An Intro to ChatGPT for Healthcare Professionals

I'm in healthcare...What is ChatGPT??I don't know what it is but am hearing all about it.?What do I need to know???

ChatGPT is a popular new software app that lets you ask questions and make requests as if you're interacting with another person rather than a piece of technology.?It's powered by AI (artificial intelligence), and it has many potential applications both in your personal and work life.

As a healthcare professional, what do you need to know about??In this article I'll use simple terms to explain how ChatGPT works, and then give you some tips for using it in a healthcare setting.?My goal is for you to get comfortable with this new type of technology, plus understand its strengths and weaknesses so you can keep yourself, your patients and clients, and your organization safe.?Let's begin!

How ChatGPT Works

What is ChatGPT??ChatGPT is a software app made by the company OpenAI.?It's available to consumers to use for free - just go to chat.openai.com and sign up.?(GPT stands for Generative Pre-trained Transformer - not important to know for now!)

How does it work??The technology behind ChatGPT is called a "large language model" or LLM for short.?Basically, it's an algorithm, or a set of instructions that are followed to solve a problem.?The types of problems that LLMs are designed to solve involve words and language.?Because of this, we're able to interact with them using normal everyday language, which makes them seem "intelligent" (hence "artificial intelligence").

How do LLMs understand language??LLMs get their functionality by "learning" the relationships between words and phrases.?They do this using complex statistical models.?This is what makes them able to make really good predictions of responses to language-based prompts.?It's important to distinguish this from knowing the "truth."?LLMs don't know the "truth" per se.?The quality (i.e. accuracy) of their responses is tied to the data they are trained on.

What does it mean to train an LLM??Training an LLM means feeding it a massive dataset of text and other information and then letting it learn the patterns and connections between words and phrases.?How much text, exactly??For ChatGPT, think the entire internet plus millions of books (fiction and non-fiction) and articles (filtered to remove harmful or offensive content).?So, yeah, a lot.

What happens when you ask ChatGPT a question??It first tries to understand what you're asking by breaking your question down into its individual words and phrases. Then, it looks for information in its knowledge base that's relevant to your question. Finally, it generates a response based on the information it found.

Wet your feet

Hopefully by now you're curious to see how this actually works.?If you haven't already played around with it, follow these steps:

1. Go to chat.openai.com
2. Create a free account
3. In the search box at the bottom of the page type in your first question.?Here are some good ones:

• "What are the advantages and disadvantages to consider before buying a smart watch?"
• "What are some ways to make instant noodles even more delicious?"
• "What are some team bonding activities for our summer work retreat?"
• "Please write me a welcome email for a new nurse on my team"

Your first instinct might be to type in requests that look like your Google searches, which tend to be short and simple.?But, instead, pretend that on the other side of the search box is a smart friend/librarian/professor/expert and ask it questions like you would to them in a conversation.?I'd love to hear what you try and what you think of the responses!

Keep your patients and clients safe

If your organization hasn't blocked access to ChatGPT at work, it means you can probably access it on the same computers as your EHR (electronic health record system), which opens the door to two risks:

1. HIPAA violations:? Whatever you enter into the search box becomes part of the corpus of knowledge that ChatGPT trains on.?In other words, assume that anything you type into ChatGPT could become public.?Yikes!?That means never enter PHI (protected health information) into a search.?How might this happen??It could be as innocuous as copying a paragraph from a specialist's medical note in your EHR, pasting it into ChatGPT, and asking "Please explain this medical note in plain language."?Oops!
2. Patient safety:? Be aware of the limitations of ChatGPT - it is not a replacement for human judgment.?It doesn't always return correct information.?You can see how these prompts could result in unintended outcomes if incorrectly answered by ChatGPT or used in the wrong context:

• "The doctor has ordered 200 mg of ibuprofen for a 75-pound child. What is the correct dosage?"
• "Can you tell me if there are any drug interactions between metformin and this new medication?"
• "Can you help me develop a nutritional intervention plan for an anemic patient?"

Technology has been a part of healthcare since French physician René Laennec invented the stethoscope in 1861. Then came antibiotics, X-rays, genetics, implants, and many more...and now there is AI.?ChatGPT is a fun and useful way to experience it directly in your personal life and healthcare work.

But like all tools it's important to have a basic understanding of how it works and its limitations so you can keep yourself, your patients and clients, and your organization safe.?I hope this article helped grow your knowledge on this topic and I welcome your comments and feedback.?Thank you!