The race beyond discovery: creating impact in science and from science
Professor Tan Chorh Chuan Chairman, Agency for Science, Technology and Research (A*STAR)
Artificial intelligence, automation and convergence science are transforming research and its application. How can we use them well to drive greater impact in—and from—science?
The world is spending more on research than ever. According to the World Intellectual Property Office, the UN agency for innovation and creativity, global R&D expenditure reached an estimated US$2.87trn in 2024, nearly three times higher than in 2000.
At the same time, the tools researchers use are also changing. Artificial intelligence (AI), for example, is increasingly adopted for tasks ranging across literature reviews, hypothesis generation, design of experiments, data collection and analysis, and scientific inference. A 2025 survey published by Wiley found that the overall usage of AI tools had risen from 57% in 2024 to 84% in 2025, with nearly three quarters saying it has enhanced the quantity and quality of their work.
Augmenting discovery
One key advantage that AI offers is that it can process vast amounts of data and draw connections across multiple scientific fields, something research teams find challenging to do as they typically work within narrow clusters of disciplines. This ability makes AI a powerful tool in advancing convergence science, which brings together concepts and methods from different research areas to address complex problems. Google DeepMind’s AlphaFold is a prime example: developed by combining AI and big data with structural biology and physics, the AI model can predict the three-dimensional shape of a protein from its amino acid sequence with high accuracy.
In fields such as material science, biology and chemistry, AI combined with automation is being integrated at different stages of the research process, including tasks that were previously carried out exclusively by trained researchers. For example, AI models are being increasingly used to generate hypotheses, which are then tested through self-driving laboratories. While still nascent, these labs are becoming more widely adopted to automate increasingly complex research tasks.
The real test, though, is whether AI and automation will significantly accelerate scientific breakthroughs and real-world applications. In a paper published in Nature this year led by Qianyue Hao, an analysis of more than 41 million publications in six natural science fields found that scientists who used AI published three times as many papers and received nearly five times as many citations than those who did not. However, Mr Hao’s team reported that adoption of AI also reduced the range of scientific topics studied by 4.6%. One possible explanation is that AI-augmented research gravitated towards data-rich subjects while automation focused on established fields rather than exploring new ones.
In a paper published by Mr Musslick and his colleagues in PNAS, the automation of research tasks relies on four main factors: the availability and quality of data inputs, the complexity of the computation, the complexity of the hardware, and whether the scientific goals can be clearly defined for machines. There are efforts underway to apply AI and automation to research areas where these four factors are most readily addressed, but substantial work will still be needed to enable automation beyond them.
Pathway to innovation and progress
The application of AI and automation to research is still at a relatively early phase, but it has immense potential to accelerate scientific discovery and enable translation that would otherwise not be possible. Recognising this, Singapore’s Research, Innovation and Enterprise 2030 (RIE2030) plan—which sets out related goals, priorities and initiatives from 2026 to 2030—will invest significantly in data, AI and advanced computing to strengthen the country’s research capabilities, and to advance national economic and strategic priorities.
Presently, Singapore has a system that provides researchers access to anonymised health, and related education, environment and socioeconomic data. Under RIE2030, the platform will expand beyond health to other research areas.
As part of the plan, Singapore will also continue to invest significantly in AI talent and basic research, alongside adopting AI to uplift and transform key economic and strategic sectors. One example is the establishment of the Institute of Advanced Intelligence and Computing (IAIC) by the Agency for Science, Technology and Research (A*STAR). IAIC will build synergistic capabilities in AI, data, and advanced computing, and apply these to priority sectors such as advanced manufacturing, semiconductors, healthcare, and sustainability.
Globally, the continued development of self-driving automated laboratories will depend on overcoming bottlenecks. Validation, for example, is a key challenge. As AI models become capable of predicting more materials or molecules with distinctive properties, there needs to be an efficient way to assess and improve their accuracy. A*STAR is building platforms to close this gap, including MARVEL-AI that pairs computational design with rapid, high-throughput physical validation for materials discovery, and ATOMIC that uses automated single-cell analyses to test AI-generated hypotheses about disease mechanisms and how cells respond to genetic or drug interventions.
Placing humans at the centre
With AI and automation becoming more pervasive, a critical question remains: how do we keep human talent at the heart of the research and innovation system?
The growing role of AI in the cognitive aspects of research makes it crucial to ensure that human researchers remain actively at the core of the process—from determining agendas and the key questions to overseeing outcomes and intended impact. Basic research should enable human understanding of what is complex, and not just increase predictions through automation.
The greatest gains will come from combining the strengths of both: when AI and automation expand what science can solve and deepen our understanding, while human researchers continue to decide which questions matter, and what answers and solutions can create real-world impact.
