What are the risks?
Is it really safe to take medical advice from an ai? They often halucinate, and have biases. How can you be sure using this model wont have harmfull effects?
Is it really safe to take medical advice from an ai?
no
They often halucinate, and have biases. How can you be sure using this model wont have harmfull effects?
Treat all ai generated content as if you created it. The onus of care delivery is on a doctor, not an AI model I trained in a weekend with a few thousand dollars worth of GPUs.
Then im curious what the real world application is of a model like this. Or is it purely for research?
The real world application is as a research tool.
As a stretch application, you could used this as a differential diagnosis.
@rombodawg
BTW, I am currently training a falcon version of this LoRA that should have much better output. Currently a checkpoint at 4500 steps is available at medfalconv2-1a-40b-lora.
The full EPOCH should be ready by the weekend, stay tuned for medfalconv2-final.
Lets go