Would you like to:
What is AI?
Artificial Intelligence (AI) is a broad term used to describe a range of technologies such as machine learning, neural networks, and more. Lately, attention has been caught by Large Language Models.
There are many areas of health care where AI is already being used effectively and safely, such as processing large amounts of anonymised patient data to identify patterns or trends in treatment needs, rapid comparison and analysis of diagnostic images, and even generating templates for patient notes and communications.
Safe and responsible use of AI
The GPT in ChatGPT stands for Generative Pre-trained Transformer. This particular type of program is trained on large datasets--especially very large language samples, such as websites, books, and other reference materials. That's why they are called Large Language Models (LLMs). It uses this dataset to predict likely, relevant answers to questions (or prompts) that you pose it.
If you use ChatGPT, it may seem like the program is "thinking", but it is actually doing huge, fast calculations to generate likely answers.
Importantly, likely possible answers are not the same as accurate or truthful answers. ChatGPT almost always sounds convincing, but often generates false or non-existent "information". Recent research has shown that "prompt drift" may be taking place, whereby answers from LLMs shift over time (possibly becoming even less reliable).
Opportunities and challenges of LLMs / ChatGPT for healthcare
Other resources
Critical Appraisal of AI Tools
A good starting place is the T.R.U.E. checklist -- Is It True? Is It Reproducible? Is It Useful? Is It Explainable?
This page collects a range of assessment tools that may be useful when appraising AI.
Interested in The No. 1 Question to Ask When Evaluating AI Tools?
Reddy, S. (2023). Evaluating large language models for use in healthcare: A framework for translational value assessment. Informatics in Medicine Unlocked, 41. https://doi.org/10.1016/j.imu.2023.101304
