Rubén Armañanzas reviews the present and the challenges of artificial intelligence in medicine at the LabMeCrazy! Science Film Festival

Artificial intelligence is not a recent trend, but rather a discipline almost a century of history that is now beginning to show its full potential in complex fields such as oncology and neurological diseases. This was one of the central ideas of the talk "Artificial intelligence in medicine: Examples and challenges in cancer and neurological diseases," given by Rubén Armañanzas, researcher Institute of data Science data Artificial Intelligence (DATAI), as framework LabMeCrazy! Science Film Festival, an initiative of the Science Museum – University of Navarra Museum.

During his speech at Civivox Iturrama, Armañanzas offered a clear and accessible overview of the origins of artificial intelligence, its main current approaches, and some of its most promising applications in research , highlighting both its advances and its limitations.

From Alan Turing to today's artificial intelligence

The researcher by placing the birth of artificial intelligence in the work of Alan Turing, considered the father of computer science. "In 1936, Turing laid the instructions of computing with his work computable numbers, when computers did not yet exist," he explained. Years later, in 1950, Turing himself would publicly pose a question that would shape the future of the field: Can machines think?

That discussion in 1956, during a summer school at Dartmouth University, where the term artificial intelligence was first coined. As Armañanzas recalled, the initial definition already pointed to one of today's great challenges: "To simulate intelligence in non-biological systems, we must first understand how our own intelligence works, and today we are still a long way from completely deciphering the human brain."

Three major approaches to artificial intelligence

The researcher three major families of techniques used today in an educational manner. He compared the first, based on learning from data , to the game "Who's Who?", where successive decisions are made to identify patterns. "The machine does the same thing we humans do, but at a speed and with a capacity that we cannot achieve," he said.

The second approach reinforcement learning, used in highly controlled environments such as strategy games or certain financial models. "It works especially well when the rules of the system are very well defined; that's why it still has limitations in open contexts such as autonomous driving," he warned.

Finally, he referred to large language models, such as ChatGPT, which he described as "semantic robots." "They are not people, nor do they think like humans; they are systems that operate in the realm of language and transform the information we give them to provide us with answers, summaries, or new texts," he clarified.

Real-world applications in oncology

The final part of the talk focused on applications of artificial intelligence in medicine, especially in oncology, a field in which Armañanzas' team is actively working within the Institute. One of the examples presented was the use of AI to predict which patients will respond best to immunotherapy, a therapeutic strategy that activates the immune system against tumor cells, but which is only effective in approximately 25–30% of cases.

"If we can anticipate and know which patients are most likely to respond to immunotherapy, we help oncologists make better decisions before starting treatment," he explained. In this regard, his group participated in thedevelopment risk indices for lung and bladder cancer that allow the identification of patient profiles with a higher probability of a positive response.

Armañanzas also presented a project on breast cancer, focusing on particularly aggressive tumors, which seeks to integrate multiple types of data , microbiota, clinical information, among others—to detect patterns invisible to the human eye. "The idea is to build a more complete picture of each patient and use artificial intelligence to discover relationships that help us personalize treatments," he said.

A tool great potential, but with limitations

Throughout the talk, he emphasized the importance of maintaining a realistic view of artificial intelligence. "We are not close to replicating human intelligence, but we are at a core topic moment core topic using these tools in a manager useful way, especially in research , research he concluded.

The lecture part of the LabMeCrazy! Science Film Festival, which combines cinema, science, and knowledge dissemination bring some of the great scientific debates of our time to the general public.