01/17/2026
(From John Williams)-- I got an email regarding an AI interpretation of a scientific paper I authored and placed on BioRxiv. After a bit of vetting to make sure it was real, it was real, so I looked at the ideas and hypotheses generated by it. They were ludicrous. The AI placed Ivory-billed and Pileated Woodpeckers both in the genus Campephilus. It also generated a hypothesis that the Ivory-billed would make different kent sounds based on whether it was eating beetle larvae or plant material. This is what I would call "far-fetched" which again is in the realm of ludicrous. I was stunned that AI got its interpretation so wrong. Below is the letter. It is also timely to post links to what I have contributed in the last four years to the IBWO search and knowledge set, below the letter. :) John
"Dear Author,
We are emailing you from Knowledge Lab (https://knowledgelab.org/) at the University of Chicago. We are studying how AI helps scientists extend their work in new directions. We came across your preprint "Spectrograms of ivory-billed woodpecker (campephilus principalis) kent sounds from seven expeditions show 587 hertz as a pattern" and applied custom AI models on it, which generated 5 new extensions (i.e., ideas proposed by AI based on the same context of your paper). Our pilot study of 400 scientists (excluding you) indicates that around 60% found at least one of the proposed extensions thought-provoking and actionable.
Would you like to see these extensions? They are included in the survey link below. The extensions are absolutely free and with no obligations -- although we would sincerely appreciate your thoughts on their quality. If you consent to review them and participate in the survey (link below), you may do so at your convenience within the next 7 days. If you do not consent, please feel free to disregard this email (and we will not collect any data from you and your research).
Click the survey link below, review them, and fill out your thoughts:
Take the survey
Or copy and paste the URL below into your internet browser:
https://uchicago.co1.qualtrics.com/jfe/form/SV_1yPc7XXPRcCh6Xs?Q_DL=n0vxYM7hgSBcamZ_1yPc7XXPRcCh6Xs_CGC_fL0keUwjj8jc5xD&Q_CHL=email
We would like to assure you that this email is not phishing. This project is conducted in good faith and has no commercial purpose (under the regulation of UChicago IRB25-1372, titled โThe Capabilities and Potential of AI for Automating Scientific Idealization: A Large-Scale Human-in-the-Loop Studyโ). Our goal is simply to help train AI systems to better understand how scientists evaluate good science. We will share the final draft, technical details, and aggregate results from scientists once the study is complete. If you have any questions or would like to verify the legitimacy of this project, you can simply reply to this email, which will reach Honglin Bao, a data science PhD student in Knowledge Lab at the University of Chicago, at [email protected]
We sincerely appreciate your feedback.
Warm regards,
Knowledge Lab & Data Science Institute
University of Chicago"
(The actual AI-generated results are not accessible after I completed and sent in a (negative) evaluation.)
Links to papers-- https://www.biorxiv.org/search/ivory-billed%252Bwoodpecker
https://www.scienceopen.com/hosted-document?doi=10.14293/PR2199.001366.v1