Japanese used bookstores see 5x sales surge as books are being bought by the ton
Used bookstores in Japan are reporting a fivefold surge in bulk orders, with tens of tons of titles shipped to U.S. firms that scan the pages for AI training and then destroy the physical books. Commenters debate whether this is an efficient way to preserve otherwise unsellable stock or a troubling form of cultural erasure that concentrates knowledge inside proprietary AI systems, especially when the resulting scans are not publicly accessible. Many also highlight the legal and ethical gray areas around copyright, fair use, and whether rare or out‑of‑print works should be pulped at all.
Scale of AI-Driven Book Buying
- Used bookstores in Japan report large bulk orders, 5x sales spikes, and exports of tens of tons of books to the US.
- Commenters infer these are mainly for AI training datasets, often non-fiction and reference works, including in underrepresented languages like Japanese.
Why Old Books Matter for AI
- Argued benefits: higher information density than random internet text, more careful thinking and writing, unique domain knowledge not online, and stylistic diversity.
- Some see books as “encapsulated cognition” and “memories of society” that models can learn from.
- Others doubt the marginal utility, noting diminishing returns and the existence of large modern corpora.
Destruction, Logistics, and What Used to Happen
- Books are often scanned destructively: spines cut off for faster, cleaner scanning; pages then shredded or pulped.
- Several note that unsold or weeded books were already being pulped or turned into products like toilet paper; scanning is seen by some as only marginally better.
- Limited demand, storage costs, and weight make broad donation impractical in many cases.
Legal and Copyright Dimensions
- Discussion centers on fair use and a recent court ruling that favored format-shifting purchased print books to digital, with destruction of the original strengthening the fair-use case.
- Distinction emphasized: training on copyrighted works has been found legal; redistributing full scans is generally not.
- Some incorrectly think destruction is legally required; others clarify it’s not mandatory but strongly incentivized.
Ethical and Cultural Concerns
- Many are disturbed by deliberate destruction of physical books, comparing it symbolically to historical book burnings or cultural erasure.
- Fears include a “dark age” where original sources vanish and only AI-mediated, possibly distorted knowledge remains.
- Others counter that national libraries (e.g., legal-deposit institutions) retain copies, and many of these commercial copies would have been pulped anyway.
Access, Openness, and Missed Opportunities
- Several wish the scans were made public or deposited with trusted institutions, framing this as a lost chance for a digital commons.
- Others note this is largely blocked by copyright, cost, and legal risk in identifying and clearing rights.
Competitive and Structural Implications
- Some argue proprietary scanned corpora will become a key moat for large AI firms, locking out future competitors and open-source efforts.
- A darker view asserts that destroying physical sources is strategically beneficial: it centralizes cultural access into AI services.
- Others are skeptical, suggesting many of these books are low-value or redundant, and that talk of deliberate cultural annihilation is overstated.