The New York Times sued OpenAI and Microsoft for copyright infringement, a new front in the debate over the use of published work to train AI. https://t.co/u8qZ247dCl

— The New York Times (@nytimes) December 27, 2023

Allegation is that ChatGPT is "trained" with New York Times paid articles and copies those articles.


Will be interesting to see how this develops.

I think I am with the NYT on this one... please don't train our future overlords on that garbage...

Interesting. A few thoughts:

• Were the articles that ChatGPT used considered to be in the public realm? Available to anyone to consume?
• Did they pay for a subscription to the NYT for an instance of ChatGPT to use?
• If a person learns something from a bunch of articles and then profits from it, or uses it to make a political argument, do they sue that person for the use of the information they consumed from the NYT?

Great! So now Chatgpt learned how to lie. yo

the first two don't really matter, because that doesn't give one a license to copy.

it would be akin to hiring some writers to write a sci-fi movie, having them watch empoire strikes back as training for what type of movie you wanted, then the writers wrote a story about han solo and luke skywalker fighting darth vader and the empire on a frozen planet called hoth, using the exact dialogue from the movie.

• 
  

  Quote:

  ---

  If a person learns something from a bunch of articles and then profits from it, or uses it to make a political argument, do they sue that person for the use of the information they consumed from the NYT?

  ---

not sure what you are asking here, but they aren't suing NYT for just "using" the information. they are suing them for copying it.

How did ChatGPT copy it? Is copying without subsequent distribution itself a violation of IP laws? I'm a lawyer also but woefully ignorant on not only IP laws, but exactly what ChatGPT was doing with the materials it was using to train its learning module.

BMX Bandit said:

---

the first two don't really matter, because that doesn't give one a license to copy.

it would be akin to hiring some writers to write a sci-fi movie, having them watch empoire strikes back as training for what type of movie you wanted, then the writers wrote a story about han solo and luke skywalker fighting darth vader and the empire on a frozen planet called hoth, using the exact dialogue from the movie.

• 
  

  Quote:

  ---

  If a person learns something from a bunch of articles and then profits from it, or uses it to make a political argument, do they sue that person for the use of the information they consumed from the NYT?

  ---

not sure what you are asking here, but they aren't suing NYT for just "using" the information. they are suing them for copying it.

---

I just read the article in the WSJ about this. If they are indeed copying the articles, then they should sue their asses off. That would be the same as using a Getty image without paying for it...

I think it's going to be defined by what "copying" is. If they are storing the articles in a database that chatgpt simply retrieves and regurgitates I think that's copying. But is that what's happening here?

Liberal rag soon won't need humans to produce liberal rag

It's an interesting topic.

If ChatGPT is copying things outright it's problematic. If ChatGPT is using the information and generating new content based off what it has learned, is that any different from a person learning information and using it? Should it provide attribution from the sources it uses, and is that enough? How does this compare with a search engine returning results and showing the first few lines and linking to the source? Would that be good enough for ChatGPT, just quote a few lines and link the source article?

The fact that ChatGPT uses the NYT to train just diminishes the trust in chatgpt for me, not that there was much to begin with. Garbage in, garbage out, this just confirms that garbage is indeed part of the input.

Jabin said:

---

Is copying without subsequent distribution itself a violation of IP laws?

---

Yes.

kb2001 said:

---

It's an interesting topic.

If ChatGPT is copying things outright it's problematic. If ChatGPT is using the information and generating new content based off what it has learned, is that any different from a person learning information and using it? Should it provide attribution from the sources it uses, and is that enough? How does this compare with a search engine returning results and showing the first few lines and linking to the source? Would that be good enough for ChatGPT, just quote a few lines and link the source article?

The fact that ChatGPT uses the NYT to train just diminishes the trust in chatgpt for me, not that there was much to begin with. Garbage in, garbage out, this just confirms that garbage is indeed part of the input.

---

If you are using ChatGPT to simply retrieve and spit out answers or solutions then you are using it wrong.

