3 ms·

What do people imagine can be done about it at this point? Offer a concrete suggestion. Any law or tax against this will give a huge advantage to other countr

What do people imagine can be done about it at this point? Offer a concrete suggestion. Any law or tax against this will give a huge advantage to other countries. It's already over, there's no going back to a world where this didn't happen. Let's just hope some good comes of it.

hgs3 5mo ago
How about requiring AI companies to pay creators for training rights? Alternatively, models trained on the commons must be owned by the commons. Right now these AI companies are trying to have it both ways: it’s The People’s Data for training on comrade but ownership is privatized.
- quantummagic 5mo ago
  Practically speaking, who is going to enforce such a regime? Do you really want to give Chinese companies such a huge competitive advantage, that they aren't subject to the same costs as western companies? How do you even sort out which "creators" are owed, and how much? It's next to impossible, and would drown the legal system in litigation; it would likely cause more problems than it solves. On top of which you can find open weights for most, if not all, of the scraped material already. If you make those illegal to use, or prohibitively expensive, you just destroyed local LLM legality, and put the technology firmly in the hands of only the monopolists.
  - hgs3 5mo ago
    If models are trained on the collective whole, they must be owned by the collective whole. If you believe funding creators for the training of private models is too slow, inconvenient, or creates a global disadvantage, then embrace collective ownership.
    - quantummagic 5mo ago
      Sure, I wish everything was perfectly fair too. But how do you practically and REALISTICALLY proceed, and ensure you don't end up doing more damage than benefit? The road to hell is paved with good intentions. Everyone seems much more focused on complaining, and talking about "what's fair", than actually proposing concrete steps that would lead to a better world, without a significant risk of creating a worse one.
      - hgs3 5mo ago
        > concrete steps Start by legally compelling companies that trained on unlicensed data to either (1) license the data, (2) publish their model, or (3) destroy their model.
        quantummagic 5mo ago
        > Start by legally compelling companies that trained on unlicensed data to either (1) license the data, (2) publish their model, or (3) destroy their model. You are lost in an imaginary world where everything is simple and has no negative consequences. First off, there is NOBODY who has that power over all the companies in the world. So immediately you are creating an imbalance between companies and potentially destroying your domestic industry; with long term negative consequences for the people you're supposed to be protecting. Secondly, you might be creating a situation where it's impossible to ever create a competitor to those companies who are already entrenched monopolists, potentially even making it impossible to ever run self-trained or local LLM's. Also, you just unilaterally made it legal to publish all copyrighted work (since that's what you believe their model to be) to the general public, presumably in a way that can be used by everyone; further eroding copyright law in one fell swoop. You've completely disregarded the legal issues around what constitutes "unlicensed data", and how much is required before triggering your new law, and what that would mean for the legal system potentially being inundated. You're reacting way too emotionally and flippantly, with no apparent thought about what harm you are doing and how you might actually be making things worse, not better.
        hgs3 5mo ago
kouteiheika 5mo ago
> What do people imagine can be done about it at this point? Offer a concrete suggestion. Simple. Free the companies from copyright liability, but after X amount of time they are required to release everything into the commons. The weights, the training scripts and the full training data (appropriately processed so that it can only be used for training and not for people to easily pirate whatever works were used). They'd still get a monopoly on their model for a little bit to recoup their training costs, but in the end would be forced to give back what they took.
- quantummagic 5mo ago
  I'm sympathetic, since I think copyright laws are far too extensive and generous. But it's not simple, there are a lot of companies that won't fall under your jurisdiction, and the question is if that will give them a competitive advantage that kills the industry for you, and ultimately costs you more than you gain.

hgs3 5mo ago

How about requiring AI companies to pay creators for training rights? Alternatively, models trained on the commons must be owned by the commons. Right now these AI companies are trying to have it both ways: it’s The People’s Data for training on comrade but ownership is privatized.

quantummagic 5mo ago
Practically speaking, who is going to enforce such a regime? Do you really want to give Chinese companies such a huge competitive advantage, that they aren't subject to the same costs as western companies? How do you even sort out which "creators" are owed, and how much? It's next to impossible, and would drown the legal system in litigation; it would likely cause more problems than it solves. On top of which you can find open weights for most, if not all, of the scraped material already. If you make those illegal to use, or prohibitively expensive, you just destroyed local LLM legality, and put the technology firmly in the hands of only the monopolists.
- hgs3 5mo ago
  If models are trained on the collective whole, they must be owned by the collective whole. If you believe funding creators for the training of private models is too slow, inconvenient, or creates a global disadvantage, then embrace collective ownership.
  - quantummagic 5mo ago
    Sure, I wish everything was perfectly fair too. But how do you practically and REALISTICALLY proceed, and ensure you don't end up doing more damage than benefit? The road to hell is paved with good intentions. Everyone seems much more focused on complaining, and talking about "what's fair", than actually proposing concrete steps that would lead to a better world, without a significant risk of creating a worse one.
    - hgs3 5mo ago
      > concrete steps Start by legally compelling companies that trained on unlicensed data to either (1) license the data, (2) publish their model, or (3) destroy their model.
      - quantummagic 5mo ago
        > Start by legally compelling companies that trained on unlicensed data to either (1) license the data, (2) publish their model, or (3) destroy their model. You are lost in an imaginary world where everything is simple and has no negative consequences. First off, there is NOBODY who has that power over all the companies in the world. So immediately you are creating an imbalance between companies and potentially destroying your domestic industry; with long term negative consequences for the people you're supposed to be protecting. Secondly, you might be creating a situation where it's impossible to ever create a competitor to those companies who are already entrenched monopolists, potentially even making it impossible to ever run self-trained or local LLM's. Also, you just unilaterally made it legal to publish all copyrighted work (since that's what you believe their model to be) to the general public, presumably in a way that can be used by everyone; further eroding copyright law in one fell swoop. You've completely disregarded the legal issues around what constitutes "unlicensed data", and how much is required before triggering your new law, and what that would mean for the legal system potentially being inundated. You're reacting way too emotionally and flippantly, with no apparent thought about what harm you are doing and how you might actually be making things worse, not better.
        hgs3 5mo ago

kouteiheika 5mo ago

> What do people imagine can be done about it at this point? Offer a concrete suggestion. Simple. Free the companies from copyright liability, but after X amount of time they are required to release everything into the commons. The weights, the training scripts and the full training data (appropriately processed so that it can only be used for training and not for people to easily pirate whatever works were used). They'd still get a monopoly on their model for a little bit to recoup their training costs, but in the end would be forced to give back what they took.

quantummagic 5mo ago
I'm sympathetic, since I think copyright laws are far too extensive and generous. But it's not simple, there are a lot of companies that won't fall under your jurisdiction, and the question is if that will give them a competitive advantage that kills the industry for you, and ultimately costs you more than you gain.