alt Hacker NewsTheft/stolen implies it's been taken and the owner deprived of it. Pirated or copied is more appropriate, but still not quite exact as it's been encoded into what amounts to lossy memory. If I ask a person to tell me about a paywalled article, they'll share what they can remember and likely get some facts wrong. An LLM is the same (these days; GPT-3 was another story).
I do agree that the methods of obtaining much of the material (eg. the published works torrents) were stupidly wrong and they should pay for that. However, legally acquired media being "read" by the training process isn't so different than it being read by a voracious reader: it's lossily encoded in (what amounts to) memories.
The insane scraping levels and behaviors ... not wrong. If only there were companies who already did that responsibly they could contract with for data. Hmm. (But they have to "own" it, as is the SV fashion these days.)
Replies
It was what's commonly called illegal piracy on a commercial basis and gigantic scale. That normally leads to hefty prison sentences. I've always been against this legal classification and think piracy should be allowed for non-commercial purposes or at least only be considered a very minor crime, but in the present example the piracy was clearly for commercial gains, a company using the illegally obtained material to create essential features of commercial products.
Agreements with publishers after the fact shouldn't even count, particularly since all copyright holders weren't consulted. I wouldn't have given my permission, for example.
It's important to stress that the media were not legally acquired by most US frontier AI companies if not all of them.
> Theft/stolen implies it's been taken and the owner deprived of it.
Nope! Theft of the proceeds.
The media may have been legally acquired, but it was done so by violating the intentions and purposes of the content creators.
In other words, people poured their time and attention into something hoping for a return, and got this, instead. And now this is obviating the venues by which creators hoped to receive their social/cultural dividend.
The AI companies doing their own scraping wouldn't be a problem if they'd to it in a reasonable way. But instead they hammer the same site over and over for data that hasn't been changed, or request the same data in different formats, or through interfaces that aren't efficient.