Annual Report 2025 - Report - Page 34
Artificial Intelligence in Chemistry
From Atoms to Algorithms:
Having a Big Impact
How is artificial intelligence changing chemistry and science as a whole? From
the promise of faster drug development to warnings about over-reliance on algorithms,
a #LINO25 Panel Discussion showed just how transformative AI has been for the
chemistry community.
Artificial intelligence is transforming how research
is conducted in chemistry and beyond. AI-based tools
such as AlphaFold have already changed how chemists
approach complex problems. Yet, such models are only
one part of a larger evolution in how data, computation,
and theory interact. The discussion, moderated by Derek
Muller, creator of the science channel Veritasium, revealed both optimism and caution: from the promise of
accelerated drug development to warnings about excessive dependence on algorithms and the need to preserve
critical thinking in the age of automation.
When asked to define artificial intelligence, John M.
Jumper, who was awarded the 2024 Nobel Prize for his
pioneering work on AlphaFold, described AI as “machine
learning – data plus programme coming together to make
a function.” Fellow Laureate Michael Levitt, awarded
in 2013 for his contributions to computational biology,
added with humour that “even a calculator” could qualify as AI, depending on one’s definition. Young Scientist
Animashree Anandkumar, Professor of Computing at
Caltech, took a broader view, emphasising adaptability as
the core feature of intelligence: “Intelligence is the ability to learn and adapt to surroundings,” she explained.
32 | Talking Chemistry That Matters
For scientific applications, she argued, AI must operate
within the “guardrails of physics.” Her research combines
physical laws with historical data to dramatically accelerate predictions, as in weather forecasting, where AI has
reduced computation time by several orders of magnitude while maintaining scientific accuracy.
Jumper agreed that computational power alone does
not explain AI’s success. The real progress, he noted,
comes from improved data quality and better algorithms.
The evolution of AI in science, he suggested, reflects a
deeper shift toward integrating data-driven methods
with fundamental theory. All panelists shared the view
that AI should be regarded as a tool – one that can extend
the reach of science, but only if applied responsibly.
Nobel Laureate Joachim Frank, honoured in 2017 for
his development of cryo-electron microscopy, described
himself as “absolutely enthusiastic” about using AI to
guide and accelerate experimental analysis. AI predictions, he explained, can help align complex structural
data, a task that has traditionally required painstaking
manual work. Still, Frank stressed that human oversight remains indispensable: “Even when AI gives you a
high-confidence prediction, you need to interpret what
