cancerGO

🤖What I learnt meeting with >30 healthtech start-ups this year. A common theme is that ~ half of them pitched interesting ideas that unfortunately would not be a product I'd want to use as a clinician or would fit in meaningfully into a clinical work flow... 😐There was also a recurring theme of overly ambitious #LLM use cases - It was really painful hearing a number of companies (that raised a modest amount £££) tell me how they'd like to revolutionise healthcare by creating a model to diagnose any undifferentiated patient, spanning over all specialties. It quickly became painfully clear the person had no understanding of EHRs, no understanding of the scientific literature, and no medical knowledge. Can we just perfect #LLMs for administrative tasks before jumping into clinical decision making? 👨🏻⚕️It was also interesting to see majority had no clinicians in the founding team, and even when they did, there was a clear gap between clinical and technical expertise so the two sides were not communicating and sharing ideas effectively. It's a message repeated over and over again by y-combinator and other VCs alike - 1. Build a product that users want 2. For healthcare I think tech needs to be developed use case by use case. Instead of jumping on an ambitious undifferentiated model for all patients. (similar to what Dr. Zhong Wei Khor recently alluded to) 3. The core team should consist of a subject matter specialist to direct the development in a meaningful way, but I'd go as far as saying that the clinician needs to have at least an intermediate level of understanding of data systems, AI models and limitations. (this is something both Keith Grimes and I have explored, and will be an incoming episode on the Dev and doc podcast 👀) 4. My personal opinion, but we can't just keep building healthcare products based on a closed source proprietary model, to me this is not future proofing our technology, and puts the power of people's health into the hands of one big for-profit conglomerate. ----------- Hello 👋 we run a Substack and Podcast, Dev and Doc: AI for healthcare. This includes latest news, education, and deep dives in AI for healthcare! 👨🏻⚕️🤖 Substack- https://lnkd.in/e_XV4Pbz YT - https://lnkd.in/ey3mxsuN Spotify - https://lnkd.in/ecFRTN9R