GPT agents in healthcare: What you need to know

Updated: May 16, 2023

While ChatGPT is the most popular example of a GPT agent, using agents in a U.S. healthcare setting warrants discussion.

What’s happening: GPT agents are AI systems that use a GPT model to perform tasks like generating text, answering questions, translating languages, or engaging in conversation. 

Why it matters: Imagine having a GPT agent with the knowledge of say, WebMD, connected to your watch or phone, yielding realtime medical advice based on your activities. 

What's GPT again?: GPT stands for "generative pre-trained transformer." It's a type of artificial intelligence language model developed by OpenAI. It has three components:

1. Generative: The model can generate coherent and relevant text, given a specific input or prompt.
2. Pre-trained: The model has been trained on a massive dataset of text from the internet before being fine-tuned for specific tasks or applications.
3. Transformer: This refers to the underlying architecture used in the model, which is based on the Transformer architecture introduced by Vaswani et al. in a 2017 paper titled "Attention is All You Need." The Transformer architecture has become the foundation for many state-of-the-art natural language processing models, including GPT.
   

Caveats: 

• At present, OpenAI does not sign a business associate agreement (BAA).
• Information submitted via ChatGPT is stored by OpenAI (for training purposes).
• If you use the OpenAI API, you can request for information not to be stored (opt-out).
• GPT agents do not have a memory. In other words, they can't recall specific facts or details from past exchanges.
• ChatGPT is often wrong, even GPT-4.

How ChatGPT handles data:

According to OpenAI's data policies as of March 1, 2023:

• OpenAI will not use data submitted by customers via API to train or improve our models, unless explicitly opted in to share data..
• Any data sent through the API will be retained for abuse and misuse monitoring purposes for a maximum of 30 days, after which it will be deleted (unless otherwise required by law).
• Non-API data is used in AI training by default, but users can opt out of having their data used in training by submitting this form.

The bottom line: Opportunity in healthcare for GPT agents exists:

1. Opt-out of having OpenAI store information submitted via its API.
2. Create an agent that can store memory, either via a database or a memory-augmented neural network.
3. For now, the healthcare GPT agent should be actively monitored by humans for accuracy.

See also: Epic, Microsoft bring generative AI to EHRs