As AI output floods the internet, will training data stop representing human work?
ThisFuzzBall078 · reddit · 2026-11-01
An amateur poster raises the synthetic-data feedback-loop question: LLMs improve by studying human-made examples, but as AI output becomes indistinguishable from human work, humans increasingly publish AI-generated content online. As future training data fills with AI output, will it stop representing the human distribution models are trying to learn — a real training-data problem? This is the popular framing of ongoing "model collapse" and data-contamination research worth following.
More from AGI Musings
- Could superintelligence park Mars in Earth's orbit instead of us flying there? — r0ck3t23 · 2026-11-01
- DIY UBI math with real data: about $2,000 per person after 30 years, not a living wage — edmrunmachine · 2026-11-01
- Technologists split into 18-month thinkers and 30-70-year thinkers, says dev — fkasummer · 2026-11-01
- Reddit suffers worst day since IPO, stock plunges 29% as CEO blames Google AI Mode for traffic loss — oran_ge · 2026-11-01
- Is human persistence a nonnegotiable condition of a good cosmic future? — danfaggella · 2026-11-01
- Grady Booch maps computing's six limits — and says ethics is OpenAI, Anthropic and xAI's present problem — Grady_Booch · 2026-11-01