The MAD Podcast with Matt Turck//Ryan Greenblatt
AI Could Take Over in 2029. Is It Already Too Late? | Ryan Greenblatt
Ryan Greenblatt argues the most plausible failure mode for advanced AI is not abstract superintelligence, but competent scheming that leads to takeover if AI R&D and industrial capacity are automated too quickly. He lays out a 2029-focused risk timeline, a controversial U.S.-China compute deal in AI 2040 Plan A, and a technical agenda centered on AI control, monitoring, and alignment-faking prevention.
