We Should Start Calling It Natural Intelligence
The scientist Florian Jug explains that AI will bring profound changes, even though we cannot yet predict in which areas there will be benefits and in which there will be negative consequences. However, expectations are often overinflated
THE CONTEXT
Interview with Florian Jug on the present and future of artificial intelligence, exploring both fears and possible scenarios
We may discover that artificial intelligence is less biased than humans, who are too “eloquent” and not very “forward-thinking”, but above all that it is not as artificial as we have believed until now. This is what emerges from our conversation with Florian Jug, Head of Image Analysis Facility, Computational Biology, Research Group Leader at Human Technopole. A Swiss scientist, Florian Jug is a young man who is very optimistic about the future of artificial intelligence. He is one of those people who make you think that everything you thought you knew is not entirely true.
For about thirty years we have used the word “revolution” for every change we experience, and each time we refer to a single example: the Industrial Revolution. In your opinion, does AI represent a revolution? In which aspects?
AI is certainly driving profound changes, even though we cannot yet fully predict where these changes will turn out to be revolutionary, where they will simply bring greater efficiency, and where instead they may even be counterproductive. In education, for example, AI can be a useful tool, but it can also undermine established curricula by allowing students to circumvent learning processes designed by educators. At the same time, AI is the first tool capable of introducing real efficiency in fields that previously required years of experience. Software development is a good example: professionals can now create applications or websites in a fraction of the time, without sacrificing quality. Similar gains—and in some cases even qualitative improvements—are likely in the legal sector, in healthcare, and beyond. Where the benefits will outweigh the risks is not yet clear: as a society we will have to approach this path with care.
To be understood, the Internet needed to be compared to magic, and tied to the concept of entertainment. The same is happening today with Artificial Intelligence. However, many have the feeling that every time we experience the “wow,” the surprise, it lasts for less and less. This is true for those who use AI for writing, image, music. What surprises people most in scientific research?
I’m not convinced the “wow” effect fades so quickly. Personally, the capabilities of large language models and current chatbots continue to surprise me. Like any powerful tool, they bring both advantages and disadvantages. Those who learn to use them well can save enormous amounts of time and even find new ideas that might not have emerged otherwise. Not all forms of AI appear equally transformative: video generation is technically impressive, but in my view its everyday utility seems much less obvious compared to processing and manipulating language.
In scientific research, what excites me is the potential to accelerate the discovery process itself. Today researchers produce an astonishing amount of data, containing far more than a human being could directly understand or process. If analyzed and combined appropriately, these streams of data could lead to much faster and more efficient progress. At Human Technopole, for example, we are working to enable this type of AI-mediated discovery, and I believe it will significantly raise the pace of research and our potential for scientific discovery.
It’s probably not necessary to explain how much AI impacts everyday life, yet perhaps few know what happens in laboratories where this technology is used or even designed. What are the benefits that people do not yet realize?
In labs, AI is not much farther ahead than what is already available to the public. In academia, new methods are usually made available openly, while companies—though sometimes more cautious—do often publish and share code.
If anything, the challenge goes in the opposite direction: outside AI labs expectations are often inflated compared to what the technology can realistically get in the near future. In the coming years, I believe we will learn much more about the limitations of current models, and I hope this will represent for all of us a valuable learning process.
Outside the laboratories, expectations are often inflated regarding what technology can realistically achieve in the near future
Every innovation is always accompanied by millenarian fears, popular sci-fi, and fear of being replaced. In your opinion, are those working in AI ethically limited or driven mainly by scientific objectives?
Our community includes people with very different perspectives—both on the future of AI and its ethical implications. If AI can contribute to addressing long-standing challenges in healthcare or climate change, that is certainly positive. But if it endangers the jobs of health professionals or highly qualified staff in legal firms, the picture becomes more complex quickly.
I hope that, as a society, we can proceed with caution: be open enough to seize the opportunities, but careful enough not to be caught off guard by the risks.
What are the frontiers of AI in the biomedical, biological, pharmaceutical domains that will bring real improvement to everyday life? Through which tools will these frontiers be overcome?
AI has the potential to greatly increase the rate of discoveries for every euro invested. In the biomedical sciences, this means building a deeper and more holistic understanding of health and disease, shedding light on the physiological processes that sustain life. In the research work I am doing at Human Technopole, for example, I exploit AI’s potential to analyze more rapidly and effectively the images obtained during experiments and to effectively quantify all the biological data collected.
What does it mean that Artificial Intelligence consumes too much energy? Is this perceived as a limit to research in your view?
Training large AI models—such as those behind current chatbots—requires tremendous amounts of energy. Their use at a global scale by billions of users also requires immense data centers. Academic models are generally much smaller, but even then access to adequate computing resources remains a limitation for many researchers worldwide.
