The Next Wave - NotebookLM Just Killed After Effects? + ChatGPT vs Anthropic War
Summary
Matt Wolfe and Joe Fear analyze recent AI model releases, including OpenAI’s GPT 5.3 and 5.4, and Google’s Gemini 3.1 Flash. They explain training phases like RLHF and distillation, highlighting how new updates improve conversational flow, context windows, and agent efficiency. The episode also covers Notebook LM’s new cinematic video generation feature available on Google’s Ultra plan.
Topics discussed
Intro: Breaking down the week's biggest AI stories
Grok update vs. new OpenAI and Google models
How AI models are trained: Pre-training, Fine-tuning, RLHF
GPT 5.3 Instant: Less cringe, more direct answers
GPT 5.4: 1M context window for coding and agents
Tool search efficiency and the rise of AI agents
Testing GPT 5.4: Oscar-winning movie recommendations
Gemini 3.1 Flash: Fast, cheap, and great for apps
Summary of the three new model releases
Notebook LM update: Cinematic video generation
Notebook LM demo: 'Birds Aren't Real' conspiracy video
Anthropic/Claude: Pentagon blacklist and user migration
Netflix partners with Interpositive for AI filmmaking
Meta Ray-Ban privacy scandal: Footage reviewed in Kenya
Listen ad-free on Castria