THE ESSENTIALS IN 30 SECONDS
- OpenAI will hand out 50,000 new grants, each covering a researcher and up to two collaborators, and now aims to reach 200,000 researchers, double its July target.
- New grants come with ChatGPT Plus instead of the Pro tier given to pilot users; applications open November 23 and are first come, first served.
- In OpenAI's study of more than 1,800 pilot scientists, survey respondents estimated their work would have taken nearly twice as long without AI, a self-reported figure.
OpenAI announced on November 16 that it is enlarging ChatGPT for Academic Researchers, the programme it piloted in July to give scientists free access to its assistant. Alongside the expansion, the company released a study of how pilot participants actually used the tool.
More grants, a cheaper plan
OpenAI plans to award 50,000 additional grants. Each one covers a researcher plus up to two collaborators, so the batch could reach as many as 150,000 people. The overall goal is now 200,000 researchers, twice the figure set when the pilot opened. According to The Next Web, about 20,000 researchers in 33 countries have joined so far.
The terms are changing. Pilot participants received ChatGPT Pro, OpenAI's most expensive tier, whereas new grantees will get ChatGPT Plus. OpenAI says a review of usage showed that most pilot users would have stayed within Plus limits, and that the switch lets it support more teams. Selection also changes: the lottery used for the pilot is dropped, and applications, opening on November 23, will be accepted in order of arrival from eligible candidates. OpenAI says data from the programme is not used to train its models by default.
What the study found
The accompanying report, "Science at the Frontier", was produced by OpenAI's economic research team and draws on more than 1,800 scientists from the pilot. The Next Web reports that reading and learning accounted for 28% of their science-related messages on average, and writing for 27%. Surveyed participants estimated that the same work, at the same quality, would have taken nearly twice as long without the assistant.
Those time savings are the respondents' own estimates rather than measurements, and OpenAI acknowledges that the study does not show that AI speeds up scientific discovery. It also points out that money, staff and laboratory capacity still limit how many new ideas a team can test.
Context
The move adds to a series of science initiatives from OpenAI, including free seats on its GPT-Rosalind biology model for eligible US university labs and a $5 million contribution to a Simons Foundation effort helping mathematicians check AI-generated results. Competitors are pursuing the same audience: The Next Web points to Anthropic's Claude Science workbench, launched in June, and Google DeepMind opening its science models to 200 European researchers in October.
