Explore advanced techniques for fine-tuning GPT-3 on medical texts to improve performance on multiple cognitive tasks and reduce hallucination in this 54-minute video. Learn how to create medical question-answering bots, predict medication prescriptions, extract information from patient charts using NLP, and fine-tune models for specific tasks. Discover methods for building a Single Source of Truth for Home Care, processing medical transcripts, and generating medical diagnoses. Gain insights into data generation, formatting for training, and the challenges of fine-tuning models like Curie. Ideal for AI enthusiasts and healthcare professionals interested in applying natural language processing to medical applications.
Fine-Tune Multiple Cognitive Tasks with GPT-3 on Medical Texts and Reduce Hallucination
David Shapiro ~ AI via YouTube
Overview
Syllabus
- GPT-3 Hallucinating
- Creating a medical question answering bot
- Using GPT-3 to predict whether or not medications were prescribed
- Using NLP to automatically extract information from patient charts
- Fine-tuning a model to be less creative
- Fine-tuning a model to do two cognitive tasks at once
- Building a Single Source of Truth for Home Care
- Determining whether or not follow-up medical tests were ordered
- Creating a mock-up for data
- GPT3 Completion
- Building a program to process medical transcripts
- The Dmark
- Creating a bot to generate medical diagnoses
- Building a chatbot to answer questions about a job
- Building a medical chatbot
- Python generate data
- Python generate data
- Saving work in progress
- Building a machine that can write
- Formatting the data for training
- Fine-tuning a machine learning model
- The difficulty of diagnosing with Curie
- The ineffectiveness of curie with more training data
Taught by
David Shapiro ~ AI
