Topics

Confiding in AI chatbots backfires. Stanford research finds users with smaller social circles feel more isolated, not less, after seeking emotional support.

Drowning in research papers? Generative AI is reshaping how scholars search, summarize, and synthesize—transforming information overload into discovery.

LLMs hallucinate confidently. Uncertainty estimation flags errors in NLP & code—strong on open-source models, misses nuances in GPT-4o. A plug-and-play step toward safer, self-aware AI.

A study of AI language models finds that safety training meant to stop bots claiming to be conscious also suppresses broader human-like beliefs and empathy toward non-human minds.

The company reports large gains on tests of computer use, coding, science, and cyber defense, and is opening the model first to a limited group of users.

Transformers learn in context, not through retraining. New research proves they are universal learners, capable of adapting to any task directly from examples, masked or unmasked.

Why do LLMs sound smart but often get it wrong? A decades-old knowledge system offers a blueprint for truly dependable AI.

Researchers plugged several machine-learning models into one safety-checked loop on a working tokamak and showed it could act far faster than a person.

Silicon Data's benchmark index hit $0.97 on August 31—a record low and less than half the summer peak as the AI price war intensifies.

OpenAI's Astra is the first model to cross its "Critical" cybersecurity threshold. It found two zero‑days autonomously. Advanced cyber features will be limited to testers.