3 ms·

> But nobody was ever going to that Didn't Google have a long standing project to do just that? https://en.wikipedia.org/wiki/Google_Books https://en.wikipedi

by jayd16 1y ago

> But nobody was ever going to that Didn't Google have a long standing project to do just that? https://en.wikipedia.org/wiki/Google_Books https://en.wikipedia.org/wiki/Google_Books

efskap 1y ago
Crazy to think we've been helping train AI through captchas long before the "click all squares containing" ones.
- a2128 1y ago
  "stop spam. read books." is a very ironic phrase to look back on considering the amount of spam on the internet that LLMs have enabled
miohtama 1y ago
This lawsuit also makes sure that only parties that can train an AI with good enough training material are now - Google - Anthropic - Any Chinese company who do not care about copyright laws What is the cost of buying and scanning books? Copyright law needs to be fixed and its ridiculous hundred years tenure chopped away.
- slow_typist 1y ago
  Training a Model on 100+ years old literature only could be an interesting experience though.
  - rollcat 1y ago
    ’Twould wax yet more marvellous to ye beholders.
  - IshKebab 1y ago
    It's been done. https://github.com/haykgrigo3/TimeCapsuleLLM https://github.com/haykgrigo3/TimeCapsuleLLM
- godelski 1y ago
  From TFA > Anthropic also agreed to delete the pirated works it downloaded and stored. Also > As part of the settlement, Anthropic said that it did not use any pirated works to build A.I. technologies that were publicly released.
  - Iolaum 1y ago
    Reminds me when Facebook said to EU that they did not have the technology to merge FB and Whatsapp accounts when they bought Whatapp.
  - kelnos 1y ago
    That's not really the point, though, is it? Now Anthropic can afford to buy books and get them scanned. They likely didn't have the money or time to do that before. And even if they didn't use the illegally-obtained work to train any of the models they released, of course they used them to train unreleased prototypes and to make progress at improving their models and training methods. By engaging in illegal activity, they advanced their business faster and more cheaply than they otherwise would have been able to. With this settlement, other new AI companies will see it on the record that they could face penalties if they do this, and will have to go the slower, more expensive route -- if they can even afford to do so. It might not make it impossible, but it makes the moat around the current incumbents just that much wider.
  - DrillShopper 1y ago
    > As part of the settlement, Anthropic said that it did not use any pirated works to build A.I. technologies that were publicly released. Oh so now we're at "just trust me bro" levels of absurdity
godelski 1y ago
From TFA The Google Books project also faced a copyright lawsuit, which was eventually decided in favor of Google. After contacting major publishers about possibly licensing their books, [former head of the Google Books project] bought physical books in bulk from distributors and retailers, according to court documents. He then hired outside organizations to dissemble the books, scan them and create digital copies that could be used to train the company’s AI. technologies. Judge Alsup ruled that this approach was fair use under the law. But he also found the company’s previous approach — downloading and storing books from shadow libraries like Library Genesis and Pirate Library Mirror — was illegal.
- imwm 1y ago
  Disassemble*
- rchaud 1y ago
  That wasn't done as a play for venture capital. The Google Books project began before eBooks existed; in the 2000s, they spent money on all kinds of projects that had no real strategy for monetization. I remember Google Books being a valuable resource as it digitized books that were out of print. Back when they actually cared about making information available widely.
- Thorrez 1y ago
  Yeah. Weird that rchaud said "But nobody was ever going to that" when the article talks about someone doing it.

by jayd16 1y ago

> But nobody was ever going to that Didn't Google have a long standing project to do just that? https://en.wikipedia.org/wiki/Google_Books https://en.wikipedia.org/wiki/Google_Books

efskap 1y ago

Crazy to think we've been helping train AI through captchas long before the "click all squares containing" ones.

a2128 1y ago
"stop spam. read books." is a very ironic phrase to look back on considering the amount of spam on the internet that LLMs have enabled

miohtama 1y ago

This lawsuit also makes sure that only parties that can train an AI with good enough training material are now - Google - Anthropic - Any Chinese company who do not care about copyright laws What is the cost of buying and scanning books? Copyright law needs to be fixed and its ridiculous hundred years tenure chopped away.

slow_typist 1y ago
Training a Model on 100+ years old literature only could be an interesting experience though.
- rollcat 1y ago
  ’Twould wax yet more marvellous to ye beholders.
- IshKebab 1y ago
  It's been done. https://github.com/haykgrigo3/TimeCapsuleLLM https://github.com/haykgrigo3/TimeCapsuleLLM
godelski 1y ago
From TFA > Anthropic also agreed to delete the pirated works it downloaded and stored. Also > As part of the settlement, Anthropic said that it did not use any pirated works to build A.I. technologies that were publicly released.
- Iolaum 1y ago
  Reminds me when Facebook said to EU that they did not have the technology to merge FB and Whatsapp accounts when they bought Whatapp.
- kelnos 1y ago
  That's not really the point, though, is it? Now Anthropic can afford to buy books and get them scanned. They likely didn't have the money or time to do that before. And even if they didn't use the illegally-obtained work to train any of the models they released, of course they used them to train unreleased prototypes and to make progress at improving their models and training methods. By engaging in illegal activity, they advanced their business faster and more cheaply than they otherwise would have been able to. With this settlement, other new AI companies will see it on the record that they could face penalties if they do this, and will have to go the slower, more expensive route -- if they can even afford to do so. It might not make it impossible, but it makes the moat around the current incumbents just that much wider.
- DrillShopper 1y ago
  > As part of the settlement, Anthropic said that it did not use any pirated works to build A.I. technologies that were publicly released. Oh so now we're at "just trust me bro" levels of absurdity

godelski 1y ago

From TFA The Google Books project also faced a copyright lawsuit, which was eventually decided in favor of Google. After contacting major publishers about possibly licensing their books, [former head of the Google Books project] bought physical books in bulk from distributors and retailers, according to court documents. He then hired outside organizations to dissemble the books, scan them and create digital copies that could be used to train the company’s AI. technologies. Judge Alsup ruled that this approach was fair use under the law. But he also found the company’s previous approach — downloading and storing books from shadow libraries like Library Genesis and Pirate Library Mirror — was illegal.

imwm 1y ago
Disassemble*
rchaud 1y ago
That wasn't done as a play for venture capital. The Google Books project began before eBooks existed; in the 2000s, they spent money on all kinds of projects that had no real strategy for monetization. I remember Google Books being a valuable resource as it digitized books that were out of print. Back when they actually cared about making information available widely.
Thorrez 1y ago
Yeah. Weird that rchaud said "But nobody was ever going to that" when the article talks about someone doing it.