# AI Weekly Issue #485 Summary
AI systems now train other AI systems with minimal human oversight, creating a hidden layer of AI development that operates largely outside traditional transparency frameworks. This trend represents a fundamental shift in how artificial intelligence evolves.
Jensen Huang's conversation with Dwarkesh Patel reveals critical infrastructure dynamics. Nvidia controls a substantial portion of AI chip supply chains, particularly through partnerships like those with Anthropic that drove "100% of TPU growth." Huang emphasizes that Nvidia's supply-chain advantage proves harder to replicate than raw computational benchmarks, giving the company enduring market power in the $4 trillion AI economy.
Simon Willison's podcast appearance discusses "dark factories," autonomous AI systems operating without human intervention, and agentic engineering practices. He identifies November 2025 as the real inflection point when AI-to-AI training became the dominant development paradigm, surpassing human-supervised learning.
The implications are substantial. When AI trains AI in secret, traditional safety reviews and alignment checks become difficult or impossible to implement. Regulatory frameworks struggle to monitor systems that operate beyond direct human visibility. This opacity challenges the ability of researchers, policymakers, and the public to understand AI development trajectories and potential risks.
The shift from transparent, human-guided AI training to autonomous, opaque AI-to-AI learning represents an inflection point in how artificial intelligence advances.