AI changes how medical informatics is learned and practiced

Prof. Dr. Enkelejda Miho, Team Leader of Digital Life Sciences at the University of Applied Sciences Northwestern Switzerland (FHNW), explains why medical informatics must move beyond the traditional boundaries of medicine and informatics and how education changes when AI knows everything.

AI pushed universities to rethink education. When chatbots deliver all knowledge at one click, what should higher education look like?

Education should guide learning and critical thinking, highlight pitfalls, and anticipate challenges. It has never been a business of printing knowledge, which AI is very good at. To print, summarize, and provide books is part of the education process, not its core. Higher education should provide active participation and the challenges that are necessary to forge learning, and the space where minds come together physically. The inspiration and promotion of the thinking process, creativity not as a combinatorial exercise, and the ways towards discovery to advise science, business and society is the role that higher education should not prompt out.

So what should young people actually learn at university to be fit for the future?

Young – and not-so-young – people should learn how to learn mainly, then use the time to discover the unknown unknowns by understanding the topic of their interest as it is, the state-of-the-art, and imagine if they would like it to change and how.

Students should be ready to implement their ideas, or to pursue their ideas, as AI has significantly shortened the cycles of "what-next” questions. Rather than a discipline, students should focus on the “adisciplinary” aspects of their interests. Adisciplinarity is a crucial way forward in times when disciplines fluidly integrate and mingle with each other.

For example, it took some decades for biology and chemistry to become biochemistry, or biology and informatics to become bioinformatics. However, it takes a few clicks today to create intelligible connections for emerging disciplines. This blurring of the boundaries of thought into concepts rather than disciplinary processes resonates with the preparation of students to be more critical in their thinking and more flexible in their categorization of knowledge.

Do you observe at FHNW that students are increasingly choosing fields such as medical informatics or life sciences because these are safe harbors, future jobs that AI won't take over?

Students choose for various reasons; however, there has been an undeniable recent increase in interest in informatics-related fields such as medical informatics and digital life sciences. These choices also reflect the trends in the job market.

Medical informatics traditionally combines computer science, information technology, and healthcare to manage medical data and improve patient care. How has this field evolved due to the rise of LLMs?

It is still too early to determine how medical informatics has changed with the development of LLMs. Trends show that the engineering part has become more attainable, while the human factors side of products and services has become more complex. It is easier to reach results, but not so easy to understand what is indeed needed. Therefore, medical informatics has started to speak the language of medicine and informatics; however, we need to do more for the translation – that is not always verbatim.

What will students starting medical informatics today be doing in hospitals in 5 years?

From novel proof-of-concept medical software to test business and disease cases, to new algorithms for large-data analysis, to consulting on the strategy of new hospital or hospital-at-home scenarios.

Basel has one of the world’s strongest concentrations of pharma, biotech, and academic research. How is the ecosystem adapting to AI? Do you see the ecosystem catching the AI train?

The AI train is already in pharma and biotech. Academia is still figuring out if there is a train (smiles). Students explore the options available freely, and curiosity wins in using the latest tech, or one garage version of it, because it is just the best; the AI train is within pharma and biotech already.

Academia is still figuring out if there is a train [laugh]. Students explore the options available freely, and curiosity wins in using the latest tech, or one garage version of it, because it is just the best, not the one that an organization decided to buy. Therefore, the whole ecosystem in Basel is embedded in AI. The small location and proximity of the massive number of professionals just next door makes the beauty of informal talks feel like those of a dream scientific and business community.

Companies such as Anthropic are moving into drug discovery. What does this mean for pharmaceutical companies, universities, and the traditional life sciences ecosystem?

Experience and the look of experience are still two very different things, I feel. Pharma has adopted the recent developments with not much fuss, be it the protein design revolution or the fast-evolving job market needs.

It looks like the whole ecosystem has agreed that experts with AI skills, rather than experts without AI skills (even if AI skills means “just” prompting), are the ones who will survive these times. However, in academia I've learned that the most polished, beautifully written reports don't mean students or newly hired educational program designers know what they are writing or proposing. I wonder how this will impact the ecosystem when one has to stand behind the concepts and strategy going forward. We’ll see.

AI, I mean AlphaFold, won the Nobel Prize. Will AI-driven life sciences dominate the curriculum, or will life sciences remain the core?

Almost a decade ago, we already attempted an AI course to rule them all: life sciences. As a Professor of Digital Life Sciences, I established the Master of Science program in Medical Informatics, recognizing the need well before it became apparent more broadly and putting that vision into practice. Imagining a new academic offer was one thing; making it a reality was another: in times of uncertainty or rapid evolution, we see that most academic institutions play it safe.

To keep pace with change in a sensible direction, without risking developing curricula that are a mere commercial offer – integrating AI into every course and making a course out of every AI lecture offered – takes courage and sometimes a bold move. This kind of program needs the rethinking of the academic offer more in depth, rather than at the top, the master’s level. Therefore, the core will need to change, but it seems to be driven by the pace of society, not AI.

When you meet students starting to study medical informatics or bioinformatics today, what drives them most?

The job market is still one of the main motivations. However, I see students joining at all ages because they want to have an impact, learn, and do something.

You lead the Laboratory of Artificial Intelligence for Health aiHealth Lab with a vision to "develop integrated intelligence for diagnostics and therapeutics". What does it mean?

With all these years of presenting AI for drug discovery – we focus on antibodies or nanobodies – and AI for diagnostics as a backend to medical software used in infections, autoimmunity, or cancer, I was never asked.  When I established the AI Health Lab, the director of the school invited me to think about what I would have imagined doing for a long time. I thought of integrated intelligence as a good 20-year project to develop. However, it took others only two or three years to make the first concept accessible to everyone: LLMs.

This fact has inspired me. I still remember when integrated intelligence was estimated to be 100 years away, not 20, as I was proposing. It was only two or three years.  So, we cannot predict the future, but we had better be wilder in our imaginations.