Turkeys don’t vote for Christmas
I predict that AI-enabled technologies, LLMs and robots will destroy the current role of traditionally trained doctors.
Do you know the quote ‘turkeys don’t vote for Christmas’? Why would doctors and healthcare professionals (HCPs) do something against their self-interests? This MS-Selfie Newsletter, the first of 2025, explains why HCPs are reluctant to incorporate new technologies to transform MS practice.
The context
One of my development goals for 2024 was to explore how AI will affect medical practice, particularly neurology. As 2024 is over and after spending many hours reading AI research and playing around with different large language models (LLM), these are some of my initial thoughts on how LLMs will affect the practice of neurology and medicine in the next few years.
The current model of delivering secondary neurological care, at least in the NHS, is still based on a Victorian model of medicine.
The Victorian era saw the professionalisation of medicine with the standardisation of medical education via medical schools, the licensing of doctors and the emergence of medical specialisation. Hospitals emerged as centres of care, and public health initiatives led to broad reforms, including improved sanitation and the development of vaccines. These changes were underpinned by scientific thinking, emphasising observation and experimentation. One example was replacing the Greek theory of bodily humors with germ theory. In Victorian times, access to healthcare was primarily determined by social class. Women faced significant barriers in the medical profession and often received inadequate care. Thankfully, this has changed, but I don’t need to remind you that health inequality remains a pervasive problem and is now linked more to cultural issues, particularly cognitive biases. A current example of this is medical gaslighting, which disproportionately affects women (please see ‘Gaslighting: an institutional problem’, 25-Aug-2022)
However, the Victorian era laid the foundation for modern medicine. While many of their practices seem crude by today's standards, their emphasis on scientific inquiry, public health, and professionalisation paved the way for the remarkable medical progress of the 20th and 21st centuries. However, the Victorian medical model is unsustainable, given the revolution we are experiencing in the rapid expansion of knowledge and medical innovation. It took me 17 years from leaving school to become a neurologist with a specialist interest in multiple sclerosis (MS). My medical training was comprehensive and included a broad and relatively deep base. In the 1980s, the cannon of medical knowledge was reasonably small, so this was manageable. My medical education expected me to be a generalist, i.e., I was expected to be able to deliver a baby, manage complex obstetrical problems, diagnose and manage a patient experiencing a myocardial infarction and treat someone with schizophrenia. Given the expansion in the medical canon's size, is the current medical education model the best way of delivering a responsive healthcare system? The same arguments can be applied to the allied medical professions, i.e. nursing, pharmacists, physical therapists, radiographers, dentists, etc. We need something different. AI will be the game changer.
Technology
During the COVID-19 pandemic, we started a new way of practising medicine with relatively immature technologies. Many MS Centres have now rolled back to one-to-one consultations and are essentially back to face-to-face interactions. Perverse financial incentives have driven these rollbacks; payers want to pay less for remote and asynchronous consultations (email, text messaging, …). Geographically and temporally synchronised consultations are the most inefficient and unrewarding for managing chronic conditions such as MS. I wrote in detail about those in one of my most popular MS-Selfie Newsletters of 2024; please read ‘Self-management and a rose-tinted-odometer’ (4-Jun-2024).
Over the last four years, IT platforms and healthcare technologies have vastly improved, so why have we gone backwards? Why are HCPs and healthcare systems such as the NHS so slow to adopt new technologies into clinical practice? I have written about my vision for the future, which would address these problems (please see: ‘The future of MS care’; 15-Feb-2022). The NHS, in the form of its smartphone app, has the platform to incorporate some of these ideas into clinical practice.
Embedding the next generation of LLMs (large language models) into the NHS app or similar platforms would transform MS care. Many of you have reservations about AI chatbots replacing HCP interactions (please see ‘MS-GPT - your AI-enabled MS guide’ (22-Nov-2023) and the associated commentary). Since the publication of this newsletter, the large language models (LLMs) underpinning these chatbots have advanced rapidly. The one that impresses me most is Google’s NotebookLM, built with Gemini 2.0, a commercially available LLM. NotebookLM allows HCPs or you to create your very own personalised chatbot.
As an example, I created an MS-SelfieLM using NotebookLM. The resources included were (1) MS-Selfie Newsletters and Case Studies, (2) the MS-Selfie microsite, (3) the MS-Selfie self-management guide and (4) the latest 2024 ‘Brain Health: Time Matters’ policy document. Here is a podcast summary that MS-SelfieLM produced, and the following is an example of the output I got from a few simple questions I asked MS-SelfieLM.
The good thing about NotebookLM is that it references all the statements it produces so that you can check the source. This feature allows you to check and ensure the model is not hallucinating. The latter is a concern of many people about using LLMs to help HCPs and people with MS use these tools for making decisions about their care.
Please note that the MS-SeflieLM answers to these questions are not being made using all data available to Gemini 2.0 on the web but by simply analysing the four resources I included in this model. Imagine how powerful MS-SelfieLM would become if I included the text from the latest edition of McAlpine’s textbook on multiple sclerosis and many other trusted resources. In short, you would have a very reliable and trustworthy companion.
