The requirements for an LLM in medical are assistance rather than automation. Can be mixture of models.

The two competencies are...

1. correct visual and information analysis
2. correct and high level intelligence, the medical knowledge base

For instance...

• Zero Shot Diagnosis: operator uploads a photo and the LLM correctly guesses the condition. Operator uploads a video or a stream and the LLM describes the video's biological process, identifying conditions. Watching osteoblasts moving to site, identifying them and describing what they are doing by interpreting the video, not the text book definition.

• Medical Knowledge Base: an expert system, the model knows all aspects of medicine and can deliver encyclopedic correct medical responses, highly intelligent and nuanced. Accurate in recall of medical facts, guidelines, and best practices. Up-to-date knowledge of current medical literature and research. Understands anatomy, physiology, pathology, and pharmacology.

• Diagnostic Reasoning: along with zero-shot diagnoses, LLM can analyse and advise patient data, test results and advise, form different diagnoses, suggest appropriate diagnostic tests and procedures, treatment options or drugs and these options are correct and evidence-based. Nuanced, understands drug interactions and contraindications to tailor treatment plans to individual patient factors.

• Research: Designing clinical trials and assistance in clinical trials. Understanding regulatory bodies and avoiding friction. Assist in research. Integration with medical machinery to perform lab work.

• Image and video to text, summarization, detailed report: uploading a video or image and the LLM returns text of what it is observiing, biotech research lab images and medical images.

• General Mode v. Expert Mode: Be friendly and safe to provide elementary advise for anyone.

• Open Medical LLM Leaderboard
• A Practical Guide for Medical Large Language Models