Yann LeCun confirms his new ‘world model’ startup, reportedly seeks $5B+ valuation
Yann LeCun's New Startup Aims to Revolutionize AI with World Models
Renowned AI scientist Yann LeCun has confirmed the launch of his new startup, Advanced Machine Intelligence (AMI), which is working on world model AI. LeCun, a professor at New York University and former VP and Chief AI Scientist at Meta, has joined forces with Alex LeBrun, co-founder and CEO of medical transcription AI startup Nabla, to create a new company that aims to tackle the structural hallucination problems of Large Language Models (LLMs).
What are World Models?
World models are an alternative to LLMs, where the AI attempts to understand its environment (aka the world) so it can simulate cause-and-effect and what-if scenarios to predict outcomes. This approach is considered more robust and reliable than LLMs, which can fabricate information due to their non-deterministic nature.
AMI's Ambitious Fundraising Goals
AMI is reportedly seeking to raise €500 million (about $586 million) at a €3 billion valuation (about $3.5 billion) right out of the gate, before even launching. This valuation is not unusual in the current AI startup landscape, where world-recognized AI scientists are attracting significant investments. For instance, former OpenAI CTO Mira Murati's startup, Thinking Machines Lab, was valued at $12 billion for its seed round last year.
LeCun's Expertise and Street Cred
LeCun's involvement in AMI is significant, given his prestigious A.M. Turing Award for his work on reinforcement learning. His expertise and street cred will undoubtedly attract top talent and investors to the company. Additionally, LeCun's partnership with LeBrun, who has a strong background in multimodal AI, will help AMI develop cutting-edge world models.
Nabla's Partnership with AMI
Nabla, the medical transcription AI startup co-founded by LeBrun, has signed a partnership with AMI to use its models as they are developed. This partnership will allow Nabla to leverage AMI's expertise in world models, while also providing AMI with access to Nabla's extensive data and expertise in medical transcription.
LeBrun's Departure from Nabla
LeBrun will be leaving his role as CEO of Nabla to join AMI as CEO. He will remain on at Nabla as chairman and chief AI scientist. LeBrun's departure from Nabla is a significant move, given the company's growth and success. However, it's clear that LeBrun is excited about the opportunity to work with LeCun and develop world models at AMI.
World Models in the Wild
Top labs and startups like Google DeepMind and Fei-Fei Li's startup, World Labs, are also developing world models. By comparison, the fundraising aspirations of AMI might appear a bit more audacious. However, the potential of world models to revolutionize AI and tackle the structural hallucination problems of LLMs makes AMI's ambitious goals seem more justified.
Implications and Forward-Looking Thoughts
The launch of AMI and its focus on world models has significant implications for the AI industry. If successful, world models could become the new standard for AI development, providing more robust and reliable results. Additionally, the partnership between LeCun and LeBrun will undoubtedly attract top talent and investors to the company, making AMI a force to be reckoned with in the AI startup landscape.
However, the success of AMI will depend on its ability to execute on its ambitious goals and develop world models that can compete with the likes of Google DeepMind and World Labs. If AMI can achieve this, it will be a significant milestone in the development of AI and a testament to the power of world models.
Conclusion
The launch of AMI and its focus on world models marks a significant milestone in the development of AI. With the involvement of LeCun and LeBrun, AMI is well-positioned to tackle the structural hallucination problems of LLMs and develop more robust and reliable AI models. However, the success of AMI will depend on its ability to execute on its ambitious goals and develop world models that can compete with the likes of Google DeepMind and World Labs.
