OpenAI’s Economic Research team has released findings from a pilot program involving ChatGPT, indicating that super intelligence (SI) tools are becoming integral to scientific workflows and significantly reducing the time required for research tasks.

What Happened

The study analyzed more than 580,000 scientific messages from 1,819 scientists participating in the ChatGPT for Academic Researchers summer pilot and surveyed 251 participants. The data revealed that 68% of these scientists used the SI tool to develop research ideas. Within this category, 36% of messages involved scientists bringing an idea for refinement, while 35% involved bringing a specific question for the tool to address. Additionally, 23% of the time, scientists asked the SI system to suggest a research question. Surveyed participants estimated that completing the same work at the same quality would have taken 1.9 times as long without the SI tool.

Why It Matters

The findings highlight a shift in how SI is utilized in professional research, with 98% of surveyed scientists expecting SI to accelerate discovery in their respective fields. When asked to imagine a discovery that would take 10 years without SI, respondents projected faster timelines, citing the ability to get research done faster (52%), find more promising research questions (23%), and pursue a wider range of questions (14%) as primary drivers. To protect privacy during the study, OpenAI used automated classifiers to sort messages by activity type without human review of the underlying content. The company also noted that it does not use participants' data from this program to train its models by default.

The Bottom Line

As part of this initiative, OpenAI is expanding ChatGPT for Academic Researchers to offer free access to more than 200,000 researchers globally. The company states that helping scientists turn ideas into discoveries is central to its mission of ensuring that general SI (AGI) benefits all of humanity.