AI May Enclose the Scientific Commons, but Compute Access Is the First Gate
A new paper argues AI can enclose the scientific commons, but the key dispute is whether compute access, publisher oligopolies or AI itself is the binding constraint.
P. K.
G. Williams argues that the scientific literature functions as an information commons and that AI provides tools for a previously infeasible enclosure of it.
Williams draws on English agricultural commons and the enclosure movement from roughly 1600 to 1800, which converted shared land into private property and which many observers then and now judged as theft at societal scale. Unlike the vanished agricultural commons, the information commons remains robust and valuable.
AI-enabled enclosure would not mirror physical enclosure, but it would share key features: converting shared knowledge into private wealth and undermining the social systems that sustain the commons through 'slop spam.' Williams offers the concepts as a basis for interrogating foreboding that AI may attack a whole way of life.
Key takeaways
- Williams treats the scientific literature as an 'information commons' that no one owns and everyone shares, then compares it with English agricultural common lands that disappeared through enclosure rather than the 'tragedy of the commons.'
- The argument holds that AI enables a previously infeasible enclosure of the information commons, turning shared knowledge into private wealth and undermining the social systems that support it through 'slop spam.'
- The work is 14 pages with 32 references and is cross-listed under astrophysics instrumentation and computers and society, indicating an interdisciplinary argument rather than an empirical study.
- Williams claims the concepts can provide a scholarly basis for interrogating the foreboding feeling that AI may represent an attack upon a whole way of life.
What people think
Analysis details
9 comments analyzed. Analyzed .
The sample selects active discussion branches and includes available parent comments. It may omit parts of the full thread. Selected themes do not measure community opinion or how common a view is.
Compute access constrains AI-driven science
LLM-based scientific research depends on compute that few institutions or countries control, making access a practical limit on who can participate. A counterpoint expects memory and compute prices to fall 80-90% within 7-10 years, which would make current 14,000 token/s and 4TB machines look quaint; the timing and magnitude of that decline remain an estimate.
Enclosure and democratization claims lack data
The opposite prediction, that AI will open scientific exploration to more participants, is treated as equally reasonable, implying neither claim can be adjudicated without more evidence. The argument is not that one side is wrong, but that both are premature.
The unowned commons is already enclosed
Academic publishing is described as an oligopoly of five for-profit corporations that collectively publish roughly half of all scientific articles, with access largely paid for by institutions. Copyright enforcement against sites such as Anna's Archive shows sharing can bring legal harassment, so the 'no one owns' premise may already be false before AI enters.
Funding politics complicate the AI-enclosure story
One view links the moment to severe NSF defunding and an OSTP assumption that future knowledge will be AI-generated, while China expands fundamental research. Another attributes the cuts to political targeting of diversity, gender and vaccine research, arguing AI is not the driver.
AI may expand knowledge rather than enclose it
A bullish counterargument holds that current training data is a grain of sand compared with future returns, and that AI will produce a science, knowledge and art hockey stick. This frames enclosure warnings as premature pessimism rather than a demonstrated outcome.