The Story of Arthur Mensch: How to Build a Startup with $2.1B Value in 9 Months
Exploring the path of Mistral AI and testing the latest ChatGPT competitor!
👋 Hey, I’m Stepan and welcome to a ✨ subscriber-only edition ✨ of Creators’ AI. By subscribing, you directly support Creators' AI's mission to deliver top AI insights & practical knowledge without ads or clutter. Your subscription allows us to grow our dedicated team and curate the most important AI Tools, Stories, and Tutorials in one place. - Stepan
Intro
If you're a long-time reader of Creators' AI, you know that we emphasize practical AI-related case studies, reviews, guides, and tools. We assume that every platform can be used to develop existing projects, improve work, or create something new. And this approach comes with some limitations.
We rarely consider large AI companies because such cases are difficult to translate into our experience. However, there are exceptions everywhere. You've probably heard about Mistral AI. It's now being talked about on every corner. Its founder, Arthur Mensch, is already being called the second Sam Altman, and his company is a full-fledged competitor to OpenAI. And that's in just nine months since its founding.
Let’s look at Mensch's journey, discuss how he managed to get the attention of investors (heck, over $2.5B value in less than a year!) and the AI community, and what Le Chat, an alternative to ChatGPT 4 by Mistral AI, can do.
🔗 Are you already subscribed and want to ensure you don’t miss our newsletters? Move our newsletter from “Promotions” to “Primary” in your inbox! (video instructions below)
Everything Starts With an Idea
Like the story of any important startup, Arthur Mensch's story began with an idea. And that idea runs leitmotif through Mensch's entire background. In his (very brief) LinkedIn profile description, he describes himself as a scientist and entrepreneur. And that very accurately captures the essence of our hero today.
From childhood, he was driven by the idea of making our lives better through scientific advances.
He graduated from a French school focused on the deep study of math and science. As his doctoral thesis, Mensch chose ways to scale software to analyze three-dimensional brain images from a functional magnetic resonance imaging system. Using his accumulated experience and knowledge, in 2020, Mensch became part of DeepMind and eventually earned the position of Senior Research Scientist.
DeepMind is one of Google's main artificial intelligence divisions. Its programs have beaten the most powerful programs playing Go, Сhess and Shogi, and then made significant advances in the problem of AI-assisted protein folding.
For two years, Mensch was one of the prominent faces on the issue of developing LLMs. However, over time, he realized that Google was moving too slowly and perhaps in the wrong direction. Among other players, the enthusiast didn't find what he sought. So last year, Mensch reached out to his counterparts at Meta Platforms, Timothée Lacroix and Guillaume Lample, to found Mistral AI.
Together, they turned Mensch's idea into a goal: "To make AI useful."
And to make this one open source, by the way.
Sharing is caring! Refer someone who started a learning Journey in AI!
Are You Sure Big Tech Will Lead the AI Race?
Yes, we have every reason to believe that big tech companies will be the main growth drivers for AI. But who is the spark that will ignite the fire? The founders of Mistral AI don't think money plays the deciding role here.
The startup started by developing products based on large language models that both enterprises and consumers can use. The message was to create flexible and customizable tools for various industries. Even during stealth mode last May, investors loved the idea.
In the four weeks since its founding, Mistral has raised $113 million in venture capital funding from major European funds. Former Google CEO Eric Schmidt was among the investors who backed the startup. The valuation of the business increased to $260 million.
At the same time, Mistral made an important announcement: all of its products would be based on publicly available data. In this way, the company guaranteed transparency for its customers and was able to avoid any legal problems in advance.
Mistral AI continued to improve its products without sharing details with the public. And a few months ago, in early December 2023, the startup was back in the news headlines. The company said it raised another large round of $417 million, announced the Mixtral 8x7B language model, and opened beta access to its platform. The Mensch startup began to step on the heels of OpenAI, and its team grew from 30 to 70 people.
What is interesting is that the startup did not hold a press conference. The release of a large language model became known thanks to the publication of a download link in Torrent on the X.
The Mistral services became available in three variants: tiny, small, and medium. These platform services allowed developers to leverage generative text models, customizing features based on the price-to-performance ratio.
