Matthew Stoller says copyrighted training data is not fair use and licensing could reshape AI
GaryMarcus · x · 2026-07-26
Gary Marcus reposted Matthew Stoller arguing that training on copyrighted materials is not fair use.
The core claim is that if licensing becomes mandatory, it could create new content business models—and that AI systems would still ultimately improve, even under a licensing regime.
More from AGI Musings
- A simple AGI-era reminder: stay curious, not overconfident — yunta_tsai · 2026-07-27
- AGI could become a 10% share of global GDP, author argues — teortaxesTex · 2026-07-27
- Reddit asks when AI will become disruptive enough for governments to intervene — shadowt1tan · 2026-07-27
- What happens if AI can implant nostalgia for people you never met? — PierceLilholt · 2026-07-26
- What happens when AI decides which of your memories get amplified? — PierceLilholt · 2026-07-26
- U.S. traffic deaths hit 36,640 last year as Waymo-style vehicles promise further declines — SuzKP · 2026-07-26