Last Week in AI #250 - Mythos Mess, GPT 5.6-Sol, GLM 5.2
Summary
Hosts Andrei Kurenkov and Jeremy Harris discuss the US government’s emerging AI licensing regime, highlighting Anthropic’s cleared release of MIFOS and OpenAI’s gated rollout of GPT-5.6. They analyze OpenAI’s new Jalapeno chip and Amazon’s potential entry into custom chip sales, while debating the geopolitical implications of AI treaties and the risks of models optimized for cyber capabilities.
Topics discussed
Intro: AI treaties, national security, and episode preview
OpenAI vs Anthropic: Licensing, benchmarks, and cheating
Government pressure on Meta and AI regulation politics
Anthropic launches Claude Tag for Slack integration
OpenAI's custom AI chip design and Amazon's silicon strategy
Micron, SK Hynix, and the HBM memory market shift
SpaceX/XAI deals and Alibaba's Qwen model updates
AI impact on labor markets and workforce training bills
AI safety: DeepMind's control framework and OpenAI politics
Anti-AI populist movements and political polarization
Research: Scaling laws, Plato's forms, and interactive models
AI in Hollywood, lobbying, and podcast outro
Listen ad-free on Castria