AI Health

Friday Roundup

The AI Health Friday Roundup highlights the week’s news and publications related to artificial intelligence, data science, public health, and clinical research.

October 2, 2026

In this week’s Duke AI Health Friday Roundup: tiny model tackles racehorse names; watermarking AI-designed proteins; stresses from measles outbreak ripple through health systems; weighing risks & benefits of menstrual tracking apps; role of AI central to priority dispute over reverse transcriptase work; taking a stand against using AI-generated responses in behavioral research; much more:

AI, STATISTICS & DATA SCIENCE

Cropped photograph showing a synthetic biology research station in a laboratory at the NASA Ames research center, showing a desk with a binocular microscope, petri dishes, and other labware on the desk and in the background. Image credit: Alexander Van Dijk (CC BY 2.0)
Image credit: Alexander Van Dijk (CC BY 2.0)
  • “Proteins designed by artificial intelligence could soon carry a hidden signature of their machine-generated origins, thanks to a watermarking system described today in Nature…. The technology, developed by researchers at Google DeepMind in London, borrows a trick that is already used to identify AI-made content across images, video, audio and text. It involves weaving a faint statistical tell into both the amino-acid sequence and 3D shape of computer-designed proteins — without noticeably compromising their function.” Nature’s Elie Dolgin reports on recent research into “watermarking” synthetic proteins that have been designed by AI.
  • “…it’s true that (even if these attempts weren’t anything to worry about) that future versions of these tools could be used to do harm. And that’s true of any new technology. I’ve been low-level worried about this since well before the days of AI assistants, and their advent definitely doesn’t make me any less worried. That said, designing a horrible pathogen is – thank God – a whole lot harder than it looks.” At his In the Pipeline blog, Derek Lowe evaluates recent concerns that AI could be used to deliberately craft dangerous biological or chemical products.
  • “The New York Times reviewed Slack messages, figures and other materials produced by Dr. Mestre documenting his research and his conversations with Claude over the past several years. They seem to show that the A.I. was extensively involved in his research….Dr. Mestre said he could not be sure that Anthropic had relied on his work to seek out jumbotrons. But he was skeptical, he said, that the A.I. agents could have managed to find them so swiftly.” New York Times science writer Carl Zimmer explores the strength of Anthropic’s recent claim that the company’s Claude AI has made original scientific discoveries in biology.
  • “Given this early data, how should we train the scientists of the future? We need to address the use of AI in the scientific process with learners early in their journey. Within my lab, we walk through the scientific process from idea generation to paper publication and discuss the appropriate use of AI at each stage. The goal is not to completely eschew AI, but, rather, to make sure trainees have the proper foundation and critical thinking skills to do science.” In an editorial for NEJM AI, Roxana Daneshjou considers the implications of AI use for the training and development of future clinicians and scientists.

BASIC SCIENCE, CLINICAL RESEARCH & PUBLIC HEALTH

Two hand-sketched, suited people chat, one in an armchair facing their back at us and holding a cigarette, the other standing, holding a pipe. There are large flowers above the armchair. The hand-drawn image is superimposed with a textured rendering of a sad-face emoji covering the standing person's face. A seemingly random network of nodes hovers above the person in the armchair. Image credit: Marcin Wilkowski / https://betterimagesofai.org / https://creativecommons.org/licenses/by/4.0/
Image credit: Marcin Wilkowski / https://betterimagesofai.org / https://creativecommons.org/licenses/by/4.0/
  • “The appeal is obvious. Silicon samples are cheaper and faster than recruiting human participants, and they let researchers study sensitive topics such as attitudes to drug use or bigotry without the ethical constraints. But are silicon samples broadly representative? We argue here that they are not—and that until three specific, fundamental concerns are resolved, the field should not accept AI output as a substitute for human data.” A perspective article published in PNAS by Hoover and colleagues stakes out the position that AI should not be used to simulate human responses in behavioral research.
  • “For many hospitals, measles’ return means heightened vigilance to keep the most vulnerable safe. Geisinger Medical Center in Danville, Pennsylvania, had to update visitation rules to its neonatal intensive care unit in August following a potential measles exposure from a hospital visitor. Now, only a baby’s parents or legal guardians are permitted to visit the NICU, and only if they do not have signs or symptoms associated with measles.” Wired’s Emily Mullin reports on the strains being felt by hospitals and health systems as the US measles outbreak continues to spread.
  • “Many of the claimed benefits remain limited or uncertain, yet these technologies and practices are already being marketed, adopted, and endorsed. In a context where women and girls continue to face barriers because of gendered social and structural factors, it is tempting to celebrate this interest in menstrual cycles…. Yet selling apps and wellness programmes as a fix for longstanding gender inequalities in healthcare and physical activity offers ineffective tech solutions rather than public investments in health and developmental opportunities for women and girls.” A review published in BMJ by Lee and colleagues suggests that on balance, menstrual tracking apps may be exposing users to more risks than benefits.

COMMUNICATIONS & Policy

Watercolor painting of a woman riding in a horse race by Frances Elizabeth Wynn, courtesy of the Museum of New South Wales.
Painting (ca 1850) byFrances Elizabeth Wynn, courtesy of the Museum of New South Wales
  • “Specifically I’m using a tiny neural net called char-rnn, originally introduced by Andrej Karpathy in 2015, and updated by my friend Dylan Hudson to work with modern libraries. I installed it on my laptop, and then for training data I turned to a dataset that I had fun with in the past, a list of 42,000 registered racehorse names from the Jockey Club. Char-rnn runs only on my cat hair-covered laptop, and it has no general internet training, only what I gave it.” It’s time to get your bids in for win, place, and show as Janelle Shane’s AI Weirdness kicks off Botober with a selection of potential racehorse names generated by a tiny LLM.
  • “There was no significant difference in comprehension of key trial details between groups receiving the comic-enhanced and standard information resource, evidencing comparable understanding. The majority of children (77%) and over half of parents (56%) preferred the comic-enhanced resource. Most parents (74-100%) felt the comic-enhanced resource helped them explain the CARE study to their child(ren). There was no significant difference in study progression or consent duration between groups.” In a research article published in the Journal of Visual Communication in Medicine, Kearns and colleagues report results from a randomized study that evaluated the use of a comic as a decision-making aid for children and parents who are weighing whether to take part in a clinical trial.
  • “I believe that science derives its authority from something more than mastering the pipeline from data generation to publication. Rather, it comes from interactions within a community of people trained to interrogate evidence and answer for the conclusions they draw from it. And it is time to formalize what this means, and which sorts of human judgment cannot (or should not) be offloaded to AI. …If we do not act, then a full replacement — completed plank by plank — may happen before we know it.” In an essay for Undark, C. Brandon Ogbunu unpacks a thought experiment from classical antiquity to interrogate the growing use of AI tools in science.

Sites@Duke Express is powered by WordPress. Read the Sites@Duke Express policies and FAQs, or request help.