Paris-based startup Mistral AI made waves this week with multiple major announcements that signal its ambition to compete head-on with AI heavyweights OpenAI and Anthropic. But its strategy already faces skepticism.
The company unveiled a new large language model called Mistral Large, claimed to rival Claude and the newly launched GPT-4. It also released its own ChatGPT competitor named Le Chat. Additionally, a partnership with Microsoft was announced, including a $16 million investment.
These developments mark Mistral AI’s rapid rise from stealthy upstart to an AI player commanding global attention. But there are signs it could become as controversial as its inspiration, OpenAI.
From Open Source to Closed Models
Founded just 10 months ago by alums of Google’s DeepMind and Meta, Mistral AI first won plaudits for its open source approach. Its initial model was released publicly along with weights.
Co-founder Antoine Bruguier declared last year: “We believe AI should benefit everyone. We want to build models out in the open that push the technology forward.”
But the bigger Mistral AI has grown, the more that open ethos has faded. Mistral Large is not open source, instead offered via a paid API like Claude and GPT-3.5.
Pricing is around 20% under GPT-4 Turbo currently. But it hints at Mistral AI’s shift towards a commercial mindset closer to OpenAI’s approach.
Le Chat Launch: Mistral AI’s Chatbot Contender
Mistral AI also made its first foray into consumer AI applications with Le Chat. Designed as an alternative to ChatGPT, the chatbot is in free public beta at chat.mistral.ai.
Users can choose between three underlying models, including Mistral Large. For now it avoids OpenAI’s issues around misinformation by lacking internet access during chats.
Antoine Bruguier introduced Le Chat by saying: “We designed this assistant to be helpful, harmless, and honest.”
But an enterprise version for businesses is also planned, which may require different safety trade-offs to maximize revenue opportunities.
Microsoft Partnership Draws EU Attention
Arguably most significant was Mistral AI unveiling a partnership with Microsoft. This includes integration of its models into Microsoft’s Azure cloud platform.
Azure customers can now access Mistral’s language models alongside OpenAI and Meta’s alternatives. For Mistral AI it promises to supercharge adoption.
But the deal also encompasses a $16 million investment from Microsoft, converting to equity in Mistral AI’s next funding round.
Given Mistral AI’s recent $2 billion valuation, this suggests Microsoft will own around a 0.8% stake. Nonetheless, it was enough to attract regulatory attention.
A European Commission spokesperson told TechCrunch it will investigate the partnership as part of its ongoing probe into Microsoft’s relationship with generative AI companies.
Controversial Funding and Unanswered Questions
Microsoft is not Mistral AI’s only backer that may raise concerns. Despite being hailed as a European AI champion, 80% of its $550 million funding so far has come from U.S. investors.
Key American backers include some of the largest West Coast tech funds like Andreessen Horowitz (a16z), Insight Partners and Tiger Global.
This reliance on the U.S. tech funding apparatus sits awkwardly with Mistral AI’s self-proclaimed European identity and independence from Big Tech.
Meanwhile, many basic questions about the company remain unanswered. Mistral AI’s blog mentions over 100 employees, but it has never named a single one beyond the co-founders.
There is an air of secrecy that contrasts with the transparency pledges of its early days. Its ambitions and capabilities around safety and ethics also remain vague aspirations rather than concrete details.
Rapid Rise Reflects Investor FOMO
What is unambiguous is Mistral AI’s blistering early growth. Securing over half a billion dollars within a year exceeds even OpenAI’s funding trajectory.
Investor fear-of-missing-out (FOMO) over the next “OpenAI of Europe” is steering huge sums towards unproven French founders.
But this breakneck speed also draws comparisons with the meteoric collapse of U.S. AI lab Anthropic last year. Its $11 billion valuation crumbled after safety issues and exaggerated capabilities around its Claude model were exposed.
If Mistral AI has similarly overpromised, while compromising ethics for rapid progress, it may meet a similar downfall.
Managing Hype Around Mistral Large Model
First indications suggest Mistral AI is following OpenAI’s playbook in generating hype around its flagship model. Conveniently benchmarked as superior to Claude but just behind GPT-4, the actual evidence is lacking.
The company admits in its blog post that “Mistral Large still has some limitations” including around reasoning tasks. But balancing realistic shortcomings with investor expectations will be a delicate balancing act.
Founder Bruguier has already echoed Sam Altman’s ambitious predictions, saying artificial general intelligence (AGI) could be developed “potentially in the next decade.” Such grand visions may return to haunt Mistral if progress stalls.
The Rush Towards General Intelligence
Taken together, Mistral AI’s announcements show it racing down the same slippery path as OpenAI towards power and profit through artificial general intelligence.
Its shift from openness to secrecy and prioritizing U.S. backers over European allies and values are all means towards this primary end goal.
But uncontrolled AGI remains widely recognized as deeply dangerous by AI safety experts. And Mistral AI currently looks less restrained than either OpenAI or Anthropic in pursuing technological supremacy.
As with OpenAI this headlong rush towards human-level AI seems destined for future trouble. Whether from internal safety failures or external interventions by regulators and policymakers.
For now Mistral AI is riding a wave of hype, wealth and influence outshining its European AI peers. But under renewed scrutiny from society and regulators, it may struggle to fulfill ambitions closely following the American playbook.
The key question is whether this self-styled European champion can forge its own responsible path on the global AI stage. Or whether FOMO-driven investor backing inevitably corrupts Mistral’s original ideals of transparency and openness.
Copyright©dhaka.ai
tags: Artificial Intelligence, Ai, Dhaka Ai, Ai In Bangladesh, Ai In Dhaka, USA
