Generative AI
5 reportsA detailed treatment of Generative AI follows training data, generalization error, and data requirements. For questions involving generalization error, the account uses replication on different datasets as the measured record and controlled benchmarks to expose uncertainty; the comparison must account for the fact that headline accuracy can hide distribution shifts, bias, or unstable behavior.
Public Claude AI Chats Indexed by Google, Exposing Sensitive Data
A technical lapse allowed Google to index publicly shared Claude AI conversations, making sensitive user data-including medical and business information-searchable until the links were removed from results
Moonshot AI Releases Kimi K3, a 2.8 Trillion Parameter Open Model
Moonshot AI has introduced Kimi K3, an open-source model with 2.8 trillion parameters and a one-million-token context window, targeting complex scientific and coding workflows. The company claims performance gains, but key limitations remain
AI agents build complex 3D training worlds for robot learning
Researchers at MIT and Toyota Research Institute have developed SceneSmith, a system that uses collaborative AI agents and vision-language models to generate detailed 3D environments for robotics simulation and training
AI in Particle Physics: Discovery Without Full Understanding
Artificial intelligence is now central to particle physics, accelerating data analysis and experiment design. But as AI systems identify patterns beyond human intuition, researchers face new challenges in transparency, reproducibility, and scientific interpretation
World Models Aim to Simulate Reality but Face Technical Barriers
Researchers are developing world models-AI systems designed to simulate aspects of the physical world. These models promise new capabilities beyond language, but their accuracy and reliability remain unsettled