[maeil]

> Not only is our understanding of physical health far less "advanced" than advertised, but the system is far too complex for any one doctor to comprehend. Medicine is far from the only field in which this is the case.

This last part, along with (from a comment elsewhere)

> [...] many doctors being terrible at statistics and failure to properly use statistics means they constantly overlook otherwise obvious causes.

means this area should be the one with dozens of AI startups chomping at the bit to improve the situation. I.e. giving suggestions for possible causes of rare medical issues. Because the existing human performance in these situations is so incredibly low mostly due to the exact things that ML models excel at, it should be very easy to at least to improve on that. If it's not happening because of US regulations against "medical advice", then base it in India or wherever it'll work. I guess that makes venture funding and profitability more difficult but if you're looking to do good, this avenue has unfathomably more potential for massive positive impact than e.g. taking a job at a traditional existing non-profit. I reckon it should also be very possible to get funding from exactly such non-profits, especially those aimed at rare medical conditions. Effective altruism-oriented groups should also be interested.

If the issue is the unobtainability of data due to privacy laws, even a ChatGPT-like model trained purely on Google Scholar, medical literature and other relevant sources has enormous potential and again should readily outperform the baseline.