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.
January 30, 2025
In this week’s Duke AI Health Friday Roundup: AI consumes AI content, converges on generic representations; debut of AlphaGenome; heritability of human lifespan may be underestimated; measles infections spread in SC; gaming disorder or gambling disorder?; disclosure of AI use in medical publications; new approaches for restoring public trust in science; robot hand detaches from body, scuttles off by itself; much more:
AI, STATISTICS & DATA SCIENCE
- “It’s a particularly alarming predicament considering the tidal wave of AI slop drowning out human-made content on the internet. While proponents of AI argue that humans will always be the ‘final arbiter of creative decisions,’ per Elgammal, algorithms are already starting to float AI-generated content to the top, a homogenization that could greatly hamper creativity.” Futurism’s Victor Tangermann reports on recent research by Hintze and colleagues that demonstrates how AIs iterating on a task that incorporates both AI-created input and output tends to converge toward a bland average.
- “We present a reversible finger design that allows grasping from both sides, with finger placement optimized to distribute roles and satisfy grasping and crawling constraints. In this work, we show that a detachable, reversible hand reliably combines manipulation and locomotion, enabling beyond-reach multi-item retrieval during crawling, one-handed tool use, and seamless bridging of stationary manipulation with autonomous mobility for industrial, service, and exploratory robotics.” Bid farewell to sleep: In a paper published in Nature Communications, Gao and colleagues describe a robot with a detachable hand capable of crawling by itself and accomplishing grasping tasks while removed from its larger robot body.
- “The win ratio provides a single metric for quantifying a treatment effect measured by a hierarchical composite outcome. Hierarchical composite outcomes are of increasing interest because of their ability to order events by clinical importance and thus be more patient centered than simple composite outcomes while achieving similar increases in study power relative to using only the most important component outcome as the single primary end point.” An article by Cunningham, Fuest, and Burke published in JAMA as part of the journal’s Statistics in Medicine series walks readers through the use of the win ratio for evaluating hierarchical matched and unmatched composite outcomes.
- “On Wednesday, the researchers unveiled AlphaGenome in the journal Nature. They trained their A.I. on a vast wealth of molecular data, enabling it to make predictions about thousands of genes. For instance, AlphaGenome can predict whether a mutation will shut off a gene or switch it on at the wrong time — a crucial question for understanding cancer and other diseases….AlphaGenome will be useful…. But not all scientists trust A.I. programs like AlphaGenome to help them understand the genome.” In an article for the New York Times, Carl Zimmer reports on the debut of DeepMind’s AlphaGenome, a predictive tool for genomic research.
BASIC SCIENCE, CLINICAL RESEARCH & PUBLIC HEALTH
- “The findings provide evidence for the gambling hypothesis and suggest that the majority of the global literature on gaming disorder is at risk of contamination. This is owing to the conceptual distinction between gaming and gambling being unclearly presented for participants. In other words, it is often unknown whether participants have reported their gambling, gaming, or both behaviours when they have answered questions on disordered, problematic and other such classifiers of gaming.” A research report published in Royal Society Open Science by Adamkovič and colleagues explores whether an inadvertent conflation of gaming and gambling behaviors has tainted research aimed at measuring the existence and extent of gaming-related psychological disorders.
- “…current estimates of heritability [of intrinsic human longevity] are low—twin studies show heritability of only 20 to 25%, and recent large pedigree studies suggest it is as low as 6%. Here we show that these estimates are confounded by extrinsic mortality—deaths caused by extrinsic factors such as accidents or infections. We use mathematical modeling and analyses of twin cohorts raised together and apart to correct for this factor, revealing that heritability of human life span due to intrinsic mortality is above 50%.” A research article published in Science by Shenhar and colleagues suggests that analytical mistakes may have caused systemic underestimates of the heritability of human lifespan.
- “Most of those sickened are children and teenagers, health officials said. At least 23 schools in South Carolina have students in quarantine….’Without better vaccination coverage to achieve herd immunity more widely, we can anticipate that many more susceptible people will contract measles in South Carolina, that we’ll continue to see spread potentially to other areas of the state,’ Bell said.” NBC’s Erika Edwards reports on the as-yet uncontrolled measles outbreak in South Carolina, which has just recently passed the previous largest US outbreak in recent years.
- “Parents’ firearm injuries were associated with increases in psychiatric disorders and mental health visits in their children, as compared with closely matched youths without exposure to parental firearm injury. These changes were larger among children whose parents sustained more-severe injuries and among female youths.” A case-control study published in the New England Journal of Medicine by Karandinos and colleagues presents findings that suggest an increased mental health toll on the children of parents who suffer firearm-related injuries.
COMMUNICATIONS & Policy
- “Collectively these studies provide a useful snapshot of AI use in clinical research studies and other manuscripts….Although actual use is likely higher due to underreporting, AI use is clearly increasing, as both the JAMA Network and BMJ journals observed an increase over time with the most recently available estimates of approximately 6%. The majority of submissions used AI to improve writing and refine language, but the increasing availability of generative AI tools is likely to lead more investigators to use AI for more advanced tasks typically considered key, intellect-driven aspects of the scientific process…” An editor’s note in JAMA by Malani and Ross introduces a pair of articles, one by AlFayyad and colleagues and one by Perlis and colleagues, that examine disclosure practices by authors who use AI during the manuscript drafting process.
- “…unlike the Orange Book, the Yellow Book would not serve as a focal point for litigation about market entry. Instead, it would provide additional information about intended use, device risks, and strategic (or anticompetitive) firm behavior. The Yellow Book would also provide a source of information for researchers and scholars to evaluate the uncertain and understudied role of patents in device innovation and regulation.” A JAMA viewpoint article by Simon and colleagues proposes creating a patent database for medical devices analogous to the FDA’s Orange Book database for drug patents.
- “What if…university-based scientists and engineers tried to repair public trust by adopting a more inclusive understanding of technological innovation itself? Nowadays, we tend to focus on new, high-tech, transformative, and material interventions. But what if we adopted a more expansive approach that takes low-tech and even incremental solutions seriously, including those developed by members of the communities surrounding our campuses?” At Science, Shobita Parthasarathy suggests a new pathway for efforts aimed at restoring public trust in academic science.