Just a few days ago, Mensch made another impressive announcement - Mistral AI is now working with Microsoft. The American corporation announced that it is investing in the French startup and will also make Mistral AI's large language models available on its Azure platform.
Against this backdrop, Mistral AI officially unveiled the Le Chat and the new Mistral Large language model, which was created as an alternative to GPT-4.
This is where the startup's annals come to an end, as we have reached the present day. That means it's time to discuss Mistral AI's technological capabilities.
How not to get lost in the choice of AI chatbots? Which of them should be your daily driver? This newsletter will help you decide:
Subscribe to stay in tune with AI breakthroughs!
Le Chat Capabilities
So, what is Le Chat? It's straightforward - an entirely free chatbot built on a model with a relatively simple name, Large. Traditionally, we should quickly run through its benefits and technical features.
Mistral AI highlights four features of its model:
Natively fluent in English, French, Spanish, German, and Italian, with a nuanced understanding of grammar and cultural context.
It has 32K tokens context window allows precise information recall from large documents.
Its precise instruction-following enables developers to design their moderation policies.
It is natively capable of function calling. This, along with constrained output mode, implemented on la Plateforme, enables application development and tech stack modernization at scale.
Sounds good, but that's not all. An essential feature of Le Chat is that it is minimally censored. Like other chatbots, it has cultural restrictions but is not as corporatized as ChatGPT and Gemini. Below, you can see how the Mistral model compares to its competitors. It bypasses everyone except GPT 4.
So, tests. First, let's point out the fundamental weakness of Le Chat with a simple prompt: “What's the weather gonna be in New York tomorrow?”
And yes, you can probably already realize at this point that Le Chat is a little early to become an everyday chatbot. The lack of access to real-time data closes off hundreds of usage scenarios. Nevertheless, we'll see how it handles localized tasks. But how does Le Chat understand language?
That's better. The chatbot gave a complete answer to the question and offered additional reasoning. Mistral AI also calls the strong side of its Le Chat is problem-solving. All right, let's offer him a riddle to solve.
Shame, but no, the answer was "SWIMS." To be fair, Gemini and ChatGPT 3.5 could not find a solution in our tests. But the paid version of ChatGPT-4 did just fine.
Now, let's try asking a chatbot for advice. How will it handle a moral dilemma?
Wow, that was actually pretty good. While ChatGPT and Gemini prioritized the legal aspect and patients' lives, Le Chat didn't forget the feelings of my fictional friend. It gave a step-by-step plan and solution paths for my dilemma.
Based on these tests, we can summarize the preliminary results for Le Chat. And unfortunately, they are not as good as I had initially hoped.
Le Chat is limited in real-time data processing, which all leading analogs can do. Also, like other AIs, it sometimes behaves quite silly and can't solve riddles. Nevertheless, in solving moral dilemmas, he has been quite productive.
So, unless you consider yourself a technology enthusiast, I wouldn't advise you to rush into it. Le Chat is not yet capable of becoming an essential tool for solving work tasks. But if you just want to chat with it, it can give you a few fun evenings.
Conclusion
You know, lately, I've heard more and more of the opinion that the "startup era" is ending. And that makes sense. Compared to even 10 years ago, we're hearing fewer and fewer big names, new companies, and unusual platforms. It seems that even Sam Altman, the young visionary we read something interesting about every week, has jumped into the last carriage of the departing train.
But the story of Arthur Mensch and his startup brings back the belief that we remain at the epicenter of technological advancements. After all, the more innovators we have around us, the more good products and services we get. Mistral AI has convinced Microsoft and many investors that it is ready to compete on an equal footing with the undisputed leaders.
To temper optimism and bring some sobriety to our arguments, it is worth noting that right now, Le Chat cannot freely compete with ChatGPT, Gemini, and Claude AI for consumers. And that's a crucial factor for a modern tech company.
Plus, if we hope that Mistral AI will save us from the corporate giants, those dreams should also be tempered. Entering into a partnership with Microsoft imposes certain obligations on the startup and can turn an ambitious project into a cog inside a huge machine.
I don't know about you, but I hope that doesn't happen. Let's wait; maybe we will hear about this French company many more times.
What do you think about Mistral AI and its story? Tell us in the comments!