenly everything stops working. Because I know you you fell into a local minima. Well, it's it's it's unclear to me that complexity leads to a collapse. Usually, complexity leads to doing nothing because there's no self-organization and no willpower to build something. So I I'm not I'm not too worried about existential risk. Obviously, this is a dynamic conversation.
27:24 Speaker 01 If we can make a model which is growingly intelligent, then maybe you're at a singularity level. There's no evidence whatsoever that we are on the way of doing that, of making that happen. So I I'm I I think it's a very open conversation we should have. I personally don't see the scientific evidence and as a scientist, I trust only what I can see.
27:46 Speaker 02 Sure. And then I guess you mentioned agents, which I think is an area of a lot of activity right now. It feels like a number of things that are related to agents are still a little bit of ways in the future. In other words, it feels like an area with enormous promise, but it's still quite early. Are there any big technological innovations or things that you're working on that you think will really help expedite a world that moves more towards agents-based use in a more broad sense?
28:11 Speaker 01 I think making a model smaller is definitely a way to make agents work. Because one one problem you have with agents that very quickly start to, if you run an agent on GPT-4, you're going to run out of money very quickly. And so if you divide by a hundred, the, well, the cost of compute, well, you you can start to build the more interesting things. What we see with agents is mode collapse, so not very interesting mode collapse. They start repeating themselves and they fall into loops. So definitely, there's some research to be made there. There's some research to be made on. Making models more capable of reasoning and making them more capable of adapting the amount of compute they put onto the difficulty of the task. And this can be aged and solved somehow.
28:56 Speaker 01 So it's definitely an avenue of research that we're exploring.
28:59 Speaker 02 Yeah. Going back to Mistral, you know, one of the things that you've talked about a little bit is the platform that you've been building around the models that you train. Can you tell us a little bit more about that and some of the directions that's heading in?
29:10 Speaker 01 Yeah, so we know that hosting is not, hosting models isn't easy. There's a lot of work to be done on the inference side to make serving efficient. There's a lot of work to be done on the training side because you do need to come up with architectures that are memory efficient, for instance. That's what Mistral is good at that because it has this part of attention mechanism that makes it more memory efficient. So that's some work that you can do on the training side, but in order to reap all the benefits of a good model, you do need to work a lot on the inference part. And so we are actively working on that part to make it as efficient as possible, to build a platform that will be very cost efficient.
29:54 Speaker 01 And so you do need to have a good platform, well, good code, good inference code. The other thing that you can propose to customers is the fact that you do time sharing across customers. So when you want to play around with a model, it's, if you want to make it completely safe, you do, you should spin it up on an instance of a cloud, of a cloud provider. But if you just want to play around with it, you can access an API. It's going to be less costly because just a single H hundred can serve hundreds of customers. So I think there's some demand, a lot of demand for experimentation and APIs, and that's something that we started to build alongside the self hosted platform that we direct to other enterprise customers.
30:38 Sarah Guo? Your team is based in France. You have said before that you think there's a opportunity for a really important AI company that is French and European and serving the world. I don't know if that is like a mainstream point of view before the early success of Mistral. Can you can you talk about why you think that might work?
31:01 Speaker 01 I think some very strong point of Europe on that domain is talent. As it turns out, France, UK, Poland are very good at training mathematicians. And as it turns out, mathematicians are very good at making AI. So which means that there's a lot of French people and English people and Polish people in AI. And many of them wants to stay in Europe. Their family is there. It's the food is better. You have many advantages. I can't list them. Would be too long. And so obviously, we've been seeing the emergence of an AI ecosystem in London.
31:45 Speaker 01 I think very much thanks to DeepMind. And then in Paris, also thanks to DeepMind and to Meta that settled a lab there and to a lot of entrepreneurs that started to come back. So today, we have, I think, hundreds of startups in Paris. So this is not the level of the Silicon Valley, obviously, but but we start to have an ecosystem in place with investors, with operators investing as well. So it's the same kind of flywheel that made San Francisco and the Bay Area successes is starting to spin in France. And I'm very glad that we are participating to it.
32:26 Sarah Guo? This has been a great conversation, Arthur. I always find you inspiring. I'm very grateful to be an investor. Thanks for doing this.
32:34 Speaker 01 Well, thank you for having me and looking forward to seeing you soon.
32:38 Sarah Guo? Find us on Twitter at NoPriorsPod. Subscribe to our YouTube channel if you want to see our faces. Follow the show on Apple Podcasts, Spotify, or wherever you listen. That way you get a new episode every week. And sign up for emails or find transcripts for every episode at no-priors.com.
Previous What is Digital Life? with OpenAI Co-Founder & Chief Scientist Ilya Sutskever Next AI Agents That Reason and Code with Imbue Co-Founders Kanjun Qiu and Josh Albrecht Page updated October 12, 2026 · transcript and summary are machine-generated