Turning scientists into pilots with AI. Interview with Dashun Wang

In an essay published this year, Dashun Wang, a professor at the Kellogg School of Management and the McCormick School of Engineering at Northwestern University, compares artificial intelligence agents to ‘airplanes for the mind’: they are machines heavier than air that ought not to fly, ‘but somehow they fly today. They offer great promise and opportunities. They lift us into the sky, show us connections we would otherwise miss, and enable us to achieve things at great speed that we would otherwise not have done. But they also carry additional risks. For example, if they crash, it is a catastrophe for everyone on board, with collateral damage on the ground as well”.

The point of thinking of them as aeroplanes, however, “is that aeroplanes also created pilots. In the same way, these AI agents will probably create a new kind of scientist. “My prediction is that the scientists who adapt most quickly to this new reality could be the ones who pave the way for tomorrow’s discoveries,” Wang tells Il Foglio, a few hours before receiving the 2026 Lagrange Prize from the CRT Foundation in Turin, the international award for the science of complex systems and data. According to the professor, just as aeroplanes gave rise to pilots, and pilots, through a deliberate process, built the institutions surrounding aviation – control towers, safety measures, checklists, airports, black boxes, transport agencies – so too will scientists adapt the institutions of science and build systems around AI that make it reliable and safe.

Dashun Wang received the Lagrange Prize yesterday for his contribution to transforming our understanding of science and innovation by making scientific research itself a subject of quantitative study – that is, the ‘science of science’ – and in particular for his studies on how major scientific discoveries arise in the age of AI, how to manage this transition and what might happen to the future of science, scientists and the relationship between science and society. “We are trying to understand the overall impact and how to make all of this more responsible, safer and more reliable”, just like boarding a flight without fear, as AI takes on more and more scientific tasks that were previously carried out by humans.

Wang explains to Il Foglio that the idea of ‘science: science itself’ is not new; the innovation lies in using large-scale data, AI tools and scientific methods to understand how science works and where major discoveries originate. “By combining data and AI, we have realised that together they tell very illuminating stories,” he says: “How innovative careers develop, how collaborations contribute to discovery, how progress can emerge from multiple interlinked factors. And these insights are not just of interest to the scientific community: they often have implications for science policy and for innovation in business too.” For example, in predicting what is known as a ‘hot streak’ – that golden period, the most creative and successful phase of a career: is it possible to understand when the time is right for a person to produce their best work? “I began analysing the careers of artists, scientists and film directors, and in all of them we found something very universal: the greatest successes of a career tend to occur in rapid succession, usually over a period of four or five years,” explains Wang. The same thing happened to Einstein and Van Gogh; it is a recurring pattern that has nothing magical about it, but can be explained by science (and AI): it is preceded by a phase of exploration and in-depth study, which appear to be a crucial factor in predicting the start of a ‘hot streak’. There is a tendency to explore different themes and styles before focusing on just one: it is this transition that marks the beginning. “There are patterns we can identify in the data that explain why they occur,” and this is possible thanks to the speed of artificial intelligence, which can also mean higher quality, because it approaches things from multiple angles simultaneously.

“We are moving towards a world in which major discoveries will increasingly arise from partnerships between humans and machines,” Wang is certain, “but science is not just about solving problems. It is also about deciding which questions are worth asking, recognising when a new direction becomes possible, and having the intuition and courage to pursue it.”

This is why the question is no longer whether to use AI, but how. According to a study published by Wang and some of his colleagues in the academic journal PNAS, amongst the funding applications submitted in recent years by two American universities to the main US federal research agencies, some showing clearer signs of AI use were more likely to secure funding, and proposals with clearer signs of AI also tended to bear a closer resemblance to projects that had already been funded in the past. “There is a risk of science becoming more conformist,” admits Wang, “machines can help scientists formulate their ideas in ways that are more likely to succeed, but in doing so they push them towards directions that the system already recognises and rewards. Funding and evaluation systems must leave room for ideas that appear unusual or risky: we must create conditions that encourage exploration rather than conformity.” Now, when it comes to AI, we still need to build the control towers so that we can board the aeroplane without fear.