Worth reading
Outside work we decided to keep. The daily AI news feed drops an article after a week; anything a researcher here will still want in a year is promoted to this shelf instead, where it stays. Papers, lab write-ups, and notices worth your time.
See what is on the feed todayA 330-million-parameter time series foundation model, trained on more than a trillion time points, that now forecasts several related series at once rather than one at a time, …
The U.S. Food and Drug Administration today issued a discussion paper on considerations for the regulation of generative artificial intelligence (GenAI)-enabled medical devices, seeking feedback from interested parties …
Aurora 1.5 adds 22 more variables, hourly temporal resolution, and probabilistic ensemble forecasting to the Aurora foundation model, making it more useful for real-world weather, climate, and energy …
Two studies of laypeople using an AI dermatology tool. Naming a condition correctly rose from 8 to 23 percent against ordinary web search, but knowing what to do …
Two NHS studies, published in Nature Cancer, of AI reading screening mammograms: detection rose from 7.54 to 9.33 cancers per thousand women, with an estimated 46 percent fewer …
A plain linear model that predicts whether an EV charging port will be free in a given number of minutes, cutting prediction error by about 20 percent at …
A toolkit for training and auditing differentially private models in JAX, with the gradient clipping and noise machinery built in, so large models can be trained with a …