Here's a company scraping millions of books, including millions in copyright, analyzing data privately, and using it for commercial gain. (Ngram viewer and Google Books don't have a fee, but are used in other products and add value to Google's ecosystem.) Compare to Prosecraft which made $0. - ThreadSky

tedunderwood.me • 679 days ago

Here's a company scraping millions of books, including millions in copyright, analyzing data privately, and using it for commercial gain. (Ngram viewer and Google Books don't have a fee, but are used in other products and add value to Google's ecosystem.) Compare to Prosecraft which made $0.

Comments

seanhp.bsky.social•679 days ago

Also, are we really going to pretend like corporations such as Google have the same legal treatment as everyone else. No way. As researchers or entrepreneurs, we should be aware that there are sharks in these waters and that we can’t afford a boat.

tedunderwood.me•679 days ago

This is very true, but I’d draw a different inference. Part of the problem with counting on IP law to protect individuals is that IP law is very responsive to deep pockets—in both directions. It’s going to protect corporations.

tedunderwood.me•679 days ago

E.g. legal exposure re: data is part of the reason open-source academic AI teams like EleutherAI (lacking a legal dept) can’t compete with OpenAI. The effect of tightening IP law will not to be to slow AI progress, imo, but to further concentrate control over it.

seanhp.bsky.social•679 days ago

I’m interested in what you’re saying. I wonder if this is an inference based on previous comparable rule changes, or if this is just something that sounds reasonable. If it’s the former, then indie hackers and researchers beware. If it’s the latter, then I’d love to be directed to evidence.

tedunderwood.me•679 days ago

If you can follow a link to Twitter, here’s Stella Biderman talking about the uneven effects of IP law right now on EleutherAI.

tedunderwood.me•679 days ago

That only shows that people who are trying to build AI in an open, accountable way feel frustrated that IP law binds them more than it binds big corporations. It's still possible to draw the conclusion "good — and now we need to make it bind the corporations." Viability of that open to debate.

tedunderwood.me•679 days ago

The Authors' Guild already attempted to sue Google over this, and lost. Researchers are pushing back right now because it's necessary to remind people what the law actually is. Transformative use is a factor in fair use; it can be legal to make statements *about* copyrighted text, even for profit.

robertmacanthony.bsky.social•679 days ago

Setting aside this current incident, the Google Books case relied in part on the fact that there was no market harm and (the court believed) a probably market benefit for the protected works. Cases involving AI products that compete with the underlying works could come out differently.

chloebubbles.gay•679 days ago

thanks ted for sharing some much needed elucidation on this debacle.

seanhp.bsky.social•679 days ago

I’m no legal scholar, but I’m aware of several data scraping/fair use cases against Google, OpenAI and Github that are ongoing and related to training LLMs. The law hasn’t caught up to this problem yet, but we’ll probably know soon, and many other non-AI projects will probably be similarly impacted.

Comments

Posting Rules

Reply