4 ms·

I'm in two minds with these kinds of fines (maybe not super relevant for this specific fine). 1. Do these big megacorps just not care about rules, and they loo

I'm in two minds with these kinds of fines (maybe not super relevant for this specific fine). 1. Do these big megacorps just not care about rules, and they look at fines as a cost of doing business while the steam roll through industries. 2. Have lawmakers completely failed for the past decades to effectively referee tech, and so instead of having sane regulation and clear rules everything has to get decided in courts with fines. Probably both are true, and the real losers are small businesses trying to get started.

m00x 3y ago
The laws around training an LLM on publicly accessible IP have never been developed and haven't been developed, so the judge has to decide from common law. It's the same issue as crypto <> SEC. Companies ask lawmakers to set the rules, but they drag their feet, they release their products, and end up getting sued. It has been decided already in more technically knowledgeable countries that training on data is not the same as copying the data. It's obvious if you understand the concept of data compression, but judges don't always see it that way.
- anon84873628 3y ago
  Or that we already train humans on copyrighted data and then employ that human intelligence for commercial use. You can argue that AI is a categorically different scale that requires special regulation to protect our societal values (and I think I do) but that is a different claim and implementation than copyright enforcement.
  - petre 3y ago
    It should be viewed as derivarive work. It equally applies to humans. It shouldn't be classified as fair use because the works have been used in whole to train the model. 'A "derivative work" is a work based upon one or more preexisting works, such as a translation, musical arrangement, dramatization, fictionalization, motion picture version, sound recording, art reproduction, abridgment, condensation, or any other form in which a work may be recast, transformed, or adapted.' (US Copyright Act of 1976)
    - hnfong 3y ago
      You’re saying after a student learns physics from a copyrighted textbook they have to pay the textbook publisher royalties to work as a physicist?

m00x 3y ago

The laws around training an LLM on publicly accessible IP have never been developed and haven't been developed, so the judge has to decide from common law. It's the same issue as crypto <> SEC. Companies ask lawmakers to set the rules, but they drag their feet, they release their products, and end up getting sued. It has been decided already in more technically knowledgeable countries that training on data is not the same as copying the data. It's obvious if you understand the concept of data compression, but judges don't always see it that way.

anon84873628 3y ago
Or that we already train humans on copyrighted data and then employ that human intelligence for commercial use. You can argue that AI is a categorically different scale that requires special regulation to protect our societal values (and I think I do) but that is a different claim and implementation than copyright enforcement.
- petre 3y ago
  It should be viewed as derivarive work. It equally applies to humans. It shouldn't be classified as fair use because the works have been used in whole to train the model. 'A "derivative work" is a work based upon one or more preexisting works, such as a translation, musical arrangement, dramatization, fictionalization, motion picture version, sound recording, art reproduction, abridgment, condensation, or any other form in which a work may be recast, transformed, or adapted.' (US Copyright Act of 1976)
  - hnfong 3y ago
    You’re saying after a student learns physics from a copyrighted textbook they have to pay the textbook publisher royalties to work as a physicist?