AI with minimum hallucinations and an output that can be audited.
How by using Knowledge Graph as source of truth and RAG for answer audit
First practical application medical field, end result an AI that's capable of clinical diagnosis and can assist medical students in their training.
AI that utilizes Knowledge Graph with 5K nodes (medical terms) and 25K relationships. Answers that can be verified via RAG audit of KG.
Potential application to other specialized areas of human knowledge. Model is available for testing at:
https://huggingface.co/spaces/cmtopbas/medical-slm-testing
An answer at HF might take up to a minute, but less than 3 secs on a dedicated GPU
I am looking for medical schools and/or clinics for a free of charge test run.
Also co-founders with a medical background and experience in marketing.
[link] [comments]