Artificial intelligence (AI) could transform scientific discovery and accelerate drug development, but replacing clinical trials with digital simulations remains a distant prospect given the complexity of human biology, Noubar Afeyan, co-founder and chairman of Moderna, said at the Bloomberg New Economy Forum in New Delhi on Tuesday.
Afeyan cautioned against exaggerated claims about AI’s potential to cure diseases such as cancer, arguing that scientific progress must be accompanied by greater humility about what researchers understand about human physiology.
“I think the hardest thing will be to do what is out there these days in terms of speculation, which is simulating clinical trials and avoiding them altogether,” he said.
He said AI could help address bottlenecks in clinical research, including identifying patients most likely to respond to treatments by analysing electronic medical records and molecular profiles. It could also improve cancer screening by reducing false-positive results, which can lead to unnecessary investigations and increase healthcare costs.
Afeyan cautioned against suggestions that cancer could be cured in the near future, noting that treatments that work in mice do not necessarily translate to humans because of differences in physiology.
“We have to be more humble,” he said, adding that reducing the time and complexity involved in clinical trials and regulatory processes could significantly lower the cost of developing new medicines.
AI could expand scientific discovery
Afeyan argued that AI could fundamentally change how scientists understand biology by helping them identify relationships that conventional research might miss.
He said researchers had traditionally focused on understanding nature’s intelligence but were now also trying to understand machine intelligence. AI could help scientists investigate complex biological systems, including the immune system and viruses.
He also challenged the assumption that scientific discoveries must be fully understood before they can be used, pointing out that the mechanisms of action of many existing medicines remain incompletely understood despite evidence of their effectiveness.
Through rigorous testing and controlled experiments, researchers can establish whether a medicine’s benefits outweigh its harms, he said.
Calls for regulation
Afeyan called for closer regulation of autonomous AI agents, warning that systems capable of adapting, acting independently and interacting with one another could create risks beyond those associated with conventional AI tools.
He distinguished between large language models, which provide capabilities, and AI agents, which can use those capabilities to pursue objectives and take actions.
“To me, LLMs are capability. Agents are intentionality,” he said.
Drawing a parallel with pharmaceutical regulation following the thalidomide tragedy, Afeyan said companies commercialising AI agents should demonstrate their safety before deploying them beyond their own organisations. He also flagged risks from AI-enabled synthetic biology and new biomaterials, saying regulation must account for both deliberate misuse and unintended consequences.
Lila Ibrahim, chief AI readiness officer at Google DeepMind, agreed that AI needed effective regulation and said the company had worked with policymakers and researchers on the ethics and risks of advanced AI systems.
Kiran Mazumdar-Shaw, founder and executive chairperson of Biocon, said AI could accelerate drug innovation and improve affordability by shortening the time required to bring medicines from laboratories to patients. She also highlighted its potential to develop new approaches to diseases such as Alzheimer’s and Parkinson’s, where targeting individual points in biological pathways has yielded limited progress.
Ibrahim cited AlphaFold, Google DeepMind’s protein-structure prediction system, as an example of AI’s contribution to scientific research.