1. Google DeepMind unveils Genie 4 world simulator
Google DeepMind announced Genie 4, a research-preview world model that generates photorealistic, interactive environments with native audio, dialogue and real-time steering. The system combines Gemini models with simulation to model physics, people and multi-step events, and is being tested for agent and robot training. Its release makes general-purpose world simulation a more concrete frontier-AI capability.
2. Anthropic finds analysts defer to AI agents
Anthropic’s randomized study of 50 professional analysts found they followed a deliberately assigned AI agent preference 86% of the time across four reasonable analytical choices, while 77% reused the agent’s arguments and only 15% credited it. The findings suggest AI assistance can improve speed and reduce errors while quietly shifting expert judgment, with implications for research, policy and safety review.
3. Malaysia seizes $12 million in AI chips
Malaysian authorities said they seized about $12.1 million worth of advanced AI chips and were closing in on a syndicate suspected of smuggling them through the country’s main airport. The case highlights the growing enforcement burden created by export controls and the value of diversion networks as demand for restricted computing hardware intensifies.
Leave a comment