New research from Dartmouth reveals that artificial intelligence can now corrupt public opinion surveys at scale—passing every quality check, mimicking real humans, and manipulating results without leaving a trace. The findings, published in the Proceedings of the National Academy of Sciences, show just how vulnerable polling has become.

A SISSA collaboration between physics and neuroscience has developed a unified model explaining how the brain stores and retrieves perceptual memories. The study shows that a single neural network with Hebbian plasticity can generate opposite perceptual biases—attraction toward past experiences and repulsion away from them. The model reproduces human and rodent data across multiple tasks (working memory, reference memory, and a “one-back” task) without task-specific adjustments, suggesting that the underlying memory mechanism is shared and that only the readout of the network’s activity changes to support different cognitive functions.

Short-form videos can trigger suicidal thoughts in vulnerable viewers; new AI model predicts which videos pose higher risk before they reach large audiences; approach guides platforms’ moderation teams and improves user safety

For decades, it was impossible for clinicians to know which babies with major heart defects might eventually have developmental abnormalities, like ADHD and learning differences. Now, by adapting a simple and safe oxygen test, UC San Francisco researchers believe they can identify those most at risk, potentially leading to early detection and life-changing treatments.