98% of Scientists Expect AI to Speed Up Scientific Discoveries: OpenAI Study Finds
More than one in five respondents believe AI could help achieve a scientific discovery within three years that would otherwise take 10 years.
Artificial intelligence (AI) has emerged as a potential accelerator of scientific discoveries, with 98% of surveyed scientists expecting it to advance research in their fields, according to a new OpenAI study.
The findings from the report, Science at the Frontier: How Scientists Use ChatGPT, examining how researchers are using AI in scientific work, highlighted how tools such as ChatGPT are helping researchers complete routine tasks faster and explore scientific questions, although practical constraints remain.
The report draws on an analysis of 580,186 scientific messages from 1,819 researchers and a survey of 251 scientists participating in OpenAI’s ChatGPT for Academic Researchers programme.
More than one in five respondents believe AI could help achieve a scientific discovery within three years that would otherwise take 10 years. Researchers also estimated that AI enabled them to complete tasks at comparable levels of quality in nearly half the time.
The study found that scientists use ChatGPT across multiple stages of research, including reviewing scientific literature, drafting and editing text, analysing evidence, developing software and generating research ideas.
Reading and writing together accounted for more than half of the average scientist’s scientific messages.
AI has also supported research ideation. Some scientists have used ChatGPT to generate potential research questions, while others have turned to the tool to refine ideas they had already developed.
According to the report, these applications could help researchers explore a broader range of topics and identify promising directions for further investigation.
However, faster task completion does not mean scientists can rely on AI outputs without review.
Researchers sought clarification, verification or revisions after around 32% of AI responses, indicating that human judgement continues to play an important role in evaluating and improving AI-generated results.
The study also identified persistent barriers to scientific productivity. Limited time or staffing was cited by 51.8% of respondents, while 43.8% identified funding as a constraint. Only 6.4% said they faced no major bottlenecks.
The findings suggest that AI could help researchers make more efficient use of their time, but faster workflows alone may not guarantee faster scientific breakthroughs.
Turning productivity gains into discoveries will also depend on researchers having adequate funding, personnel and the resources needed to conduct experiments.
Stay tuned for more such updates on Digital Health News