In work reported Sept. 16 in Nature, researchers used Paper2Agent to convert 74 of 100 computational biology papers into working tool sets, Science News reports. When a paper comes with code, methods and data, the system packages them for an AI assistant to use with new data. The remaining papers faced obstacles such as missing code or data, or software that would not run.
Paper2Agent packages a paper’s associated code into reusable tools, tests them against reference outputs and writes step-by-step instructions for an assistant to follow, according to Science News. That differs from a general-purpose assistant generating new code on the fly.
For the papers the researchers converted, the tool sets ran analyses of the published work and answered questions about the results more accurately and more than twice as fast as the same agent working from the original papers and accompanying code, Science News reports. The comparison concerns those converted papers.
James Zou, a computer scientist at Stanford University, called the system a “virtual author of the paper,” available around the clock. Peter Clark, a researcher at the Allen Institute for Artificial Intelligence who was not involved in developing Paper2Agent, called it “an exciting step” but said it specializes in reusing methods and code rather than answering general questions. For open-ended questions, Clark said he would still contact a paper’s authors about things they had not put in the paper, Science News reports.
