AI could accelerate a decade of scientific breakthroughs into three years, OpenAI finds
Nearly 98% of scientists expect AI to speed up research, while more than one in five believe discoveries that normally take 10 years could emerge within three.
By Amin Masoodi · Updated

Artificial intelligence could fundamentally change the pace of scientific progress, potentially reducing the time needed to achieve major discoveries from a decade to just a few years, according to a new study by OpenAI.
The findings point to growing confidence among researchers that AI can do more than automate routine academic work, with the technology increasingly being viewed as a tool capable of expanding scientific exploration, improving research efficiency and helping identify new avenues of investigation. OpenAI's study, titled Science at the Frontier: How Scientists Use ChatGPT, found that 98% of surveyed scientists expect AI to accelerate discoveries in their fields. More significantly, 22% believe breakthroughs that would ordinarily require 10 years could be achieved within three years with AI assistance.
The research draws on an analysis of 580,186 scientific messages exchanged by 1,819 researchers participating in OpenAI's ChatGPT for Academic Researchers programme, alongside a survey of 251 scientists. Researchers also reported that AI could enable them to complete comparable tasks to the same standard in nearly half the time, suggesting considerable potential to reduce the workload associated with scientific investigation.
AI changes the research process
The growing role of AI is evident in how scientists are incorporating it into their everyday work. Rather than limiting its use to basic information searches, researchers are turning to ChatGPT for literature reviews, writing assistance, evidence analysis, software development and the exploration of research ideas.
Reading and writing accounted for more than half of the scientific interactions examined in the study, making them the most frequent applications of the technology. AI is also finding a place in the early stages of scientific investigation, particularly when researchers are developing or refining hypotheses.
The study found that 36% of idea-development interactions involved scientists presenting their own concepts to ChatGPT for improvement, while 23% involved asking the system to generate potential research questions. This suggests that AI is increasingly being used as a thinking and research aid, helping scientists examine possibilities, refine arguments and explore questions beyond their initial research plans.
Scientists differ on how quickly discoveries could emerge
Although optimism about AI's contribution to science is widespread, expectations about the speed of future breakthroughs vary across disciplines.
When asked to consider how long a scientific discovery requiring 10 years of work without AI might take with the technology, 22% of respondents estimated three years or less. Around 50% expected the process to take between four and seven years, while 19% anticipated a timeline of eight to nine years.
Researchers in mathematics and statistics expressed particularly strong confidence in AI's potential, with 31% believing that decade-long discoveries could be achieved within three years. The corresponding proportion among computer and information sciences researchers was 28.3%.
The findings indicate that the anticipated benefits of AI are not uniform, with the pace of progress likely to depend on the nature of research, the complexity of scientific problems and the availability of tools and resources.
AI cannot replace scientific judgement
Despite the efficiency gains, the study highlights the continuing importance of human oversight in research.
Scientists did not simply accept AI-generated responses at face value. They sought corrections, clarification or verification after approximately 32% of the responses analysed. The findings reinforce the view that AI currently serves primarily as an aid to scientific work, with researchers retaining responsibility for assessing evidence, checking accuracy and determining whether suggested conclusions are scientifically credible.
Asked how AI could improve their research, 52% of scientists identified faster completion of existing tasks as the biggest benefit. Another 23% pointed to its potential to help identify more promising research questions, while 14% said it could allow them to explore a wider range of subjects.
The distinction is important because AI's contribution may extend beyond saving researchers time. Its ability to help generate and refine ideas could also influence which scientific questions are investigated in the first place.
Funding and staffing remain major obstacles
The prospect of faster discoveries, however, does not eliminate the structural challenges facing scientific research.
More than half of the scientists surveyed, or 51.8%, identified insufficient time or staffing as major constraints on their work. A further 43.8% cited inadequate funding, while only 6.4% reported having no significant research bottlenecks. These constraints could determine how effectively researchers translate AI-driven productivity improvements into actual scientific advances. Faster literature reviews, data analysis and hypothesis development may shorten parts of the research process, but experiments, specialised equipment, laboratory capacity and funding remain essential for validating discoveries.
The findings therefore suggest that AI alone cannot guarantee a dramatic acceleration in scientific progress. Research institutions will also need adequate investment, skilled personnel and infrastructure to turn technological capabilities into measurable outcomes.
OpenAI expands research access
In a parallel move to encourage wider adoption of AI in academia, OpenAI announced that its ChatGPT for Academic Researchers programme would expand to provide free access to more than 200,000 researchers worldwide.
The programme is intended to support academics and scientific institutions in incorporating AI into different stages of their work, from reviewing existing research and analysing data to developing hypotheses and communicating findings. Antimicrobial researcher César de la Fuente has also highlighted the potential of AI tools such as ChatGPT to help scientists explore ideas and advance research.
However, the study's projections should be viewed as scientists' expectations rather than proof that AI has already reduced the time required to achieve major breakthroughs. Its findings are based on reported experiences, AI interactions and researchers' assessments of future possibilities.
Even with that distinction, the results point to a potentially significant transformation in scientific research. As AI becomes more deeply integrated into academic work, its influence may be measured not only by how quickly researchers complete existing tasks but also by the new questions they are able to investigate and the discoveries they may eventually achieve.


.jpg&w=1920&q=75)

.jpg&w=1920&q=75)


