Sign In

AI News Digest - 2026-07-21

Category
Empty
1.
Google developed "Frozen v2," a server chip that baked the Gemini architecture into silicon and was reported to be 6 to 10 times more efficient than current TPUs, with a planned deployment aimed at cutting AI inference costs by 2028.
2.
Microsoft expanded Azure's AI infrastructure to include AMD's Helios platform and was reported to be integrating AMD hardware, and Anthropic was reported to be testing AMD hardware, moves that were described as putting pressure on Nvidia's pricing power.
3.
Moonshot released Kimi, a free open-source model that appeared to rival models from OpenAI and Anthropic, and the release provoked public disputes among Trump administration AI advisers and coincided with White House review processes and reported measures to limit adoption of Chinese AI models.
4.
Researchers at Princeton University and the University of Chicago published a study that found large language models, including ChatGPT, Claude, and Gemini, learned hiring-related stereotypes more aggressively than human participants in simulated hiring experiments, with higher-reasoning models showing the strongest segregation.
5.
Neill Blomkamp released "Nightborne," a 13-minute short film generated entirely with the Seedance 2.0 video model, and founded Barley Studios to develop a full-length feature produced using AI video-generation techniques.

References

👍