AI companies destroy physical books – let's scan rare books before it's too late
TL;DR Highlight
AI firms reportedly scan then destroy books; comments split over cost versus copyright.
Who Should Read
AI service developers curious about training data provenance and copyright issues
Core Mechanics
- No article text was provided
- Only comments were given, so the original claims cannot be verified
- No facts from the article are stated here
- What a quoted first paragraph says
- Several AI companies buy secondhand books via intermediaries, scan and destroy them
- The stated goal, as quoted, is training data from before 2022 "untouched by machines"
Community Reactions
- Opinion that destruction is about cost
- One commenter said non-destructive scanning can cost up to 10x more
- They noted Google never destroyed the books it scanned
- Suggestion to preserve only rare books
- One commenter said books with thousands of copies can be cut up
- They proposed non-destructive scanning for unique or limited editions
- Copyright blamed by some commenters
- Some said rights holders lock books up, forcing AI firms to shred them
- Some said copyright also blocks releasing the scans
- Some called the scale overblown
- One said most are specialized books that sat unused for decades
- They still found a single corporate-held copy problematic
How to Apply
- List the sources of your training and RAG data
- Open your corpus folder today and tabulate each document's source and license
- Flag documents with unknown provenance
- Separate unknown-license data from RAG
- Add a license field to metadata
- Add a retrieval filter that excludes unknown-license documents
Terminology
Related Papers
The Judge Can Debias Itself: Awareness and Self-Debiasing for Reliable LLM-as-a-Judge
Warning an LLM judge about bias or letting it self-critique improved 20 of 21 model–bias combinations.
90% of US working hours are spent on tasks not covered well by AI benchmarks
Mapping 29 benchmarks to job tasks shows 90% of US working hours lack good coverage
Postgres 19 delayed after AI reviewers find more bugs than humans can fix
LLM code review surfaced so many bugs that Postgres 19 slipped to November and 8-10 features were reverted
Show HN: Gemini 4 Argon plays StarCraft
LLMs wrote StarCraft bots in one hour; GPT-6 Astra scored highest.
A Plausible Explanation Is Enough to Mislead Multimodal Judges
Adding plausible explanations to fixed answers cut MLLM judges' wrong-answer rejection by 12.7–13.8 points.
Neural networks are clean
Looking at neurons directly shows most have one main job, challenging superposition.