Science at the Frontier: How scientists use ChatGPT | OpenAI

Open original

OpenAI analyzed over 580,000 scientific messages from 1,819 researchers in a summer pilot of ChatGPT for Academic Researchers, and surveyed 251 of them. The main takeaway is that scientists use the tool across the whole research process, and most expect it to speed up discovery in their fields.

  • Surveyed scientists estimated the same work, at the same quality, would have taken 1.9 times as long without AI.
  • 68% of scientists used ChatGPT to develop research ideas, and 98% of those answering expect AI to accelerate discovery in their field.
  • Scientists stayed involved after delegating tasks, and their stated ambitions go beyond saving time to finding more promising questions. OpenAI is also expanding free access to more than 200,000 researchers.

OpenAI's Economic Research team looked at more than 580,000 scientific messages sent by 1,819 scientists during the summer pilot of ChatGPT for Academic Researchers, and surveyed 251 of them. The headline numbers are that 68% of scientists used ChatGPT to develop research ideas, and 98% of those who answered a question about discovery timelines expected AI to speed up discovery in their field. Participants estimated that the same work, at the same quality, would have taken 1.9 times as long without AI. Usage spanned the whole research process, from reviewing literature to analyzing evidence, writing code, and communicating findings. The method matters for how much weight to give these figures. Researchers did not read the messages. Automated classifiers sorted each one by activity type, such as literature review, coding, or data analysis, and the team saw only those categories. Each scientist's share of messages was averaged with equal weight, so a heavy user doesn't dominate the picture. Participant data was not used to train models by default. The sample is self-selected, though: these were people in a pilot program, and the speedup figure is the scientists' own estimate, not a measured timing. That's worth keeping in mind before treating 1.9x as a hard number. What stands out to me is that the stated ambitions go beyond saving time. Respondents talked about finding more promising questions and pursuing a wider range of research, which is a different claim from "doing the same work faster." Alongside the report, OpenAI is expanding the program to free access for more than 200,000 researchers worldwide. The study is a useful snapshot of how one group of early adopters works, but it can't yet tell us whether those expected gains in discovery actually show up in published results.

Ask about this article