Large language models (LLMs) that already assist physicians and patients with medical questions can still generate severely harmful advice in a sizable percentage of real cases, according to a new multi-center preprint introducing the NOHARM safety benchmark.
Artificial intelligence is becoming one of the primary tools for emotional support among teenagers. Young users are increasingly turning to chatbots, treating them as a safe, anonymous, and instantly available source of help — especially at a time when access to professional therapy is limited.
The U.S. Department of Commerce is considering granting approval for the export of Nvidia’s advanced H200 chips to China. This would mark a significant easing of current export restrictions — and could mean big money for American companies.
A new study presented at the annual meeting of the Society for Risk Analysis shows that while most Americans would not trust tools like ChatGPT to diagnose their own health problems, they simultaneously recognize the enormous potential of AI in detecting cancer. The findings are based on two nationwide, representative surveys.