If openAI paid Wikipedia & others $x before charging their customers $20 or $200 per month, it would have been better. Having said that how OpenAI’s case is different from what Google is doing with their crawling & making money on Ads?
If openAI paid Wikipedia & others $x before charging their customers $20 or $200 per month, it would have been better. Having said that how OpenAI’s case is different from what Google is doing with their crawling & making money on Ads?
Google has already been through countless lawsuits because of its indexing, and what we see now is the sum total of all the wins/losses/settlements constantly happening in every jurisdiction in the world. The number of words they can show in the snippets, how they must respond to takedown requests, how they must share revenue for content, what they can and cannot cache, how personal info is removed upon request, how someone's house is blurred out, what content fee they must pay news publishers...everything is regulated. So "Google does it" is not at all an excuse that OpenAI or anyone else can use.
Presumably the difference is index (listing links) vs reproduction (actually returning content).
Also, it’s easier to remove copyright material if it’s not all crammed into an LLM first. Eg. If someone wanted to remove their website from Google, you can do that incrementally without rebuilding the entire index, whereas it’s a lot harder post-LLM (post-processing is probabilistic at best).
Exchange of value.
Websites tend to be okay with it because they accrue a benefit of Google’s crawling - they get traffic back. When websites don’t feel that Google keeps the traffic for themselves, websites tend to get upset https://www.theregister.com/2020/03/11/yelp_congress_google/
LLM training just takes and keeps all benefit to themselves. Wikipedia (or news site) get no traffic or anything back in return.
Not quite good enough I think, the copyright holders are not Wikipedia, but the individual contributors.