137
Pay-per-output? AI firms blindsided by beefed up robots.txt instructions. (arstechnica.com)
submitted 1 year ago by to c/technology@lemmy.world
26 comments fedilink hide all child comments
you are viewing a single comment's thread
view the rest of the comments
[–] [S] 6 points 1 year ago (5 children)

I think the idea is that all parties would find it beneficial:

Leeds told Ars that the RSL standard doesn't just benefit publishers, though. It also solves a problem for AI companies, which have complained in litigation over AI scraping that there is no effective way to license content across the web.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 5 child comments
    •
  • [–] 15 points 1 year ago (4 children)

    The thing is a robots.txt file doesn't work as licensing. There's no legal requirement to fetch the file, and no mechanism to consent or track consent.

    This is putting up a sign that says everyone must pay, and then giving it to anyone who asks for free.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 4 child comments
    •
  • [–] [S] 2 points 1 year ago (3 children)

    The thing is if all parties find the terms agreeable it doesn’t matter if it’s legally binding.

    It’s more like putting a price on the shelf at the grocery store. Not every one will agree the price is agreeable and you might still get shoplifters but it doesn’t mean it’s a waste of time to list the price.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] 8 points 1 year ago (2 children)

    It really does matter if it's legally binding if you're talking about content licensing. That's the whole thing with a licensing agreement: it's a legal agreement.

    The store analogy isn't quite right. Leaving a store with something you haven't purchased with the consent of the store is explicitly illegal.
    With a website, it's more like if the "shoplifter" walked in, didn't request a price sheet, picked up what they wanted and went to the cashier who explicitly gave it to them without payment.

    The crux of the issue is that the website is still providing the information even if the requester never agreed or was even presented with the terms.
    If your site wants to make access to something conditional then it needs to actually enforce that restriction.

    It's why the current AI training situation is unlikely to be resolved without laws to address it explicitly.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 0 points 1 year ago (1 child)

    I think the analogy is apt. If you post a price on goods, and somebody walks into a store, picks up the item, and walks out without paying, they can't simply say, "Well, I didn't care to read the price, and nobody presented me with a contract, so I just took it," as a valid defense. There's sometimes an explicit agreement upon terms, sure, but there are times where that agreement is implicit: they put a price on a thing, I pay it, else it's stealing. I don't need to sign a contract every time I get groceries.

    I do, however, agree that this will only have teeth once it's argued and upheld in court the first (few) time(s). If nothing else, it's good to see people trying to solve the problem, rather than just throwing up their hands and letting billionaires run amok with virtual impunity. Maybe this won't work to reign in AI tech bros, but maybe it will inspire the things that do.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 2 points 1 year ago

    Except that with the website example it's not that they're ignoring the price or just walking out with the item. It's that the item was not labeled with a price, nor were they informed of the price. Then, rather than just walking out, they requested the item and it was delivered to them with no attempt to collect payment.

    The key part of a website is that the user cannot take something. The site has to give it to them.
    A more apt retail analogy might be you go to a website. You see a scooter you like, so you click "I want it!". The site then asks for your address and a few days later you get a scooter in the mail.
    That's not theft, it's a free scooter. If the site accused you of theft because you didn't navigate to an unlinked page they didn't tell you about to find the prices, or try to figure out payment before requesting, you'd rightly be pretty miffed.

    The shoplifting analogy doesn't work because it's not shoplifting if the vendor gives it to you knowingly and you never misrepresented the cost or tried to avoid paying. Additionally, taking someone's property without their permission is explicitly illegal, and we have a subcategory that explicitly spells out how retail fraud works and is illegal.

    Under our current system the way to prevent someone from having your thing without paying or meeting some other criteria first is to collect payment or check that criteria before giving it to them.

    To allow people to have things on their website freely available to humans but to prevent grabbing and using it for training will require a new law of some sort.

  • permalink
  • fedilink
  • source
  • parent