AI and the Enclosure of the Scientific Commons

Discussion overview

Commenters spent less time accepting the article’s enclosure thesis than arguing over whether AI is actually concentrating scientific power or making the literature newly usable. The strongest pro-AI accounts were practical: assistants can help a non-specialist search, translate, summarize, and connect papers across fields, turning nominally open research into something people can work with. Some see that as a genuine expansion of the scientific commons, not its destruction, and argue that open-weight models further complicate a simple monopoly story.

Skeptics countered that this accessibility is conditional on corporate models, scarce compute, and systems trained to produce agreeable summaries rather than reliable scientific understanding. They also challenged the grand claim that models contain all knowledge: much expertise is tacit, local, and never written down, so prompt-based tools cannot reach it. Optimistic predictions of an imminent scientific and cultural explosion drew particular ridicule after years of promised breakthroughs without corresponding marvels. Several readers agreed that both liberation and enclosure remain plausible, making firm forecasts premature. The thread’s sharper institutional concern was that scientific literature has never been fully unowned in practice, and AI could replace dispersed but imperfect access with dependence on a handful of firms controlling frontier infrastructure.

Points of disagreement

The core fight is whether AI opens science or encloses it. Supporters argue that AI makes an overwhelming body of open literature intelligible and useful to far more people, while skeptics argue that meaningful AI-enabled research depends on concentrated compute and proprietary systems, leaving users dependent on a few gatekeepers.