Could the NHS incorporate disease-specific but curated LLMs into its NHS App? This would give HCPs and patients using the LLMs more confidence, as the NHS would curate the LLMs. These LLMs will transform healthcare; in fact, they will revolutionise healthcare.
At the same time, these LLMs will transform healthcare on the coalface; they will revolutionise medical education. Why should we spend years teaching medical students fact-based learning when they can outsource fact-finding to LLMs? This would allow HCPs to be trained much more quickly. Maybe the UK Government's decision to introduce physician associates (PAs) in the early 2000s was very forward-thinking. PAs armed with LLMs will become as good as doctors over time. PAs and LLMs will increase capacity and alleviate pressure on doctors. Tasks traditionally performed by doctors will be taken over by LLM-enabled PAs, which will improve patient access to healthcare services. LLM-enabled PAs will bring a unique skill set to the healthcare team, complementing the work of doctors and other professionals, leading to more comprehensive and efficient patient care. PAs will streamline healthcare delivery and reduce costs by taking on less complex tasks, freeing doctors to focus on more specialised care. Despite the pros of PAs, doctors in the UK have started to rebel against PAs in earnest. I suspect this rebellion is partially because doctors are feeling threatened. Doctors and other HCPs should be reminded of the process of creative destruction.
Austrian economist Joseph Schumpeter coined the term "creative destruction" in 1942. He described it as "the process of industrial mutation that incessantly revolutionises the economic structure from within, incessantly destroying the old one, incessantly creating a new one." I predict that AI-enabled technologies, LLMs and robots will destroy the current role of traditionally trained doctors. We must adapt or die.
How many of you use LLMs to explore the MS information landscape? Is this information reliable enough to help you self-manage your MS? What has your HCP’s response been to you using LLMs for self-management? I have created a short survey that will take you less than 2 minutes to complete to get a handle on this adoption of LLMs into clinical practice. Your help in completing this survey will be greatly appreciated.
Thank you, and happy new year.
Subscriptions and donations
MS-Selfie newsletters and access to the MS-Selfie microsite are free. In comparison, weekly off-topic Q&A sessions are restricted to paying subscribers. Subscriptions are being used to run and maintain the MS Selfie microsite, as I don’t have time to do it myself. You must be a paying subscriber if people want to ask questions unrelated to the Newsletters or Podcasts. If you can’t afford to become a paying subscriber, please email a request for a complimentary subscription (ms-selfie@giovannoni.net).
Important Links
🖋 Medium
General Disclaimer
Please note that the opinions expressed here are those of Professor Giovannoni and do not necessarily reflect the positions of Queen Mary University of London or Barts Health NHS Trust. The advice is intended as general and should not be interpreted as personal clinical advice. If you have problems, please tell your healthcare professional, who will be able to help you.
I can't help but feel that the turkeys here are the patients rather than the doctors. Why are pwMS sceptical about AI? Allow me to speculate:
1. After more than a decade of cufs and underfunding, pwMS sense that AI appointments are being used to replace actual human contact and appointments. You appear to be suggesting this.
2. Who is introducing AI and why? We only need to glance at recent events in the US for an example of the malicious political goals of tech oligarchs, who are using small change (by their standards) to buy elections with the goal of taking an axe to the state.
3. As social scientist Michel Foucault convincingly explains, science is not objective and reflects the prejudices and priorities of the people who created it. You also refer to this in your piece when you show how medicine was originally - and still is, to a certain extent - misogynist. Various studies have been conducted on how LLMs mimic and imitate the racism and sexism of their host societies.
4. All of this is not to say I'm techphobic and that i don't see a role for AI in medicine. I also broadly agree with your diagnosis of contemporary medicine and it matches my experience as a pwMS. However, the point of AI should be to improve our lives, not to save time. I can see how AI could detect lesions on an MRI scan that might be missed by the human eye, but there it is acting as an aid to the human doctor not as their replacement. My ideas on this are influenced by thinkers such as Evgeny Morosov.
I don't use AI at the moment because I have access to all the top journals: Nature, Science, Cell, Lancet, Brain, Annals of Neurology etc. BUT I just made a recent accidental discovery. There is this new AI platform GregoryAI which someone created to bring all the best MS research from all these top journals together where you can see the latest research and search for anything specific you want. I tried using chat-gpt a few times but it gave me some silly answers so I stopped using it. But I agree with you, these AI technologies are already getting so good that one can see getting information from AI is going to give more depth, breadth, and detailed, valuable information to the patient than a single human HCP could manage. And having this AI in your pocket and accessible anytime you want at no cost is going to be revolutionary. I think all this is likely to assist and enhance medical care and as you say free up space for more detailed research by HCP's. I don't think its going to replace human doctors, but the whole system of education and practice as you say I think will change. I mean this is really going to happen in so many industries that in 10-20 years so many professions will be learning and practicing in new and different ways.