AI Training: Synthesis Like Humans, or Theft? A Sharp Copyright Clash on X
JHochderffer · x · 2026-10-13
A heated X exchange over AI training data: the author argues AI doesn't copy books for resale but synthesizes multiple sources into new output, "just like humans," driving productivity and scientific gains — "if AI is theft, the thief broke in and left their treasure chest." The quoted reply pushes back: humans can't legally access the movies, music, and books AI trains on for free — we pay just to perceive them, let alone exploit them commercially. A concise snapshot of the core copyright debate.
More from AGI Musings
- DeepRSI benchmark pushes agents to 48-hour ML tasks to measure recursive self-improvement — garrytan · 2026-10-13
- Bindu Reddy: only competition will force Anthropic and OpenAI to release their internal models — bindureddy · 2026-10-13
- Study: micro entities and individual inventors are most likely to use AI in patent filings — robseamans · 2026-10-13
- Grad student: academia is cooked and prestigious jobs may only have 2 years left — felpix_ · 2026-10-13
- Stanford Analyzes 250K Claude Conversations: Most Users Collaborate Deeply, Not Shortcut — StanfordHAI · 2026-10-13
- Karpathy says AI builds everything at OpenAI and Anthropic — yet both keep hiring thousands — bindureddy · 2026-10-13