Pioneering the Next Frontier of AI: Healthcare Transformation through World Models
In a significant move that signals a potential paradigm shift in artificial intelligence applications, Yann LeCun, one of the world’s leading AI experts, has stepped away from his position as Meta’s chief AI scientist to launch Advanced Machine Intelligence Labs (AMI). The startup, currently seeking funding at a $3.5 billion valuation, aims to develop a revolutionary type of AI called “world models” – systems that learn from videos and spatial data alongside text to better understand and interpret the complexities of the real world. LeCun has made a noteworthy choice for CEO: Alex LeBrun, cofounder of Nabla, a Paris-based health tech company specializing in AI-powered transcription of doctor visits. This appointment underscores AMI’s strategic focus on healthcare as a primary application domain for their cutting-edge technology, with the company’s first partnership being with Nabla itself.
The intersection of advanced AI and healthcare represents a compelling frontier for innovation, driven by the field’s inherent complexity and high-stakes nature. “Healthcare is my baby, and we know what problems we cannot solve today,” LeBrun explained during an interview at the J.P. Morgan Healthcare conference, highlighting his passion for addressing critical gaps in current medical technology. While stepping down as Nabla’s CEO, LeBrun will maintain connections with the company as chairman and chief AI scientist, bringing valuable continuity to the partnership. His previous experience working directly under LeCun at Meta, following the acquisition of his earlier startup, established a foundation of trust and shared vision that now fuels their collaboration at AMI. Beyond healthcare, the company is also targeting applications in manufacturing and robotics, though medical applications remain central to their mission due to the field’s unique challenges and potential for meaningful impact.
The limitations of current large language models (LLMs) in healthcare have become increasingly apparent, particularly their tendency to hallucinate or generate errors – issues that become potentially dangerous in clinical settings where lives hang in the balance. World models offer a promising alternative approach to address these shortcomings. As LeCun explained, health data is “continuous, high-dimensional and noisy” whether from medical tests or sensor readings. While generative AI has proven valuable for clinical documentation, “generative approaches do not work well for more impactful applications.” The world models being developed at AMI are designed to enable “agentic systems to predict the consequences of their actions, and to plan action sequences to accomplish a task, subject to safety guardrails,” potentially unlocking “a new category of AI applications in healthcare, where reliability, controllability and safety really matter.”
This pivot toward world models represents a reaction against what LeBrun characterizes as the industry’s fixation on language models: “The world in the past five years has become fully obsessed with LLMs. Now we have a hammer, and we see everything as a nail.” He acknowledges that for certain problems like information retrieval, LLMs excel, and specific healthcare applications like AI scribing (including Nabla’s own offerings), clinician support tools, and pattern recognition in medical imaging have benefited from existing language model technology. However, these applications represent just “1% or 5% of the problems” in healthcare, according to LeBrun. For more critical clinical applications, current AI systems lack the necessary accuracy to be safely deployed, especially when considering autonomous AI agents operating without human supervision. “Nobody would claim healthcare is working well today,” LeBrun noted, suggesting vast opportunities for improvement through advanced AI.
AMI’s business model centers on partnerships with companies to apply world models to real-world challenges, with Nabla serving as its inaugural collaborator. The French startup has demonstrated significant momentum, having raised $120 million and projecting annualized subscription revenue of $100 million within two years. Nabla’s leadership transition is already underway, with co-founder and COO Delphine Groll stepping in to lead the company until a permanent replacement for LeBrun is named. “Alex has always been the AI visionary of the company,” Groll acknowledged, highlighting the natural progression of his move to AMI while maintaining ties to Nabla. The broader world models space is attracting significant investment, with Stanford professor Fei-Fei Li’s World Labs recently emerging from stealth mode backed by $230 million in funding at a valuation exceeding $1 billion.
Looking ahead, LeBrun anticipates a relatively swift timeline for bringing AMI’s innovations to market, suggesting that “about one year to get the first things we can use in the product” is realistic. This optimism stems from the robust foundation of existing research on world models, with hundreds of academic papers published in recent years establishing the theoretical underpinnings for practical applications. “I felt that the time was right given the recent progress of research in world models to apply this to healthcare, so that is what we are about to do with Nabla,” LeBrun explained. With substantial funding reportedly being raised – the Financial Times indicated AMI is seeking nearly $600 million – the company appears well-positioned to accelerate the development and deployment of world models in healthcare and beyond, potentially transforming how AI addresses some of the most challenging problems in medicine and other complex domains that have remained resistant to current AI approaches.
