#252 - GPT 5.6, Grok 4.5, Nemotron-Labs-Diffusion, AI 2040
Summary
Hosts Andrei Kurenkov and Jeremy Harris review recent AI developments, including the releases of GPT-5.6, Grok 4.5, and Meta’s Muse Spark 1.1. They analyze the intensifying price war among providers and criticize the US government’s inconsistent, ad hoc regulatory approach to model approvals. The episode also covers Meta’s new cloud infrastructure plans and the rising adoption of Chinese open-source models, highlighting associated security risks.
Topics discussed
Intro and industry release waves
Sponsors: Factory and Outshift by Cisco
GPT-5.6 release and government pressure
Critique of AI safety discourse and politics
GPT-5.6 capabilities and jailbreaks
Grok 4.5 and competitive landscape
Market demand and Anthropic pricing
Meta's Llama model performance and safety
Instagram AI edits and Chinese model adoption
National security risks of open source models
Meta's data center strategy and compute
Compute infrastructure economics and fluidity
Energy constraints and grid regulation
New inference: Blockwise diffusion models
Tencent's open source model release
Anthropic research on model consciousness
China's shift away from open source AI
AI 2040 predictions and future outlook
Listen ad-free on Castria