I was thinking about this yesterday, after the recent citation and plagiarism controversy at Harvard: most of these current AI bots are information aggregators and after an analysis of the available data, they will often return a text response that is a copy of or close paraphrasing of the most common or popular or cited response it finds, but without attribution of links to what source it had chosen to use. They're often not really analyzing and synthesizing a unique response so much as doing an extremely fast and exhaustive search and comparative analysis of the quality or conformity of approval of possible answers it finds.

dmart90 said:

---

BMX Bandit said:

---

the first two don't really matter, because that doesn't give one a license to copy.

it would be akin to hiring some writers to write a sci-fi movie, having them watch empoire strikes back as training for what type of movie you wanted, then the writers wrote a story about han solo and luke skywalker fighting darth vader and the empire on a frozen planet called hoth, using the exact dialogue from the movie.

• 
  

  Quote:

  ---

  If a person learns something from a bunch of articles and then profits from it, or uses it to make a political argument, do they sue that person for the use of the information they consumed from the NYT?

  ---

not sure what you are asking here, but they aren't suing NYT for just "using" the information. they are suing them for copying it.

---

I just read the article in the WSJ about this. If they are indeed copying the articles, then they should sue their asses off. That would be the same as using a Getty image without paying for it...

---

They are not copying and publishing anything. Do you not understand how this AI learning works?

Also, the answers it gives are two years old. NYT should not have a copyright to a news article beyond a few months.

ChatGPT should simply give attribution like Wikipedia does.

C@LAg said:

---

Logos Stick said:

---

They are not copying and publishing anything. Do you not understand how this AI learning works?

Also, the answers it gives are two years old. NYT should not have a copyright to a news article beyond a few months.

ChatGPT should simply give attribution like Wikipedia does.

---

disagree. 100%

neither people nor AI should have unfettered "fair use" to copyright material.

it may be widely available, and even freely available (e.g. on NYT's web site), but that does not mean it is free to use.

---

They do it in Wikipedia all the time with attribution.

Two years is long enough to have the rights to some investigative report they did.

Maybe NYT should look to the Sports Illustrated model to improve their product.

C@LAg said:

---

Logos Stick said:

---

They are not copying and publishing anything. Do you not understand how this AI learning works?

Also, the answers it gives are two years old. NYT should not have a copyright to a news article beyond a few months.

ChatGPT should simply give attribution like Wikipedia does.

---

disagree. 100%

neither people nor AI should have unfettered "fair use" to copyright material.

it may be widely available, and even freely available (e.g. on NYT's web site), but that does not mean it is free to use.

---

You either begin using the information in an article as soon as you read it , or you discount it. Just like you do each post on this thread.

C@LAg said:

---

taxpreparer said:

---

C@LAg said:

---

Logos Stick said:

---

They are not copying and publishing anything. Do you not understand how this AI learning works?

Also, the answers it gives are two years old. NYT should not have a copyright to a news article beyond a few months.

ChatGPT should simply give attribution like Wikipedia does.

---

disagree. 100%

neither people nor AI should have unfettered "fair use" to copyright material.

it may be widely available, and even freely available (e.g. on NYT's web site), but that does not mean it is free to use.

---

You either begin using the information in an article as soon as you read it , or you discount it. Just like you do each post on this thread.

---

by your logic, I should be able to use any music lyrics I want to, or the music from a song I hear. just because I heard them they are fair game. I mean... I heard it over the air, so it must be free, right?

---

That would depend on HOW you use them. You hear a song and hum it to yourself, or sing it to your spouse, you are using it. It record it and sell it (without permission,) you are using it illegally.

Edit to add, we all use the information we find to make decisions all the time.

dmart90 said:

---

• If a person learns something from a bunch of articles and then profits from it, or uses it to make a political argument, do they sue that person for the use of the information they consumed from the NYT?

---

This should be where the argument begins and ends. ChatGPT doesn't copy and plagiarize like a Harvard president, it looks for word patterns within trillions of words sequences in order to create its own patterns in response to prompts. It is no different than a person reading material and then summarizing, rephrasing, or otherwise regurgitating it.

C@LAg said:

---

taxpreparer said:

---

C@LAg said:

---

taxpreparer said:

---

C@LAg said:

---

Logos Stick said:

---

They are not copying and publishing anything. Do you not understand how this AI learning works?

