Google’s AI co-scientist could enhance research, say Imperial researchers

An unreleased system designed to assist researchers has the potential to “supercharge science”, according to Imperial researchers.

A partnership between Imperial, the Fleming Initiative, and technology giant Google gave scientists access to a powerful new artificial intelligence, designed to make research faster and more efficient.

Google has published the first test results of its AI ‘co-scientist’ system, in which academics from a handful of top-universities asked a question to help them make progress in their field of biomedical research.

Our scientists are among the most talented in the world, with the curiosity and lateral thinking needed to exploit AI technologies for societal good. Starting with new avenues for biomedical research and sowing the seeds for greater scientific efficiency - the prospects could be game-changing Professor Mary Ryan Vice Provost (Research and Enterprise) and Imperial Global USA Academic Theme Lead for Advanced Materials and Cleantech

Imperial’s researchers approached this challenge laterally, after realising the AI’s answers would potentially take years to validate, asking it a question to which they already knew the answer from laboratory experiments.

The findings, which are yet to be peer-reviewed, are published online for experts and members of public to see, and the scientists will continue to work with Google to further develop the system before it is ready to be launched.

Professor Mary Ryan, Vice Provost (Research and Enterprise) and Imperial Global USA Academic Theme Lead for Advanced Materials and Cleantech, at Imperial College London, said: "The world is facing multiple complex challenges – from pandemics to environmental sustainability and food security. To address these urgent needs means accelerating traditional R&D processes and Artificial Intelligence will increasingly support scientific discovery and pioneering developments.

“Our scientists are among the most talented in the world, with the curiosity and lateral thinking needed to exploit AI technologies for societal good. Starting with new avenues for biomedical research and sowing the seeds for greater scientific efficiency - the prospects could be game-changing.

“Last year, Imperial became the first UK university to have a permanent science and tech base in the US, launching a hub in the San Francisco Bay Area. We are proud of our long legacy of transatlantic collaboration, creating and nurturing the kind of partnerships with industry partners that lead to world-changing scientific discoveries.”

Purpose-built for collaboration

Google’s AI co-scientist system does not aim to completely automate the scientific process with AI. Instead, it is purpose-built for collaboration to help experts who can converse with the tool in simple natural language, and provide feedback in a variety of ways, including directly supplying their own hypotheses to be tested experimentally by the scientists.

Dr Tiago Dias da Costa, who co-led the experimental work from Imperial’s Department of Life Sciences and the validation work with the Fleming Initiative - a joint partnership between Imperial College London and Imperial College Healthcare NHS Trust - using Google’s AI co-scientist platform, said: “In scientific research a huge part of the discovery process often means exploring numerous experimental ‘dead ends’. We spend a considerable amount of time identifying the right scientific questions to ask, designing experiments to answer them, and evaluating the results - sometimes only to uncover inconclusive or uninformative findings.”

Do more with less

The AI-co-scientist accessed multiple sources of information, including web search, research papers, graphs and databases about this research area, specialised AI systems for feedback and refinement, and private documents submitted manually.

Imperial’s researchers approached this challenge from a new angle, building on a relationship established through the Fleming Initiative. They designed their question to the AI-co-scientist, asking it to explore a topic they had been pondering for over ten years, and had been the subject of recent unpublished laboratory experiments.

Professor José Penadés, from Imperial’s Department of Infectious Disease and the Fleming Initiative who co-led the experimental work, said: “Laboratory science is resource-intensive, and with global challenges like antimicrobial resistance looming, it’s clear we need to do more with less and speed up new discoveries.

“When the Google research team approached us to test its AI platform, we realised we needed to task it with the same scientific questions that we had already explored ourselves and used as the basis of our experimental work.

“This effectively meant that the algorithm was able to look at the available evidence, analyse the possibilities, ask questions, design experiments, and propose the very same hypothesis that we arrived at through years of painstaking scientific research, but in a fraction of the time.

Enhance discoveries

Google researchers expect that AI will accelerate scientific discovery and increase, rather than decrease scientific collaboration. The tool, for instance, might help improve the time it could take research teams to conduct detailed literature review in many fields with which they might otherwise not be immediately familiar. This may in turn accelerate the time to discovery, and perhaps lead to both more discoveries by experts in their fields and lower barriers to entry for new research scientists who wish to contribute to leading work.

The scientists are working in the area of antimicrobial resistance (AMR), which is a critical global healthcare challenge with increasing rates of infections and deaths worldwide.

The Fleming Initiative, which is a partnership between Imperial College London and Imperial College Healthcare NHS Trust, will launch global programmes of work to address the drivers of AMR, develop international networks of AMR expertise, and outline strategic research themes to rapidly advance solutions to these urgent challenges, including:

• Leading state-of-the-art drug discovery using AI and machine learning and rapid high throughput experimentation for new therapeutics, and;
• Developing diagnostics to improve early detection, prevent transmission and enable specific treatments.

Dr Dias da Costa said: “What our findings show is that AI has the potential to synthesise all the available evidence and direct us to the most important questions and experimental designs. If the system works as well as we hope it could, this could be game-changing; ruling out ‘dead ends’ and effectively enabling us to progress at an extraordinary pace.”

Professor Penadés said: “This type of AI ‘co-scientist’ platform is still at an early stage, but we can already see how it has the potential to supercharge science.”

--

Read more about the AI co-scientist project in Google's blog