

How to Get to AGI
477 snips Sep 15, 2025
The journey to achieving AGI is packed with challenges, with AI showing PhD-level reasoning but faltering on basic tasks. Experts debate current capabilities while pinpointing the necessity of continuous learning and memory as crucial breakthroughs. Coding agents and real-world data usage may be key steps forward. The podcast also touches on ongoing lawsuits regarding AI impacts, leadership shifts at Apple, OpenAI's deal renegotiation with Microsoft, and significant layoffs at xAI, reflecting the ever-changing landscape of AI.
AI Snips
Chapters
Transcript
Episode notes
Cloudflare's Prediction On Content Deals
- Cloudflare CEO Matthew Prince predicted Google will pay content creators for crawling and training models within a year.
- He argued tech solutions, not legal fights, will shape web content economics.
Jagged Frontier Of Capability
- Demis Hassabis calls current models "not PhD intelligences" because they mix high-end capabilities with trivial mistakes.
- He argues true AGI should perform across the board at PhD level, not just on select tasks.
AGI Label Vs Business Value
- The term AGI matters most to labs pursuing general intelligence but is less useful for businesses focused on task-specific value.
- For companies, jagged frontiers determine how much autonomy you can safely grant agents.