(It's such a pain these days to disambiguate "AI", as both artificial intelligence and autoimmune diseases) Autoimmune diseases have non-local effects, both i

(It's such a pain these days to disambiguate "AI", as both artificial intelligence and autoimmune diseases) Autoimmune diseases have non-local effects, both in time and space, and thus are very hard to investigate or diagnose. They have population dynamics (e.g., as you mount a response to an infection), which means feedback effects. They're also sensitive to and productive of systemic stress and anxiety, so the feedback runs all the way to neurology, psychiatry, and environment. In a sense we have multiple overlapping immune systems, with triggers and effectors that interplay differently in the gut, brain, lungs, and other organs. Each organ has evolved locally-appropriate systems, and are also subject to the blood/lymph-borne systems, or might be more encapsulated/resistant like brain or distal, poorly-perfused tissues. Then it's complicated by personal history and genetics. E.g., some forms of EBV with some HLA genetics produce B-cells that happen to love a protein on the foot process of myelin; this truly does come down to a specific molecular antibody shape, now prolifically produced... So aside from the few diseases that have single point of effect, autoimmune diseases are even more personalized than cancer (which at least is tissue-local), making research into causes even more difficult. I don't doubt that AI might arm patients with enough possibilities to be a lot more persistent in seeking help, or that sometimes a possibility might be right. I do think that a lower threshold for analytical testing would help, but that would require a higher threshold for acting on results. AI has been shown to be better at image recognition than radiologists, and to have more extensive differentials for primary care doctors, but both are applying implicit likelihood filters based on what can be investigated. I'm fairly confident investing in patient-driven LLM-based diagnosis for autoimmune disease using existing tests and records is a waste of time and money. I'm looking forward to a world where we gather more data more systematically, and use analysis to find connections that can knit together disease dynamics we can validate and block. But the most impact would be a world where anyone can access all their health information as data, and where doctors don't spend their evenings with EPIC.