This is a recap of a webinar that explored the potential benefits and challenges of using large language models, including ChatGPT, in healthcare and how the tools can be leveraged to enhance patient care and improve healthcare outcomes.

The webinar panelists were:

  • Anthony Chang MD, MBA, MPH, MS, Chief Intelligence and Innovation Officer, Children’s Hospital of Orange County, Chair, American Board of AI in Medicine and Chair, AIMed
  • Matthieu Komorowski, MD, PhD, Consultant in Intensive Care, Charing Cross Hospital Clinical Senior Lecturer, Faculty of Medicine·Imperial College London
  • Harvey Castro MD, CEO, Author of ChatGPT and Healthcare
  • Alfonso Limon, PhD, Chief Liaison of R&D, Oneirix Labs

The technology for transformer tools that have some sort of awareness and memory has been in development for quite some time but the capabilities have increased so much that now everyone is interested. 

What is good about LLM’s? 

  • You can ask very complicated questions and very often they can give a clinically accurate response. The ability to communicate in human-like terms is quite outstanding. This is considering they don’t have access to the majority of electronic medical records yet or   images (until recently)
  • The applications in medical education. One example is a simulation of a patient with any condition. We should assume they will be used in education, but a critique of the response can be useful
  • Other applications include writing discharge summaries, brainstorming sessions, assistance with tailoring conversations to patients of different ages, backgrounds, and religions
  • It does have a creative component, which gives hope that one day there may be a cognitive component too. 

What are the limitations?

  • Hallucinations
  • Misuse via plagiarism, fabrication of results
  • Data privacy – assume that everything you type in will be captured and used for retraining. 

Is there a concern about an over-reliance on these tools?

  • It’s a bit early to say what the actual impact is going to be, but there is potential for automation bias. Users need to be careful to remember that the information provided is very refined “super information” but not knowledge. Yet.

How will LLM’s be used in the future?

  • They are still great translation models. You can ask them to find a query on a database without knowing SQL, you can ask them to look up records on a data frame without knowing Python to be useful. LLM’s could be a nice way to engineer solutions that traditionally required an expert in a technical field. LLM’s could be used as entry points to do your own research. 
  • Is there still some scientific rigor about programming that clinicians should have? A lot of programming tends to be template oriented. LLM’s can do this really well. What is not clear is whether ChatGPT4 is ready to assist when you want to do something new that hasn’t been done before. 

This new addition seems really novel but in 5-10 years it could look like just an amazing start of a great era in medicine and healthcare.

To watch the webinar on demand, click here

Anthony Chang, Alfonso Limon and Harvey Castro will be at the annual AIMed Global Summit, scheduled for June 4-7th 2023 in San Diego.  Book your place now! 

We believe in changing healthcare one connection at a time. If you are interested in the opinions in this piece, in connecting with the author, or the opportunity to submit an article, let us know. We love to help bring people together! [email protected]