Scientists are increasingly using AI tools as part of their jobs, but it seems that some of the potential productivity gains at one part of the research workflow may, right now, be offset by new bottlenecks created elsewhere.
A study lead by researchers at Google has looked at how scientists are using AI and its impact on productivity. It was based on 360,000 interactions by scientists with the company’s Gemini large language model (LLM), plus data on 2,600 specialised AI models focused on a range of areas such as protein structure prediction, materials discovery and weather forecasting - plus a survey of 600 scientists.
The research found that scientists are ahead of other occupations in their use of AI. Nearly half of surveyed scientists said they used AI, either LLMs or more specialised models, as part of their everyday workflow. The scientists are using LLMs to help with tasks such as writing and troubleshooting research code, statistical analysis and literature review. The specialised models, which are relatively more common in health and life sciences are used, for example, to predict disease outcomes, engineer molecular constructs, and run complex simulations.
Around three quarters of researchers report time savings, with on average almost seven hours saved per week. Most of that time goes back into research, the report said, with about eight in 10 scientists reporting higher lab output over the past three years.
But bottlenecks remain, which the report said can explain why some apparent AI productivity gains have not translated into new discoveries.
More than four out of 10 of surveyed scientists reported that bottlenecks have moved downstream over the past two years, into lab execution, clinical validation, and field data collection.
“AI tools may have increased the number of viable theories and hypotheses but not necessarily the means to test them,” the report said. That has led 41% to report a growing backlog of untested hypotheses.
Checking the work of AI is also taking time; nine out of ten of the scientists who save time spend more than a tenth of it checking AI outputs - and nearly half said 46% spend more than a quarter.
Using AI tools also seems to have some impact on the type of research done. Half of the scientists said AI pushes them toward “safer, more incremental projects” where benchmarks are established and results are reliable - despite evidence that most important scientific progress relies on high-risk exploration and novel hypotheses. In contrast just over a quarter of scientists said AI lets them take on riskier questions.
“If AI mainly lowers the cost of incremental work, it could raise the volume of papers while not meaningfully advancing the scientific frontier,” the report said. The paper said over time that AI tools might change the way science discovery is organised, and who does it. While today a researcher moves between AI tools, the next step might be an LLM that orchestrates the specialised models directly, planning experiments and returning answers for the scientist to judge.
“The human’s job shifts toward choosing the question, setting the standard of evidence, deciding what is worth the cost of a physical test,” the paper said, while acknowledging: “Our data say we are not there yet.”
Further reading
- AI, design, energy, robotics - and humanity
- Self-driving labs are coming: What does that mean for chemistry?
- AI jobs and the future of chemistry
Chemistry & Industry (C&I) magazine reports on the people, the scientific advances and the industrial innovations being harnessed to tackle society's biggest challenges. C&I covers advances in agrifood, energy, health and wellbeing, materials, sustainability and environment, as well as science careers, policy and broader innovation issues. C&I’s readers are scientific researchers, business leaders, policy makers and entrepreneurs who harness science to spark innovation.
Get the latest science and innovation news every month with a subscription to Chemistry & Industry magazine. You can subscribe to C&I here.
Visit Blog