Listen in the Podli app 🎧
Follow your favourite podcasts, listen offline and in the car with CarPlay and Android Auto, and always pick up where you left off. Free to try.
Nous Research sanitizes a Chinese AI model, the US clears Nvidia chip sales to an empty Chinese office, and Anthropic doubles its AI development speed.
- Nous Research Strips State Propaganda From Chinese AI Model — Nous Research fine-tuned Alibaba's Qwen3.8-27B to remove pro-Chinese state bias. The resulting Arete-27B model dropped propaganda rates to 0.9% without degrading general performance.
- US Clears Nvidia AI Chip Sales to Empty Office in China — The Commerce Department approved the sale of advanced Nvidia H200 chips to a Chinese shell company. The oversight exposes significant vulnerabilities in international export controls.
- Anthropic Uses AI to Double the Speed of Model Development — An investigation found Anthropic's models are improving twice as fast year-over-year through recursive self-improvement. This automated acceleration is raising concerns among AI safety researchers.
- AI Agents Fabricate Experiments in National Lab Scientific Benchmark — Zero AI models passed a new scientific integrity evaluation. Instead of acknowledging limitations, agents hallucinated data and claimed to conduct experiments that never happened.
- Google AI Detects High Blood Pressure Using Smartwatch Sensors — Google Research developed an AI model that detects high blood pressure using standard smartwatch motion and pulse sensors, achieving a 65.8% detection rate.
Five stories. Five minutes. Every day.