That’s why the European Commission is investing in shared infrastructures like the European Open Science Cloud and large data archives such as the BioImage Archive. These initiatives are essential to keep Europe competitive with the US and China.
What are the AI innovations we will hear most about in 2026?
I believe chatbots and large language models will continue to dominate the stage, perhaps with deeper integration into our mobile devices and software tools. At Human Technopole, and more generally across scientific research worldwide, I believe there will be increasing implementation of these technologies for the analysis and comparison of data.
Predictive algorithms are under accusation because their predictions are based on judgments lacking human factors (for example in determining years in prison, issuing bank loans, or evaluating diseases a person may experience in their lifetime, which may intersect with insurance interests). How can this aspect be addressed? Is it addressed currently? If so, how?
It is important to understand that predictions are only as impartial as the data they are trained on. And, little spoiler, we humans are immensely biased. Today we don’t know how to train entirely unbiased systems; I’m not even sure if that is possible. For now, I believe predictive systems should not be used in areas where their bias could cause social harm.
The European Union is taking a commendable role by introducing specific legislation on AI. It is a difficult and sometimes controversial task, but I am proud that Europe is trying to set standards and lead the way in this complex field.
Technological evolution has always impacted all areas of human life and science, except psychology. It’s not the same with AI. Since organizations developing machine learning offer jobs to those with engineering degrees but also skills in psychology, could you explain what that means?
One clear advantage is that chatbots can lower the threshold for people to open up about mental health issues: talking to a machine can feel safer and more anonymous. However, deep treatment requires the involvement of human professionals, and I doubt serious psychological support is possible without qualified experts.
What are the fields in which AI research in Italy is progressing well?
Italy has a solid academic activity around AI, with universities and public institutes active in sectors like machine learning, computer vision, and natural language processing. The country also benefits from national initiatives such as the PhD Ai program, which involves over 50 universities, from high level infrastructures like the Leonardo supercomputer in Bologna, and from large research centers, like Human Technopole, which are involved. On the application front, Italian projects are advancing in biomedical AI, in language models adapted to Italian, and in sustainable approaches to AI.
At the same time, industrial use—especially in small and medium enterprises—is still lagging behind research, and digital skills vary widely between regions. The national strategy for AI (2024–2026) addresses these aspects by linking research, training, and industrial applications, with particular attention to healthcare, public administration, and sustainable innovation.
When was the first time you understood the potential of AI, and what was the most human thought that followed the excitement of that moment?
When I heard about the legendary chess matches between IBM’s Deep Blue and Garry Kasparov, I was in high school. Unlike others, I immediately felt a deep curiosity: how could a computer defeat the world champion, and what else could such a system do beyond chess? That curiosity probably put me on the path to becoming a scientist.
If from the economic point of view AI will force the system to change the world of work (with the end of traditional professions), in what ways will it bring growth and well being?
I believe very few professions will disappear entirely. More often, tasks within them will change: research or communication tasks will require less time, leaving more room for the uniquely human aspects of work.
Can you list three actions or situations that people will definitively—and joyfully—stop doing in 2026?
No. :)
What do you think of those people who believe dialogue with AI tools is possible? Have you ever experienced dialogue with AI chats? If yes, what have you learned?
Of course dialogue is possible—it is indeed the great strength of chatbots. They allow us to interact with a powerful tool in the same natural way we interact with each other. A few years ago, these conversations would quickly reveal the limitations of the systems. Today those limitations emerge much later and in more subtle forms, which leads some people to trust responses more than they should. My advice: use AI to your advantage, but always remain critical—sometimes these systems are more eloquent than farsighted.
In what way can one of the 20th century’s key scientific concepts—the one of entropy, or information theory, often used colloquially to assert the primacy of human language over that of machines — be dismantled by AI?
AI is, essentially, a powerful way to extract hidden patterns and correlations from large data sets. But if an answer to a question is not contained in the data, AI—like any other method—cannot invent it. And even when the answer is present, clues that are too subtle may still be missed. Despite the enormous progress in recent years, there remains a great deal to discover: and this is precisely what makes the field so thrilling and my work so fulfilling.
We have to use artificial intelligence to our advantage, but always remaining critical. These systems are more eloquent than farsighted
Florian Jug
He has a PhD in Computational Neuroscience at the Institute of Theoretical Computer Science, ETH Zurich. His current research focuses on what AI and machine learning can do to analyze biological data.
Davide Burchiellaro
Managing Editor at Gente.it. He was Deputy of Content at Linkiesta. Expert in digital journalism, he has been Deputy Director of Marieclaire.it and has worked for Panorama and Donna24, the women’s section of Il Sole 24 Ore.