Generative AI and the future of medicine

Ramya Raman

8/09/2023 12:44:26 PM

RACGP WA Chair Dr Ramya Raman reflects on the disruptive potential of large language models in healthcare.

Concept picture of doctor with tablet
The new technology will have an impact on general practice, but it is not yet clear exactly how.

Generative artificial intelligence (AI) is challenging the livelihoods, values and sense of purpose for millions across many industries, yet it promotes an uplift in productivity never experienced before.
As I write this, screenwriters and actors across the American film industry are on strike, in part because generative AI threatens to leverage their previous work and potentially replace their need altogether for future work.
The trivially accessible ChatGPT service is an example of generative AI that showcases the latest capabilities in language-based content generation.
In doing so, we’re experiencing the power of modern large language models that have been trained using vast quantities of data – covering an unimaginable number of subject areas using massive amounts of computing power and energy.
ChatGPT can generate impressive responses in many human languages, using formal or informal writing styles, and can even generate programming code – which, after all, is a type of language.
Generative AI, however, is more than text generation. It refers to AI that can generate content across written human language, realistic voices, video, images and more.
While many organisations are busy blocking ChatGPT and working out how, and if, it can be safely used (e.g. should it write discharge summaries?), the reality is that generative AI is being transparently embedded in everyday solutions in our lives, and we’re not even aware of it.
Examples include Grammarly that is marketed to make you a better writer; Descript that enables me to create a video in your voice with my words; or Siri-style personal assistants and countless other examples. Solutions using generative AI models are being developed that can learn or reference medical literature, exam questions, and dare I say medical notes and records.
There is a school of thought that generative AI will wipe out knowledge workers. This is an interesting observation, because to an extent (but not totally), doctors may be classified as knowledge workers in the application of our skills, with the knowledge to understand and advise our patients on their health outcomes.
Could a ChatGP (I didn’t forget the T) App collect, augment and build patient history, converse in the patient’s preferred language, and provide a holistic consult for common conditions using vast amounts of medical data, while uplifting productivity by orders of magnitude in public health? Could this displace many specialist appointments?
All the necessary technology components are there to make this a reality, and considerable effort will have to be undertaken by our profession, regulators and technology providers to deal with the complex implications that will arise across governance, liability, workforce disruption, cybersecurity, and other areas.
In the more immediate time horizon, I anticipate generative AI will become part of our administration toolkit, built into our medical management suites with capabilities such as, ‘write a referral letter for CRS, summarising the patient’s most recent patient notes and history’ – and in seconds, a letter is ready for your review and dispatch.
If there was ever a technology that would upend our current generation and beyond, generative AI would have to be up there.
This article originally appeared in Medicus, which is produced by the AMA (WA), and is republished with permission. This article is an opinion piece, not an official RACGP position.

ChatGPT generative AI healthcare

newsGP weekly poll What is your chief concern with role substitution?

newsGP weekly poll What is your chief concern with role substitution?



Login to comment

Dr Stephen Longworth   13/09/2023 12:10:02 AM

Thanks Ramya for this thoughtful piece on AI.

A question we have all added to our consultation agenda-search strategy is: 'What have you seen on google?' We've all had patients who have terrified themselves with an online search. The internet has not put us out of business, far from it.

Symptoms with no identified physiological cause are common, often self limiting and may never reach clinical scrutiny (not least because of waiting for appointments). If AI diagnostic algorithms become easily accessible online, symptoms that would have resolved spontaneously will become subject to the same level of analysis as persistent, and probably more important symptoms, with the attendant risks of overdiagnosis, overinvestigation, and overtreatment (1).

AI won't make doctors obsolete if we are smart enough to use it to make us into better practitioners.

1. Hopcroft K. Artificial intelligence may not recognise the nuances of general practice. BMJ 2018;363:k5205