The world is being quietly rearranged by people who write very long documents.


The title they went with Automating Clinical Information Retrieval from Finnish Electronic Health Records Using Large Language Models Noisy translates that to

Finnish hospitals can use local AI to read patient records, but it makes dangerous mistakes


Open-source AI models can now pull specific patient information from electronic health records without sending data outside the hospital. This makes it faster for doctors to find what they need, but the AI still makes a few serious errors.
Doctors spend a lot of time digging through patient records. This paper shows that AI can automate that, making it much faster to get specific answers. The key is that hospitals can use these tools without sending sensitive patient data to external cloud services, which could speed up diagnosis and treatment. However, the measured error rate means human doctors still need to double-check everything the AI says.
Watch for hospitals in Finland or other countries to start piloting these local AI systems, and whether their real-world error rates match the study's findings.

If you insist
Read the original →