Also, the answers it gives are two years old. NYT should not have a copyright to a news article beyond a few months.

ChatGPT should simply give attribution like Wikipedia does.

---

disagree. 100%

neither people nor AI should have unfettered "fair use" to copyright material.

it may be widely available, and even freely available (e.g. on NYT's web site), but that does not mean it is free to use.

---

You either begin using the information in an article as soon as you read it , or you discount it. Just like you do each post on this thread.

---

by your logic, I should be able to use any music lyrics I want to, or the music from a song I hear. just because I heard them they are fair game. I mean... I heard it over the air, so it must be free, right?

---

That would depend on HOW you use them. You hear a song and hum it to yourself, or sing it to your spouse, you are using it. It record it and sell it (without permission,) you are using it illegally.

Edit to add, we all use the information we find to make decisions all the time.

---

this is recording it and using it. illegally. the content is copyrighted, even if readily available. what is so hard to understand about that.

ChatGPT is not being developed for free and the good of all..

It is monetized and will be monetized even further. They intend to sell off instances of the AI to be further developed/customized all over the place. and the data that was stolen to train the baseline AI will be used in ways no one can see, control or even necessarily predict.

it absolutely is an issue that needs to be hashed out legally now, rather than later.

---

No. This is closer to songs using the same chord progressions or different songs sounded similar because they use the same basic beats or styles.

You can ask ChatGPT to write a song about "whatever topic you want" in the style of another song. Like basically asking it to be weird al for you. So…to do that it has to know and use the lyrics of the original song you're asking about, for example. So is that a commercial use of the original lyrics being incorporated in their program or is it like a person having knowledge that they creatively manipulate to produce something new? That's basically the crux of this question.

ABATTBQ11 said:

---

C@LAg said:

---

taxpreparer said:

---

C@LAg said:

---

taxpreparer said:

---

C@LAg said:

---

Logos Stick said:

---

They are not copying and publishing anything. Do you not understand how this AI learning works?

Also, the answers it gives are two years old. NYT should not have a copyright to a news article beyond a few months.

ChatGPT should simply give attribution like Wikipedia does.

---

disagree. 100%

neither people nor AI should have unfettered "fair use" to copyright material.

it may be widely available, and even freely available (e.g. on NYT's web site), but that does not mean it is free to use.

---

You either begin using the information in an article as soon as you read it , or you discount it. Just like you do each post on this thread.

---

by your logic, I should be able to use any music lyrics I want to, or the music from a song I hear. just because I heard them they are fair game. I mean... I heard it over the air, so it must be free, right?

---

That would depend on HOW you use them. You hear a song and hum it to yourself, or sing it to your spouse, you are using it. It record it and sell it (without permission,) you are using it illegally.

Edit to add, we all use the information we find to make decisions all the time.

---

this is recording it and using it. illegally. the content is copyrighted, even if readily available. what is so hard to understand about that.

ChatGPT is not being developed for free and the good of all..

It is monetized and will be monetized even further. They intend to sell off instances of the AI to be further developed/customized all over the place. and the data that was stolen to train the baseline AI will be used in ways no one can see, control or even necessarily predict.

it absolutely is an issue that needs to be hashed out legally now, rather than later.

---

No. This is closer to songs using the same chord progressions or different songs sounded similar because they use the same basic beats or styles.

---

That's nonsense. For one, the gist of most music copyright cases focus on lyrics, not the music itself. There are chord progressions that transcend music types that are used often, but even chord progressions don't make up a song or piece of music. There's melody, rhythm, and cadences within common chord progressions. Beats or styles have little to do with song differences. As a drummer, I can tell you, I could survive as a studio/gigging drummer knowing (well) about a dozen different grooves, and three or four of those aren't used much in the US pop music scene. I could lay down rock groove tracks and resell those tracks (with some adjustments) dozens of times to artists needing a pro drummer rather than software loops for their productions. But each of those artists would have a wildly different song.

I don't think you can make a music comparison here, but your's is way off even if you could.

doubledog said:

---

If so, my respect for ChatGPT has reached a new low.

---

AI doesn't care about your respect. This is the issue.