Beyond ChatGPT for Health: Why Biology Needs Its Own World Model

​OpenAI is reporting that 1 in 4 ChatGPT users ask health questions weekly. That's billions of queries hitting a system that hallucinates 59-82% of the time on clinical tasks.

​Yann LeCun calls LLMs "doomed". Sophisticated autocomplete that can't model how the world actually works. Nowhere is this more dangerous than health, where the difference between correlation and causation is life or death.

​This dinner is for people building what comes next.

​We'll discuss:

• ​Why world models, not language models, are the path to useful health AI

• ​What it takes to train on 100M+ real patient records (hint: it's not fine-tuning GPT)

• ​The regulatory and sovereignty constraints that make health AI fundamentally different

• ​How federated learning changes the data moat equation
  
  Event hosted by Nuraxi with the entire founding team: Stefano, Das and Zeljko