The Private Capture of Public Genius
Points and comments are a snapshot, not live.
Modern AI labs profit from publicly contributed training data without compensation, echoing Bell Labs' past.
The essay compares Bell Labs' 1956 consent decree, which opened 7,820 patents royalty-free and spurred Silicon Valley, to today's AI frontier labs. Bell's monopoly-funded research produced breakthroughs like the transistor and UNIX. The author argues that AI models are trained on "the collective experience of humanity", public writing, code, and posts, without compensating contributors. OpenAI and Anthropic's revenues have grown rapidly ($47B projected for Anthropic in 2026). The essay proposes a "Corpus Fund" paying individual Americans for their data contributions, citing legal ambiguity around fair use in recent rulings like Bartz v. Anthropic.
What commenters are saying
Top commenters split into two camps: those praising the historical analogy but critiquing the narrow U.S.-only fund proposal, and those skeptical of any redistribution scheme. The top comment argues the fund should be global via the UN, not just American. The author responds, promising future essays addressing that. Another commenter flags the is-ought problem, noting the article shifts from describing what is to prescribing what should be. A third questions whether taxing AI output, rather than data input, is more sensible.