- You don’t have to be Sam Altman to build a $1 billion AI company.
- Kai-Fu Lee’s new startup and Mistral AI have shown that AI companies can scale by going open source.
- The rise of open source has made it easier to achieve the kind of success OpenAI’s enjoyed.
The opportunity to build a $1 billion AI business has never seemed this straightforward for entrepreneurs aspiring to be the next Sam Altman.
As OpenAI boss, he’s succeeded thanks to billions of dollars of backing from Microsoft, while maintaining an allure around ChatGPT by keeping the stuff that makes it tick a secret.
It’s becoming clear, however, that this closed approach isn’t the only way to build a successful AI business: being open can get you there too.
This month, Taiwanese computer scientist and AI whiz Kai-Fu Lee celebrated as his startup 01.AI achieved a $1 billion valuation less than eight months after launch, following an investment round, Bloomberg reported.
What makes 01.AI different from OpenAI was its conscious decision to go open source.
This means 01.AI’s AI model Yi-34B is more transparent about the data it’s fed, and can be benchmarked more accurately against other AI models on platforms such as Hugging Face to determine how smart it really is.
It’s not just Lee who has achieved this. Mistral AI, which is just six months old, has been in discussions to raise around $400 million in funding at a valuation of at least $2 billion, Insider reported.
The Paris-based startup, which hopes to be Europe’s biggest rival to OpenAI, has also prioritized open source. In September, it made its Mistral 7B model available for free.
The growth of these startups to unicorn status is a sign that entrepreneurs don’t need Big Tech’s billions and a black box to develop a competitive edge. Being transparent about their AI can do just the same.
“Best way to become a unicorn in record times? Train and release AI openly on @huggingface!” Clément Delangue, cofounder and CEO of Hugging Face, wrote on X.
In response, Lee credited Hugging Face as the reason his company was able to take its AI mode public “rapidly, credibly, and impactfully.”
By going open source, 01.AI has been able to show very publicly that its AI can outperform technology from the industry’s big guns. The Hugging Face leaderboard for pretrained models puts Yi-34B ahead of Meta’s open-source model Llama 2 and Abu Dhabi’s Falcon 180B.
Support for open source has become more evident in recent months amid fears that lawmakers may harm developers in the open-source community in their attempt to regulate AI.
Meta’s chief AI scientist, Yann LeCun, has been one of the most vocal proponents of the technology, describing his belief that “the benefits of open-source AI platforms in terms of progress, safety, economic development, and cultural diversity are overwhelming.”
Developers wanting to build AI businesses at scale will want open source to thrive.
Read the full article here