
AI Inside Bill Gates Has Three Ideas To Fix AI
47 snips
Aug 27, 2026 They debate Bill Gates's three policy ideas to slow AI and the political tradeoffs involved. They cover Apple’s new M6 and M5 Ultra machines built for local AI workloads. They explain the rise of open-weight models and local inference with Perplexity and Nvidia’s portable agent. They highlight humanoid robots sprinting fast and a robot that can draw blood. They touch on AI in China, AI chips, data fights, and industry claims.
AI Snips
Chapters
Books
Transcript
Episode notes
Use Local Agents To Eliminate Token Costs
- Run agents locally when you need always-on, low-token-cost workflows by using Perplexity's Portable Computer concept.
- Jason Howell explains Portable Computer runs fully local models (Quinn 3.8, pplx Branson) to avoid cloud token costs.
Local Agents Still Limited By Target Hardware
- New local-agent products are initially hardware-locked; Perplexity's portable agent runs on NVIDIA DGX Spark Linux boxes today.
- Jeff Jarvis warns Apple Macs aren't yet supported for this Perplexity portable setup, limiting consumer adoption.
Gates Says Tech Is Now Moving Too Fast
- Bill Gates warns AI pace now feels too fast compared to past tech eras and urges caution.
- Jason Howell notes Gates went from believing tech moved too slowly to saying, now technology is moving too fast, faster than we've prepared for.



