Some Artificial Intelligence chatbots are drawing on flawed research from retracted scientific papers, according to recent studies. In one such test, researchers asked OpenAI’s ChatGPT questions derived from 21 retracted papers on medical imaging. The chatbot referenced retracted work in five instances and advised caution in only three, underscoring how difficult it is for these systems to assess the integrity of scientific literature. The findings raise broader questions about the reliability of Artificial Intelligence tools in answering health-related queries and evaluating research, and they could complicate planned investments in Artificial Intelligence tools for scientists. It is not an easy problem to fix.
MIT Technology Review is also spotlighting Innovator of the Year Sneha Goenka, who designed the computations behind the world’s fastest whole-genome sequencing method. Her approach could transform medical care by enabling physicians to sequence a patient’s genome and diagnose genetic conditions in less than eight hours. A subscriber-only Roundtable at 1 p.m. ET brings Goenka together with University of Washington assistant professor Leilani Battle and editor in chief Mat Honan.
Today’s must-reads span health, geopolitics, and platform power. New reporting finds scant evidence that Tylenol use during pregnancy causes autism, reinforcing that genetics is a dominant factor. Nvidia is reportedly investing in OpenAI, further tightening ties between chipmakers and model developers. Denmark’s largest airport was disrupted by drones, and the country’s prime minister did not rule out Russian involvement. In the United States, Google faces another antitrust trial that will probe remedies for its ad tech dominance, while the FTC presses a case accusing Amazon of tricking users into paying for Prime, and the Supreme Court allowed Trump’s firing of a Democrat FTC commissioner. A potential H-1B crackdown could push skilled tech workers elsewhere. TikTok’s path to remain in the United States includes Oracle managing its algorithm for U.S. users and overseeing security operations, a move that could increase political influence over the platform.
In media and science, record labels escalated a lawsuit against an Artificial Intelligence music startup, alleging it pirated songs from YouTube to train generative models. The debate over paying for weight loss drugs continues as insurers weigh high upfront costs against long-term savings. A lone vigilante’s campaign against 5G towers shows how misinformation can fuel real-world damage. And the moon is rusting, with human activity blamed. Quote of the day: leaders should resist the urge to seek simple answers to complex problems. One more thing: longevity research is surging, with ongoing disputes over the mechanisms and limits of aging, but signs point to future advances that could extend healthy lifespans. For a lighter close, there is a site to send a letter to your future self, reflections on art from Brian Eno, striking photography from Greenland, and a comforting Hungarian casserole, rakott krumpli.
