🧠The Bulletin #67: Mistral AI's Le Chat, OpenAI's In-House Chip Initiative, and Microsoft Quantum Computing Breakthrough
In this week's bulletin - Mistral AI's Le Chat and Europe's AI future,👾 OpenAI and the in-house AI chip initiative, 💾 and Microsoft's Topoconductor breakthrough💡
Hey there! Welcome back to the VWV Bulletin! We hope you had a fun and restful President’s weekend. This week, we’re covering the latest news and moves in VC and tech—let’s get into it! 🔥
💡For this semester, keep up with our content if you’re interested in:
Demystifying & breaking into VC
Finding opportunities in the start-up world
Keeping up with VC investment news at Brown & beyond (pro tip: this is essential to breaking in and finding opportunities)
Enjoy the Bulletin! 😁
🌍 Mistral’s Le Chat and the Implications for European AI
A couple weeks ago, Mistral AI launched their new foundational model, Le Chat, a play on words French speakers will understand.🐱 Founded by three former Meta and Google AI researchers, the Paris-based startup made waves in 2023 when it raised $113M at four-weeks old in what was Europe’s largest ever seed round.🤑 Since then, Mistral has raised a total of $1.19B, with notable backers including a16z, General Catalyst, Index Ventures and major companies such as Microsoft and Databricks, and stands tall at a $6.5B valuation.
Why the Hype? ⚡️
After such extensive fundraising, Le Chat is Mistral’s first consumer AI product, and another foundational model meant to join the fight with ChatGPT. What sets it apart? If you go ahead and try it, you’ll notice immediately – it’s FAST.🏎️ Mistral makes use of its own proprietary models and specialized Cerebras inference AI chips to deliver response speeds of 1’000 words per second. Mistral has also partnered with Agence France-Presse (AFP) to allow Le Chat to deliver real time news.🗞️ Did I add that it's also open-source? While it's harder to definitively compare it to other models on problems like coding and math, its ridiculous speed and affordability (it being free) has propelled it to reach 1M downloads in just 14 days.📈
Not Just About the Model 🇪🇺
Mistral isn’t just significant in terms of its model. Le Chat is the first real challenge Europe has been able to mount against the AI market that the US has virtually dominated since the beginning of the new wave of AI hype with the release of ChatGPT.🗺️ EU tech regulation, such as the General Data Protection Regulation, the Digital Markets Act, and the AI Act, has slowed deployment of new AI solutions in the continent for fear of potential legislative backlash.🏛️ Products such as Apple Intelligence or OpenAI’s Sora are still unavailable for use in the EU. Such regulatory barriers have stifled innovation and no doubt contributed to the AI gap between the US and Europe.⛔️ Furthermore, the US leapfrogs the EU in terms of AI funding and investment. Whether this is due to the regulatory aspects, other concerns with the EU’s innovation landscape, or just a general lack of capital, the EU investing only 4% of what the US spends on AI is sure to heavily limit any type of reasonable AI innovation.
What Next? ➡️
While Mistral was lauded as the European challenger to the US’s AI dominance at the Paris AI Summit last week, the EU needs to take radical measures to ensure it does not permanently fall behind in the technological race that will surely define our century.💨 Europe has a hotbed of technological talent: DeepMind was started in the UK before it was bought by Google. But if governments must modify the regulatory burdens facing entrepreneurs and pave the way for increased investment. If they fail to do so, Europe risks being left in the dust, and given the exponential nature of AI progress, the continent risks staying there permanently. Europe cannot afford to place all its eggs in Mistral’s cat-shaped basket; it needs to create a conducive innovation ecosystem that allows its talent to thrive.🪷
🔧 OpenAI Jumps into the Foray of In-House Chipmaking
In other AI news, OpenAI is in the process of finalizing the design for its first in-house AI chip.🏠 The ChatGPT maker intends to use TSMC’s 3nm process to build its own training focused chips and fuel its innovation efforts. The initiative is being led by Richard Ho, who previously worked on Google’s AI chip initiative and who has doubled the size of the team to 40 people in recent months.
A Strategic Play ♟️
Internally, the move is seen as a strategic tool to help reduce the company’s reliance on Nvidia. Having its own in-house chips would give OpenAI powerful negotiating leverage in discussions with chip manufacturers. More than that, it gives the company room to play with in the design of the chips it uses, opening the space for application-specific innovation which, if successful, could be huge.📝
A Broader Trend in Industry
OpenAI’s initiative is by no means new. Tech giants such as Microsoft, Amazon, Meta and Google have developed their own in-house AI chips. However, success has been mixed. Google’s TPU chips have boasted a spectacular cost-to-performance ratio, but Microsoft and Meta’s efforts have been less successful despite years of developments. This remains a development to watch closely, but it should be remembered that Nvidia’s moat remains considerable.🏰 The semiconductor giant will no doubt respond to such developments, but as to how and with what success only time will tell.
🔬 Microsoft’s Quantum Computing Breakthrough
Reported progress in quantum computing has been erratic, oscillating between breakthroughs claiming a new discovery makes these next-gen devices’s delivery to the public just around the corner, to pessimism preaching patience. Yesterday, Microsoft revealed Majorana 1, its latest quantum computing chip claiming to make useful industrial applications possible within years rather than decades.⚠️
Topological Computing 🌀
In a paper published in Nature, researchers at Microsoft outline how by creating a new state of matter called “topoconductors”, a portmanteau for topological superconductors, its new chip has error-correction capabilities embedded in its physics.🧑🔬 Namely, it showed how this new state of matter allows the chip to distinguish between an odd and even number of electrons, which could function as the qubit’s analog to the 0s and 1s of classical computers. It is important to note however that the findings have not been thoroughly tested and verified by others, but the breakthrough remains significant nonetheless.🎓
Endless Possibilities 🧬
Microsoft foresees the first applications of quantum computing to be in scientific computation. It is difficult for classical computers to accurately solve certain problems in chemistry and biology, but quantum computers can leverage the bizarre effects of the quantum world to deliver the answer in seconds.⏩
Can AI Steal the Show? 🤔
Breakthroughs such as these increasingly raise the prospect that we will actively benefit from quantum computing in our lifetimes or even the near future. However, a number of researchers have argued that many of these scientific problems that traditional algorithms have been unable to efficiently solve can effectively be addressed by AI.✋ Such an approach comes at a fraction of the cost. Whereas before quantum computing faced virtually no potential competition in solving an array of scientific problems, AI solutions come at a fraction of the cost, increasing the pressure on the industry to deliver actionable solutions soon.
💻 We’re Updating Our Website
We love our website, and want to make sure it is accurate and up-to-date. Please let us know through this form if you are a part of the VWV team (Investment Committee, Advisors, Alumni) and have changes you wish to make to your personal profile.
That’s it for this week, feel free to email me taari_chandaria@brown.edu with any thoughts or inquiries! 💌
Follow us:
Instagram: @vanwickleventures
Twitter: @VanWickleV