Deciding which research projects or technologies to invest in requires a high degree of expertise. One must understand the research or technology while also being well-acquainted with the market. The day is not far off when artificial intelligence (AI) will take over the review of research funding or investment in tech startups. Universities, public funding agencies, and venture capital are increasingly utilizing AI to discover groundbreaking research.
On the 29th (local time), the international journal "Science" introduced examples where AI is used to select research with high commercialization potential. A representative example is the "Climate Solutions Catalyst (CSC)" at Imperial College London. Launched last year with a donation from a charity foundation, CSC is a program that selects climate-related research close to commercialization and provides small research funding.
The CSC is operated by an AI developed by researcher César Quilodrán Casas. The AI has been trained on various research and industry application cases similar to conversational AI, ChatGPT. Through this process, the AI identified which research in the field of green chemistry has high industrial applicability.
The research team input over 10,000 abstracts of papers by UK researchers since 2010 into the trained AI. The AI initially filtered 160 papers with high commercialization potential. Based on this, 50 papers were selected after joint reviews by experts and non-experts, and simple proposals were received from the authors of these papers. Ultimately, three projects, including research by Professor Joanna Sadler at the University of Edinburgh, received seed funding without charge or contract.
Professor Sadler is researching a technology that decomposes disposable plastic tableware into acetone using microorganisms. Acetone is primarily derived from fossil fuels. She noted, "In March, I received an email announcing 'unconditional funding of £35,000 (about 6.6 million won)' and confirmed its authenticity before accepting the proposal."
Christopher Waite, chief of scientific innovation at CSC, stated, "There are no obligations associated with the funding, and Imperial College London will not acquire equity or patent rights. This program aims to uncover groundbreaking discoveries that generally go unnoticed and provide researchers with the tools to bring their findings to market."
However, AI does not make all decisions. George Richardson, head of data science at the UK innovation foundation Nesta, said, "AI is merely a 'large filter' that narrows down possible research among thousands of papers, while the final judgment is made by humans. This method is effective in actively finding suitable candidates to address specific challenges."
Researchers also see advantages in this approach. Dashun Wang, a professor at Northwestern University in the U.S., stated, "AI-based pre-analysis could provide a fairer approach for researchers with less commercialization experience or weaker networks. While it's observed that male professors tend to have an advantage over their female colleagues in patent reviews, AI could reveal previously hidden potential and help bridge that gap."
Recently, the U.S. Federation of American Scientists (FAS) proposed to the White House Office of Science and Technology Policy to implement AI in the review of subsidies from multiple institutions. The goal is for AI to summarize complex research to aid the reviewers' fair understanding.
The fairness and efficiency of AI reviews have already been demonstrated in other fields. A study released last August indicated that AI interviewers performed better than humans in hiring. Researchers from the University of Chicago's Booth School of Business and Erasmus University in the Netherlands found that candidates selected by AI interviewers had higher acceptance and retention rates than those chosen by human interviewers. This suggests that AI was better at selecting employees than humans.
However, there are also concerns about AI reviews. Ramana Nanda, a professor at Imperial College London, remarked, "AI-based decision-making in venture capital could amplify the bias toward startups that resemble companies that have previously succeeded," adding that innovation sometimes requires 'difference.'
Due to these concerns, the National Institutes of Health (NIH) in the U.S. prohibited the use of AI in the proposal review process in 2023, and the UK Research and Innovation (UKRI) created guidelines to prevent generative AI from being used in reviews. This is because there are concerns that submitted research confidentiality could leak into AI models used for training. Richardson from the UK Innovation Foundation also stated, "More testing is needed to understand the actual impact of AI tools on outcomes."
References
Science (2025), DOI: https://doi.org/10.1126/science.zf3ks8o