Governments, companies, nonprofits should invest in free, open source AI [pdf]
Points and comments are a snapshot, not live.
Open-source AI investment is needed to prevent a closed-AI oligopoly from locking down knowledge.
David Siegel argues that the fight for open-source AI mirrors the 1980s open-source software debate. He warns that frontier AI models are increasingly closed, and that this trend threatens to lock down scientific progress. He calls for government, corporate, and nonprofit investment in free and open-source AI, including public compute grants and a rule that AI built with public money is open by default. He distinguishes between releasing code to run a model (open weights) and releasing code that built it (true open source), urging the latter.
What commenters are saying
Commenters are split. Some argue that open-source AI will fail because training costs are prohibitive and contributions require compute, not just code. Others counter that commoditization is already happening, drawing parallels to Linux and databases. A few propose prizes or cooperative models. One commenter notes that many 'open' models are like compiled binaries, not libraries. Another criticizes the article's author for previously calling the AI buildout premature.