In his daily life, Dr. Julio Mayol (Madrid, 62 years old) poses questions and answers on ChatGPT. This surgeon confirms that artificial intelligence is an indisputable ally of all professions, but in medicine it is developing a new way of understanding clinical care, more focused – ultimately – on listening to the patient. With four decades of experience, Mayol is Scientific Director of the San Carlos Health Research Institute and Professor at the Complutense University of Madrid.
As a professional, he emphasizes that AI enhances research by accelerating drug discovery and hypothesis formulation, which are key aspects of discovering new treatments. As a teacher, he warns that medical training – too long in his opinion – must urgently be reorganized in light of the arrival of new tools. His conclusion was that technology was exposing vulnerabilities in a sector that continued to operate with “19th-century processes.”
I ask: What is the real gateway to artificial intelligence in hospitals?
answer: Maybe we’ll look first at the transformation of hospital management systems, because it’s something that’s very little complicated, and it doesn’t have any direct implications for patients’ health. The entire bureaucratic and administrative part could be done by machines and humans would simply have to check every step: calls, shift records, or specifications for the general purchase of certain goods. What we need to do is see where people are feeling overwhelmed and overwhelmed. Doctors and nurses spend 50% of their time in consultation interacting with information systems; That is, they spend more time looking at screens than looking at people.
Q: There are some companies that are already designing and testing AI agents in hospitals.
A: Although the promise of their use in patients is high, agents must meet a minimum of three regulations. Aside from scientific verification, they first have to pass the GDPR (European Union’s General Data Protection Regulation). Second, they must abide by the AI law; Third, because they are biomedical devices, they require compliance with multiple drug requirements (MDR).Regulation of medical devicesfor its abbreviation in English). This makes its development and implementation very complex, taking into account that its operation must also be monitored periodically because it may sometimes be subject to deviations.
Q: So, how far are we from seeing a robot operating without human intervention?
A: We are still a long way off, but thanks to the large amounts of existing data and the development of robotic surgery, the surgeon will be supported or partially replaced by autonomous robots. Not the entire process, but parts of our procedures, in the future approximately five or ten years, will be replaced by physical artificial intelligence that can reproduce and simulate part of our interventions. Artificial intelligence will play a crucial role in work in complex fields such as, for example, neurosurgery, where the brain must be mapped very well, or in orthopedics. However, we must remember that surgery is not just a technical procedure. In fact, this is the least important thing. What is more complex is making decisions: about who should work, when to intervene, and how to do it.
Q: What if he’s wrong?
A: A strict attitude to medical information must be maintained. Generative AI is not perfect, but perhaps, as Geoffrey Hinton says, AI is an average expert who knows more than any human. Good training and organization in education allow us to develop tools with a high degree of reliability. For example, at university we trained AI in oncology that achieved a success probability of very close to 90%. On any topic related to oncology, his performance far exceeds that of almost anyone known. There are usually no people who can correctly answer 90% of questions on such complex areas.
Q: But the companies behind the big language models, like ChatGPT, haven’t solved the hallucination problem and probably never will.
A: There is a very famous phrase from the US Institute of Medicine that states that error is human nature. No human activity is exempt from making mistakes. We make mistakes in all human activities, including health care. What we need to do now is to use artificial intelligence to reduce the current error level. Otherwise, adverse effects occurring in health systems are a cause of significant morbidity, i.e. infections, and mortality.
Q: In other sectors, there is talk of replacing workers. Could something similar happen in this field?
A: We will reach the final stage where humans will only validate decisions made by agents, but not continuously, because we will not be able to keep up with that pace; We will have to decide the areas of risk and classification. This will cause a human to stop making decisions at certain levels of complexity, and a machine will do it. At that time, we humans will simply become service supervisors. This will occur in research, education, clinical practice and management. This is the development of artificial intelligence in this field.
Q: How can you particularly benefit from artificial intelligence?
A: The use of artificial intelligence represents about 90% of my professional activity. With ChatGPT or Gemini, for example, I can review every week what is published in specialized journals and compile the basic ideas I need to know to create new projects. It also allows me to analyze complex situations, so I have increased my productivity and analytical capabilities, because I use AI to optimize my bandwidth limits. What I do is constantly challenge myself to be able to have conversations that improve my hypotheses, questions, and dialogue.
Q: How are professionals trained to work this way?
A: We haven’t trained people to use this new technology, so sometimes we have bottlenecks because we don’t organize the medical visit in the way that AI helps us the most. Therefore, not only should more technology be introduced; The important thing is to change our processes. Many of the things we do now we shouldn’t do. A machine can do it perfectly, but since we have learned how to do it this way, it is very difficult for us to stop doing it.
Q: With instant access to knowledge, can a profession whose training spans nearly a decade still make sense?
A: The current model relies specifically on memorizing large amounts of information, and I think that’s what made the race so long. We really need to rethink and change medical studies. I don’t know if it will change 180 degrees, but it will be an important shift: we have to rethink the whole training.
We continue to use methodologies that go back to the 19th or 20th century, but we have to work in a world where professionals are going out to train with new tools, and are forced to think differently and take on a different role in the next 20 years. Therefore, our challenge is to reformulate the instructions of healthcare professionals, especially in my case, doctors. We cannot continue to train professionals who will work with integrated AI as part of a team with technologies and processes of the past.
Inclinations It is the EL PAÍS project, through which the newspaper aspires to open a permanent dialogue about the great future challenges facing our society. The initiative is sponsored by Abertis, Enagás, EY, Iberdrola, Iberia, Mapfre, Novartis, the Organization of Ibero-American States (OEI), Redeia, Santander, WPP Media and strategic partner Oliver Wyman.
You can subscribe here To receive the weekly EL PAÍS Tendencias newsletter, every Tuesday, from journalist Javier Sampedro.