John Carmack about open source and anti-AI activists
A prominent game developer’s claim that his million lines of open-source code were always intended as a “gift to the world” has reignited debate over how AI companies use open-source software for training. Many developers counter that most licenses attach conditions such as attribution or copyleft, and that large AI models effectively “launder” those obligations while concentrating profits and power in a few firms. Beyond licensing, contributors voice broader fears about enshittification of communities, erosion of developer livelihoods, and the growing centralization of AI infrastructure.
Open Source: Gift vs Reciprocity
- One camp agrees with framing OSS as an unconditional gift: once code is released, any use (including AI training and corporate profit) is acceptable.
- Others stress that most OSS licenses are conditional gifts: MIT/BSD require attribution; GPL/AGPL require derivatives to remain free. They see reciprocity, not pure altruism, as the core norm.
- Several distinguish between “code dumps” of end‑of‑life products and long‑term, labor‑intensive project maintenance; they argue these are different experiences and produce different attitudes.
AI Training on OSS and Licensing
- Many argue AI training ignores license conditions (attribution, copyleft), effectively “laundering” GPL and other restrictive code into proprietary outputs.
- Disagreement over whether LLMs and their outputs are “derivative works” in a copyright sense; some say yes (thus GPL applies), others say existing legal opinions favor training on lawfully obtained data.
- Some propose new licenses (“OSS + no ML training”) or revenue‑share / anti‑AI clauses, though this would not meet existing “open source” definitions.
Profit, Power, and Labor Concerns
- A major thread is capital vs labor: AI firms capture value from OSS and public content while threatening to devalue programmers’ and creatives’ livelihoods.
- Several connect this to broader wealth inequality and fear of white‑collar job losses, potential economic shocks, and political radicalization.
- Others respond that OSS always allowed commercial use; companies running Linux or using libraries for profit already did this without paying maintainers.
Quality, Community, and “Slop”
- Maintainers complain about AI‑generated “slop” PRs and issues, saying it clogs community channels and reduces meaningful collaboration.
- Some worry AI encourages disengagement from “social coding” and erodes apprenticeship, recognition, and motivation in OSS communities.
Views on Copyleft’s Future
- Some see AI as undermining copyleft and “destroying” the GPL ecosystem; others welcome this, arguing copyleft has “served its purpose.”
- There is disagreement about whether protecting “source code” itself matters, or whether only people’s rights and material conditions should be the focus.
Wealth, Privilege, and Perspective
- Multiple comments note that well‑off, famous developers can more easily treat their work as gifts and feel safe about AI, whereas many rank‑and‑file developers cannot.
- Some see current pro‑AI open‑source rhetoric as shaped by this privilege and by direct involvement in AI startups.
Broader AI Activism & Ethics
- Distinct concerns are raised about:
- Artists’ works being scraped without consent.
- Centralization of power in a few AI labs vs. hopes for open models.
- Enshittification, surveillance, water/energy use, and downstream harms (e.g., military uses).