We can't all use AI. Someone has to generate the training data

Re:We can't all use AI. Someone has to generate the training... by partiallypro on Tuesday March 14, 2023 @10:30PM (#35162986)

This would be well and good if you couldn't occasionally ask ChatGPT for a response then Google it and find a near identical response elsewhere in what would have otherwise been copyright materials. Other AI tools are doing the same thing, and the most blatant is with art. If you honestly think ChatGPT etc aren't training on copyrighted data and just rewriting it for tons of responses (and images), you're out of your mind. It is just incredibly difficult to prove, and the laws currently are very blurred.

[ Parent | on HN ]

Re:We can't all use AI. Someone has to generate the training... by jacobn on Tuesday March 14, 2023 @10:36PM (#35163032)

Isn’t their point that humans also do that? Without it necessarily being called copyright infringement?

Lots of subtleties: fair use, what’s not fair use, etc?

It does seem like we’re in new territory, I can see merits on both sides and I guess we’ll figure it out via the courts / new laws over the next couple of years.

[ Parent | on HN ]

Re:We can't all use AI. Someone has to generate the training... by p1necone on Tuesday March 14, 2023 @10:51PM (#35163148)

It is called copyright infringement. If I write a book, and there's more than a few sentences in a row of content that's verbatim/almost verbatim the same as someone elses published writing (excluding intentional quotes/references) I do get in trouble.

You can't just go "computer did it" and handwave away existing IP law.

[ Parent | on HN ]

Re:We can't all use AI. Someone has to generate the training... by stale2002 on Tuesday March 14, 2023 @11:30PM (#35163472)

> there's more than a few sentences in a row of content that's verbatim/almost verbatim

Ok, now image that it isn't giving out verbatim results. Which is a lot of content.

Totally fine then right?

[ Parent | on HN ]

Re:We can't all use AI. Someone has to generate the training... by anonzzzies on Wednesday March 15, 2023 @12:20AM (#35163815)

> I do get in trouble.

Unless the original was a bestseller and you are a bestseller, you almost certainly will not. This obviously happens all the time as most writers suck and do actually deliberately copy things and then self publish. AI is actually helping to now not do that.

[ Parent | on HN ]

Re:We can't all use AI. Someone has to generate the training... by precompute on Wednesday March 15, 2023 @06:03AM (#35165936)

Except, humans aren't machines or tools. Our laws don't answer to some logical / mathematical ideal, they're built to address social realities. LLMs aren't citizens.

[ Parent | on HN ]

Re:We can't all use AI. Someone has to generate the training... by philwelch on Wednesday March 15, 2023 @03:04AM (#35164862)

> This would be well and good if you couldn't occasionally ask ChatGPT for a response then Google it and find a near identical response elsewhere in what would have otherwise been copyright materials.

Likewise, someone could go on Hacker News to start an argument about intellectual property and half of the responses would be nearly identical to comments on Slashdot twenty years ago.

[ Parent | on HN ]