The idea of an AI-driven robot taking over our healthcare – one of the most “human” fields that we have – is not a new one. Until recently, however, the idea remained relatively firmly in the realm of science fiction books and movies.
But with the viral success of ChatGPT, the discussion of what this new generation of AI will be doing for us in the future has started to come up in the healthcare field. Now we’re not saying that you’ll be talking to a robot doctor at your next annual checkup any time soon, but we have to ask the question…
Could ChatGPT be your doctor in the future?
How Does ChatGPT Work?
Without getting too deep into AI and Machine Learning, let’s define ChatGPT. The initials “GPT” stand for “Generative Pre-training Transformer,” three fundamental properties of the world’s most famous (currently) Artificial Intelligence that already provide a description of how it works:
- Generative: More specifically than an AI, ChatGPT is a generative language model, a program that has been trained to hold natural conversations with humans.
- Pre-Training: ChatGPT was pre-trained using the largest text dataset ever used for this purpose. Furthermore, the company behind ChatGPT, OpenAI, also trained its product through human feedback.
- Transformer: ChatGPT is built on transformers, which are a type of artificial neural network architecture that allows it to elaborate an input with the information it has been trained with and produce an output.
How is ChatGPT Similar to Doctors?
Simply put, ChatGPT must be trained with datasets in order to process inputs and generate answers that are understandable and accurate to humans. Sounds familiar? This process is similar to the path our society has outlined for anyone seeking to become a professional – in our case, a medical doctor. Doctors are among the professions that require the most years of university education, studying from books and through lessons (datasets) before they can finally talk to patients (receive inputs), provide their diagnosis, and recommend specific treatments (generate outputs).
Only in its infancy, ChatGPT is already a remarkable product, but it is still far from being sufficiently advanced to trust with healthcare decisions or diagnoses. AIs require training on specialized data sets in order to perform specialized functions, and ChatGPT just isn’t trained for healthcare purposes. In fact, when we asked ChatGPT if it could provide medical diagnoses, it responded:
“No, I’m sorry, but as an AI language model, I am not qualified to make medical diagnoses. My programming and training are focused on understanding and processing natural language to provide information, answer questions, and assist with a wide range of tasks. However, I can provide you with general information on medical conditions and their symptoms, as well as advice on seeking professional medical help. If you have a specific medical concern, it’s always best to consult with a licensed medical professional who can properly diagnose and treat your condition.”
For now, doctors all over the world can breathe easily. However, if a new version of ChatGPT was developed – one fed with a dataset of scientific information about healthcare and medicine that is thousands of times larger than any amount of material humans can effectively learn – it could mean a drastic and possibly rapid shift in how we receive our healthcare. If functionality included voice recognition and medical imaging machine learning, such a tool could not only perform initial patient assessments and triage but could potentially become a primary tool - or at least a supplementary tool - for diagnosis in hospitals and clinics.
Human doctors will always have their advantages. Being people, they always have the capacity for empathy and provide the crucial human, emotional and psychological support that patients need to recuperate. Humans are also better at innovative ideas and improvisation, both of which are crucial in ERs, operating theaters, and many other areas of healthcare. But AI models like ChatGPT may soon be able to provide critical support to their work and may even be a potential solution to over-stressed healthcare networks and national services.
The Shift to AI May Not be Seamless
Nevertheless, the implementation of AI in our daily lives may not be as smooth as one might expect. For example, in order to have a trustworthy product, AIs must be trained with reliable medical data.
Furthermore, several paradoxes may emerge from a new collaboration between humans and robots. For example, the AI may use its sophisticated algorithms and data analysis to come up with a diagnosis, whereas the medical doctor may rely on their professional experience and intuition to make a different diagnosis. Which diagnosis should take precedence if the doctor and the AI arrive at opposing conclusions? What happens if a doctor and AI team disagree with the diagnosis of another doctor and AI team? Even without a deep dive into the ethics and implementation, it is obvious that - before any AI integration occurs - experts and governments must discuss a robust framework to regulate scenarios like this one.
For these and other reasons, such a change may appear radical or even dangerous to many, but when one zooms out and considers the incredible advancements in medicine over the last century, as well as the incredibly fast rate at which things continue to change today, we think it’s safe to assume that future generations will have AI robots like ChatGPT covering some diagnoses and taking care of their health on a daily basis.
In less than 100 years, we’ve gone from the discovery of penicillin to custom-printed drugs created through genetics-powered personalized healthcare. In the last few decades alone, we’ve doubled cancer survival rates for almost every type of cancer. Nearly every discipline of our healthcare is seeing benefits from technology, as well as a shift towards data-driven, patient-centric medicine. AI will play a major part in the fourth industrial revolution, and both medical practice and research are already beginning to be more and more driven by these advancements.
Multnomah County Hospital, 1948 (Source) vs. Modern Medical Facility (Source)
Conclusion
History repeats itself: as it happened with the introduction of computers and of the internet, AI and similar other disruptive technologies are inevitably going to cause rapid and profound changes in every aspect of our society.
AI technology is already being used in many healthcare sectors to diagnose and treat illnesses, automate medical processes, to discover and print new drugs, and further advancements are likely to have a major impact on the way medical care is provided.
Data Lake is an EU-funded startup aiming to solve the issues of access to and bias in data in order to ensure these technological developments are equitable and trustworthy, and board chair Dr. Ligia Kornowska is also the leader of the “AI in Healthcare” Coalition. She was instrumental in the creation of the “AI in Clinical Practice” whitepaper, discussing the benefits, concerns, and ethical guidelines required when it comes to the use of AI in healthcare. It will take many similar efforts from governmental institutions and citizen organizations to ensure AI is properly and ethically implemented, with sufficient oversight and a patient-centric focus.
This is an exciting time for healthcare, and as we move towards a digitalized future, it is our duty as a society to prepare for the impact of these technologies and begin a discussion on how they can be useful for humankind in a safe manner.
Could ChatGPT replace your doctor? Well, it’s certainly not going to happen today or any time soon. However, as the introduction of computers and the internet brought profound changes to the way we live, work, and interact, the introduction of AI is certain to have a similarly dramatic impact on medicine and healthcare.
This article was written by the staff of Data Lake, an EU-funded startup working to provide ethical access to medical data for pro-patient research and development purposes.
Also published here.