AI has been all the rage over the past year. Maybe you have played with ChatGPT or encountered yet another news story about the increasing impact of AI on a given industry. In this article, I’d like to break down a few things:
- What is AI?
- How is AI already being used in health care and other industries?
- How can I use AI in my practice?
What is AI?
AI means artificial intelligence. In using ChatGPT, an AI product of OpenAI, to answer the question “What is artificial intelligence?” it offered this reply: “Artificial intelligence (AI) refers to the simulation of human intelligence in machines that are programmed to think and learn like humans.”
ChatGPT presented examples of AI such as speech recognition, decision-making, and translating languages. In brief, AI replaces humans in needing to convert input, such as words, sounds, numbers, or images, to create outputs, such as decisions and responses to prompts. Subsets of AI such as machine learning, natural language processing, robotics, and computer vision are responsible for the creation of tools such as ChatGPT itself.
GPTs that have dominated the news are generative pre-trained transformers. GPTs use large language models to input language data to answer prompts. Think of a GPT as an intern who is asked a question on rounds and then scrambles through Google, Pubmed, Netters, and picks a few fellow resident brains to find the answer.
While the details can be complex, most AI is created using a standard process. The steps to creating an AI model include identifying a problem, determining the data to use, training a model using the data, testing and refining the model, then iterating the model based on user interactions. Improvements in computing power and the amount of data available to train models have allowed AI to become increasingly capable and popular.
How is AI already being used in medicine and other industries?
AI may seem new to some of us, but it is already widely used. One of the simplest examples is the chatbot you see when you open your bank website or contact customer service online. In health care, there are many current uses of AI, including:
Recommendations. Electronic health record providers like Epic and Cerner offer autofill options for completing notes.
Object recognition. Analysis of radiology imaging, ophthalmology scans, and pathology slides to detect anomalies and offer diagnoses.
Speech recognition. AI scribes detect speech and convert it to notes during patient visits and telehealth.
Now consider a few examples of AI applications in industries outside of health care:
Object recognition. Many security systems now use fingerprint, eye, or facial recognition.
Recommendation. Netflix and Amazon offer recommendations based on prior purchases and browsing history.
Large language models. ChatGPT, an OpenAI product, and other chatbots allow individuals to enter prompts to generate replies but also offer custom services to many companies, including Microsoft, Instacart, and Salesforce.
Navigation. Self-driving cars, for example, Waymo, provide obvious examples of AI in action, however, Google Maps also uses AI to improve navigation for human drivers.
Prediction. Zillow and Priceline offer predictions on future house and flight prices based on many variables that change over time.
In these and other applications, the purpose of AI is to enable people to complete tasks easier, faster, and sometimes safer. However, not all AI applications are efficient or useful. AI can cause harm based on inappropriate training or use. Safety features and monitoring are essential parts of any responsible AI application. Given this power, there is limitless potential for AI improvement and future uses.
How can I use AI in my practice?
Of course, Google and Amazon use AI, but what about me? As a health care professional, how can I use AI?
While the options are limited by imagination alone, here are some examples of AI applications in health care to consider:
Custom GPT/chatbots. Create a chatbot for patients to use to answer frequently asked questions, enable scheduling, or help navigate a website or portal.
Recommendation. Use the contents of notes to recommend CPT codes and automate billing submissions.
Computer vision/object recognition. Identify anatomy during surgical procedures for surgeon training, quality assessment, and safety.
Prediction. Risk assessment for patients based on real-time or long-term monitoring.
If you have a process to automate or a question that may benefit from recognition, prediction, or recommendation, AI may help. Many businesses are springing up to help health care providers implement AI tools. For smaller projects, consider internal IT resources, hiring contractors, or even asking service providers, such as your EHR company, to help develop and use AI tools.
In my own practice as a bariatric surgeon, a small team and I are creating two separate custom GPTs for our patients, those who have had bariatric surgery and those participating in medical weight management. In preliminary use, these custom GPTs appear to offer accessible, validated answers to users and alleviate some administrative burden for our staff. We are separately working on a computer vision project to recognize surgical tools.
Besides providing improved access to diagnosis and treatment for patients, AI tools may enable all of us as health care professionals to offload the mundane or tedious parts of our work. I hope to spark in you a sense of the possibilities that can benefit us all.
Maria Iliakova is a bariatric and general surgeon.