AI Risks and Opportunities Test Scientific Integrity, Say Top Scientists
A panel of leading laureates warns that rapid adoption of AI demands rigorous verification as new data‑rich tools transform research
Artificial intelligence is accelerating scientific work, but at the same time, the rapid spread of automated tools is weakening the safeguards that keep research credible, according to several top scientists.
In a panel discussion at the 2025 Hong Kong Laureate Forum (HKLF) on November 5 in Hong Kong, two Shaw laureates and a senior academic discussed how AI is transforming scientific methods and reshaping expectations for the next generation.
They said powerful models must be accompanied by rigorous verification, particularly as scientists face growing pressure to interpret increasingly complex datasets.
“Some mathematicians believe that in ten years we will be using AI to prove theorems, not just to state them,” said Nigel Hitchin, a British mathematician and 2016 Shaw Laureate in Mathematical Sciences. “But there are all sorts of dangers associated with this. The mind has to understand what is going on as well as the computer.”
“There’s a great fan base for AI techniques in astronomy, but I also think there’s a great danger that we will use them naively,” said Matthew Bailes, an Australian astrophysicist and 2023 Shaw Laureate in Astronomy. “You need to learn about unit tests and testing that what the AI is feeding you is actually correct.”
Both laureates warned that the risks extend beyond individual errors. They pointed to opaque reasoning, soaring energy demands, and the growing temptation for students to outsource thinking to automated systems, turning AI from a tool into an unearned authority.
Their concerns set the tone for a wider examination of how AI is altering scientific practice. Hitchin said the discipline must prepare for a future in which formalised theorem libraries and machine‑generated proofs become routine, urging young researchers to “dig deeper” and understand the structures behind results rather than rely on automated outputs.
Bailes warned that over‑reliance on AI risks eroding foundational skills. He described students submitting code they did not understand, generated by chatbots, and added that astronomers must “take time to think” rather than respond instantly to automated prompts. He stressed that intellectual discipline—not computational speed—remains the core of scientific integrity.
Tak‑wah Mak, Professor of Pathology at the Li Ka Shing Faculty of Medicine, The University of Hong Kong, said the risk is even more pronounced in life sciences, where many datasets are incomplete or noisy.
He argued that AI cannot compensate for flawed inputs: “In many fields, like neurobiology and immunology, it’s so complicated. I don’t think there’s enough precise and accurate data for a machine to distill and conclude with really correct answers.”
These warnings underlined the panel’s central message: AI is now inseparable from modern research, but it must serve as a disciplined assistant—not a substitute for expertise.
The panel discussion was moderated by Kenneth Young, Emeritus Professor, Department of Physics, The Chinese University of Hong Kong.
The 2025 HKLF took place at the Hong Kong Science Park from November 5 to 8, bringing together more than 200 university students from around the world and 12 Shaw Laureates for cross‑disciplinary dialogue. The event, sponsored by the Lee Shau Kee Foundation, aims to deepen scientific exchange between young researchers and leading global thinkers.
Next‑Generation Instruments
At the panel discussion, speakers also discussed the rise of billion‑dollar scientific instruments that are reshaping global research agendas.
They highlighted how space‑based observatories, large‑scale surveys, and high‑performance computational systems are generating unprecedented volumes of data—so much so that researchers increasingly rely on automated systems for processing and classification.
Bailes noted that facilities such as these are advancing faster than traditional analytical methods can absorb.
“We are kind of drowning in data,” Bailes said, emphasizing that wide‑field surveys will reveal “so many transient phenomena that it’s difficult to even catalog them, let alone follow up on each one.”
He added that understanding the engineering behind each instrument is essential: “If you understand the instrument and its limitations, it enables you to do breakthrough science.”
These tools, he added, expand the frontier of discovery but also demand new forms of collaboration between scientists and engineers. Investment in infrastructure determines what breakthroughs become possible, and younger researchers must be ready to operate in an environment where data volume and instrument complexity far exceed previous generations.
Mentorship and Prepared Minds
Beyond tools and technology, the panel underscored the enduring value of mentorship. Panelists said their careers—and in some cases their most celebrated contributions—began when mentors placed ambitious, unsolved problems in front of them.
Hitchin explained that receiving the right problem at the right moment shaped his entire trajectory.
“I was in the right place, at the right time, with the right background to actually start making progress,” he recalled. He credited his supervisors with cultivating boldness, encouraging deep engagement with new ideas, and prioritizing understanding over exhaustive preliminary reading.
Bailes said his mentors taught him to distinguish between spurious signals and genuine phenomena long before he recognized their value. He stressed that ambition and disciplined training must coexist to prepare researchers for breakthroughs that may arrive unexpectedly.
The Role of Serendipity
While preparedness matters, the panelists agreed that chance often plays a decisive role in scientific advancement. They described how some of the most important discoveries in their fields came from unexpected data, misplaced assumptions or accidental observations.
Bailes recounted the discovery of fast radio bursts—a phenomenon “a trillion times brighter than any radio burst” previously observed. He said the event surprised the entire community, yet the team’s training allowed them to recognize the anomaly instead of dismissing it.
Mak shared a similar story, saying, “Discoveries are really 80–90% luck. You have to be in the right time, the right place. It’s almost like the stars all aligned,” crediting timing and open‑minded discussion, timing and open‑minded discussion for the identification of immune checkpoint proteins.
He described how a casual conversation with a neurologist later inspired years of integrative research. The lesson, he said, is to “listen everywhere,” because insight can come from any discipline.
The panel briefly turned to the intersection of immunology, neurology, and aging, with Mak noting that immune cells influence far more physiological processes than previously understood. He pointed to emerging evidence linking immune activity in the brain to conditions such as Alzheimer’s and said this integrative view may guide future therapies.
Mak added that studies of neurotransmitter‑producing immune cells reveal unexpected coordination between the brain and the immune system, offering what he described as a “new lens” for understanding complex diseases.
Future‑Driven Strategies
Looking ahead, Bailes urged young researchers to stay true to their strengths and develop resilience.
“You need to look in the mirror and understand what you see and not try to be somebody you’re not,” he said.
He encouraged students to choose work that stretches their abilities and offers real scientific value, rather than following trends. “You should be trying to act with integrity and maximize your opportunities with the talents you’ve been given.”
Hitchin reminded young scientists that missteps are part of the process.
“Finding out you’re wrong is actually giving you knowledge,” he said, noting that reflection and patience help ideas mature into solid contributions. He said that understanding why an approach fails can be just as enlightening as proving a theorem.
Mak emphasised the value of purpose and character in building a lasting career.
“You have to have that passion because you’re giving up a lot of other things,” he said. He added that reputation matters in every discipline: “Being respected is more important than being famous.”
Together, the speakers urged early‑career researchers to act with integrity, think independently, and pursue work that genuinely inspires them. They said the most meaningful advances will come from young scientists who pair curiosity with discipline, embrace collaboration, and stay open to insights that emerge when they least expect them.