#252 - GPT 5.6, Grok 4.5, Nemotron-Labs-Diffusion, AI 2040
Summary
Hosts Andrei Kerenkov and Jeremy Harris review major AI releases, including OpenAI’s GPT 5.6, xAI’s Grok 4.5, and Meta’s Muse Spark 1.1, highlighting an intensifying price war. They critique the US government’s ad hoc regulatory approach, noting inconsistent oversight between OpenAI and Anthropic. The episode also covers rising adoption of cheaper Chinese models like GLM and Meta’s new cloud infrastructure plans.
Topics discussed
Introduction and episode overview
Sponsors: OutShift and Box
US government greenlights GPT-5.6 release
Anthropic's evaluation of GPT-5.6 and safety concerns
xAI Grok 4.5 release and price war trends
MetaSpark 1.1, Muse Spark, and Chinese model growth
Meta's cloud infrastructure strategy and data centers
Compute economics, pricing pressure, and Grüns sponsor
FERC regulations and energy grid stress from AI
Open source models and diffusion decoding techniques
AI interpretability and global workspace theory
China curbing overseas access to AI models
AI 2040 Plan A and future AI development scenarios
Outro and closing remarks
Listen ad-free on Castria