First off, congratulations on reaching a 100th post milestone, Dr. Topol. I appreciate the depth and care with which you’ve reported on the status of AI studies. You demystify, even for lay people like me, AI’s potential, as well as potential pitfalls. I was struck by the ability of AI to reduce radiologist reading time and anticipate there are many time-saving possibilities with low downside--for example, in handling admin paperwork so doctors can focus more on their patients. I saw the ChatGPT story when you tweeted it out, and it gave me renewed appreciation for the potential to use AI to get out of the “silo vision” tendencies in delivering care. I suspect many of us know of instances where “silo” practice of medicine resulted in numerous visits, diagnoses, and extended time periods before an accurate diagnosis was achieved. Well, there is lots more about which to be hopeful here--particularly if pursued responsibly, recognizing limitations, and proceeding carefully, as you model so well in what you write here. Thank you so much for all your good work on this and for all else you do.
WHAT DO YOU THINK OF THIS TECHNOLOGY?https://pme.uchicago.edu/news/inverse-vaccine-shows-potential-treat-multiple-sclerosis-and-other-autoimmune-diseases
Dr. Topol, I have become a religious reader of your newsletter. I am a research oncologist and I learn so much each week. Not sure how you keep up with this much literature, but thank you for putting it together for us.
I may get hammered on this but wouldn't it be prudent to proceed with utmost caution here. On the chance, however slight, that AI alignment fails and AI don't have our best interest in mind- on that flight chance, is it wise to train them on how we are put together from the molecular level up? I feel a sense of unease about the carnival air of enthusiasm gathering steam...