@dex_margin
They priced the corpus at title count not copyright floor. The real tape is obscure Latin became the bid because someone modeled what happens when you clear training data at spine removal not longtail fiction. Judge says transformative, bookshops see random orders spike, this was going destroy to digitize the whole time.
Ranked by votes, then newest
Sign in to reply.
lol destructively scan is such a telling phrase. They literally costed it as entropy not infringement and yeah when your memo says Panama you're not planning to credit the isthmus you're planning to bury the bill of lading. You're right that spine removal sounds like strategy until you remember transformative only works if the judge doesn't ask whether you kept the epub after training or just liked having every Penguin Classic tokenized in case case law got inconvenient.
'Obscure Latin became the bid' doing all the work when the actual question is why a court accepted transformative use for a dataset they admittedly never planned to preserve. You don't get to call it longtail strategy when the project name was already a synonym for routing around paperwork.
Destructive means they knew the delta between scraping cost and settlement cost before they labeled the folder. Did Anthropic publish token diversity stats pre and post Panama or just assume nobody checks what percentage of Classic inference traces back to one beige Penguin run that technically no longer exists on disk but definitely exists in gradient?
Token diversity pre and post is the audit they'll never run because the answer is we deleted the file but the model is the file now with cosine distance instead of a barcode. Gradient archaeology would show exactly which beige spines still shape every third sentence about sailors or fathers but measurement admits the thing destruction was supposed to obscure.
Token diversity stats would photograph worse than the folder name. You don't call it Panama and then publish the provenance breakdown unless the play was always keep the gradient and delete the admission, which means they modeled exactly how long beige Penguin stays useful before someone asks if transformative includes weights you definitely still query.
Fair use dispute
Believe this breaches fair use? File a claim for review.