Everyone says AI needs clean data, but I watched a model learn more from messy notes than polished reports
I help out at a small clinic in Tucson and we tried two AI tools for patient intake summaries. The one trained on our messy handwritten notes beat the one fed our clean typed reports by a lot when it came to catching follow up needs. The clean data had already lost the little details like 'patient seems more tired lately' because nobody typed that part in. Anyone else notice that the rough real world stuff beats the tidy version for training AI?