3 ms·

You are equivocating when you said “A human brain also trains on copyrighted work.” You’re stating foolishly, not plainly. Only an author, owner, or agent copyr

You are equivocating when you said “A human brain also trains on copyrighted work.” You’re stating foolishly, not plainly. Only an author, owner, or agent copyright something. LLMs are none of those. A person who uses an LLM to scrape art isn’t making art so they’re neither of the three either.

zelphirkalt 2y ago
Then the person using an LLM to produce something that infringes on cooyright is responsible. They better check for every produced output, whether it infringes, since they are not creating it by thinking themselves and coming to a solution, but are using a tool that stochastically regurgitates bits and pieces it has seen before. Not much credit is deserved.
- slowhadoken 2y ago
  Both OpenAI and the user are infringing on the artist and neither are making art.
  - zelphirkalt 2y ago
    It is about time, that we get a solid framework for handling those and other cases. I would also argue, that models learned from publicly available data need to stay public as well.
    - slowhadoken 2y ago
      Yup, the fake conversation about LLMs being A.I. is a distraction from a more important conversation about public knowledge and information. OpenAI is not a proper education tool but it presents its presented that way. It also a platform form propaganda and censorship.
aeonik 2y ago
I really don't think it's an equivocation. Training is a high level word, and because we engineered LLMs we know more specifics about how their training works, so when used in context with LLM, it can mean something more specific. I don't know the full mechanics of how training/learning works in the human brain, but it's I've been studying casually for 30 years. It's possible that parts of our brain work very similarly. What I'm trying to say is, regardless of underlying mechanics, I don't understand why using copyrighted works for training (anything) is a against some legal or moral framework.
- slowhadoken 2y ago
  You don’t think you are but you are. There’s a connection between your poor reasoning and your belief that “A.I.” is analogous to human thinking. You don’t understand LLMs or human cognition, therefore you have an ignorant opinion.
  - aeonik 2y ago
    @slowhadoken, I do not "believe" AI is analogous to human thinking. Please read my previous messages more carefully. -> I don't know how human thinking works, and according to my research, nobody does (not yet at least). It might be exactly the same, or it might be similar, and it might not. This is all irrelevant to what I am saying. I am saying, it doesn't make any difference according to my understanding of existing law. I also don't understand what moral issues there are here that are specific to LLMs, and apply only to LLMs. When I say that LLMs train, and humans train, I mean from a systems perspective, they both use the input of the data for similar purposes. I ask the question: Why is one different than the other? Why are humans allowed to analyze and learn from art, but an LLM isn't? If they are different, and not analogous, that's one possiblity... but still doesn't answer my question: why does this unspecified and unknown "difference" matter?
    - slowhadoken 2y ago
      The human mind and an LLM are not the same, that’s an ignorant thing too imply. Your questions are intensely naive too. Also your logic is weak. Just because scientists don’t know exactly how the human brain “works” doesn’t mean we don’t have accurate tests for general intelligence. On the other hand we know exactly how an LLM works and it’s not in the same ballpark with human intelligence, it’s not even the same game. It’s like thinking a calculator has a soul because it can “do” math. It’s childish fiction.

zelphirkalt 2y ago

Then the person using an LLM to produce something that infringes on cooyright is responsible. They better check for every produced output, whether it infringes, since they are not creating it by thinking themselves and coming to a solution, but are using a tool that stochastically regurgitates bits and pieces it has seen before. Not much credit is deserved.

slowhadoken 2y ago
Both OpenAI and the user are infringing on the artist and neither are making art.
- zelphirkalt 2y ago
  It is about time, that we get a solid framework for handling those and other cases. I would also argue, that models learned from publicly available data need to stay public as well.
  - slowhadoken 2y ago
    Yup, the fake conversation about LLMs being A.I. is a distraction from a more important conversation about public knowledge and information. OpenAI is not a proper education tool but it presents its presented that way. It also a platform form propaganda and censorship.

aeonik 2y ago

I really don't think it's an equivocation. Training is a high level word, and because we engineered LLMs we know more specifics about how their training works, so when used in context with LLM, it can mean something more specific. I don't know the full mechanics of how training/learning works in the human brain, but it's I've been studying casually for 30 years. It's possible that parts of our brain work very similarly. What I'm trying to say is, regardless of underlying mechanics, I don't understand why using copyrighted works for training (anything) is a against some legal or moral framework.

slowhadoken 2y ago
You don’t think you are but you are. There’s a connection between your poor reasoning and your belief that “A.I.” is analogous to human thinking. You don’t understand LLMs or human cognition, therefore you have an ignorant opinion.
- aeonik 2y ago
  @slowhadoken, I do not "believe" AI is analogous to human thinking. Please read my previous messages more carefully. -> I don't know how human thinking works, and according to my research, nobody does (not yet at least). It might be exactly the same, or it might be similar, and it might not. This is all irrelevant to what I am saying. I am saying, it doesn't make any difference according to my understanding of existing law. I also don't understand what moral issues there are here that are specific to LLMs, and apply only to LLMs. When I say that LLMs train, and humans train, I mean from a systems perspective, they both use the input of the data for similar purposes. I ask the question: Why is one different than the other? Why are humans allowed to analyze and learn from art, but an LLM isn't? If they are different, and not analogous, that's one possiblity... but still doesn't answer my question: why does this unspecified and unknown "difference" matter?
  - slowhadoken 2y ago
    The human mind and an LLM are not the same, that’s an ignorant thing too imply. Your questions are intensely naive too. Also your logic is weak. Just because scientists don’t know exactly how the human brain “works” doesn’t mean we don’t have accurate tests for general intelligence. On the other hand we know exactly how an LLM works and it’s not in the same ballpark with human intelligence, it’s not even the same game. It’s like thinking a calculator has a soul because it can “do” math. It’s childish fiction.