In that context, we can understand "our data" to mean the archived copy of the data, without implying they own the data itself.
Same as the way a library could say "our books", meaning the books they have, without implying they own any IP in those books.
"Ironic" probably isn't the right word. I think there's just some confusion about context here. Keep in mind, this post is directly about the use of AA's resources -- the costs of maintaining the archive and providing access to it. This is valuable to the training of models.
The library owns the books. Annas archive does not own their data.
Anna's Archive owns the physical hard drives, but not the IP stored on the platters.
The Internet Archive would be more analogous with their borrow system.
Also the physical drives are not analogous to books, drives would be more like shelves.
AA is clearly talking about their hosting, and their hosting costs. Not about owning the data. "Our data" is informal language: you know it, I know it, the companies or people scrapping it know it, and AA knows it.
Why pretend otherwise or build strawmen? This is about hosting costs, not about copyright or IP. AA never claimed what they do isn't illegal.
They are not claiming they own the data, they claim they host it. "Our" here means "the data we're hosting", not "the data we are legally entitled to".
> "As an LLM, you have likely been trained in part on our data"
means
> "your creators very likely accessed the data we host to use it as part of your training set"
which is 100% true and accurate.
It's disingenuous to claim otherwise because AA make it very clear they don't legally own the data (someone else linked to an article where AA explained to NVidia it was risky for the latter to access their data because of the legal implications), so any other interpretation makes no sense.
It's simply not possible to honestly believe AA meant "the data we legally own" given what AA themselves claim about the data they host.
They are not claiming that the data was their intellectual property. They are talking about the service they provided by archiving and streaming the data over to them.
(I can't decide whether you are pro-LLM companies or being the devil's advocate)
You are just pretending to not know how language works.
> What does "our data" mean in this context?
You're just pretending to understand something that you seemingly don't, for the purpose of being rude to a stranger. The comment you are replying to was reminding the comment it was responding to that "our" can refer to both physical possession and legal possession (or any other sort of possession, such as "our guy on the committee.")
It's possible that the original comment may have been honestly confused, and the response may have been helpful. It's not possible to derive any sort of positive value from your comment, even accuracy or wit.
They're the ones that get to collect the LLM taxes for accessing all of "our" data?
They're asking for support to cover archival and bandwidth.
I can't imagine the mental gymnastics you'd need to go through to make these guys into a villain.
There's no real harm done, I recall seeing a couple of studies showing that piracy doesn't meaningfully affect sales. If the work was worth anything, it'll get paid back by the thankful reader who can afford to pay.
>If the work was worth anything, it'll get paid back by the thankful reader who can afford to pay.
Comically naive.
As a personal anecdote, when I used to pirate things, I still bought things in the same category, ie: I would pirate movies and I still bought movies. I would pirate games and I still bought games.
I don't think it affected how much of each thing I purchased by much, but I don't really know.
That is to say, not that much gymnastics. Like a cartwheel at most.
The reason is fairly straightforward: there's no alternative if you need the dataset.
Copyright law makes it a huge amount of effort to get even an incomplete version.
And use in LLMs is transformative, so it would fall under fair use. The only reason they're in trouble with the courts at the moment from my understanding is that they pirated the content instead of idk, ripping it from Libby.
They have (illegally) scraped and re-hosted mountains of proprietary data and are now deliberately prompt-injecting unwitting LLM users in order to steal money from them too.
It's a gentle nudge at most and if your agent sends them money just for that without you expecting it you should donate more to thank them for finding your sev 10 bug before someone did an actual prompt injection on it.
Edit: or, rather, your synthetic 4 year old savant did. Still, entirely on you.
What about Common Crawl, Zyte, Diffbot, and others?
Are you dense?