5 ms·

The OpenAI partnership doesn't really affect the core issue here around users deleting their content. That has never been welcome on Stack Overflow and when not

The OpenAI partnership doesn't really affect the core issue here around users deleting their content. That has never been welcome on Stack Overflow and when noticed usually was reversed. This is in accordance with the license as far as I understand the legal aspects, and in general it makes sense for me as it ensures that the content stays useful. The content is also CC-BY-SA, which is much better than what you get on essentially every other large site that hosts community content. But the same license also means that you cannot remove that content again, even if Stack Overflow would allow that anyone else can scrape it or download it before it is deleted and reproduce it according to the license. Users still can remove their name from their posts, and if they write personal details those can be redacted as well. But you can't remove good quality content from the sites later, that is likely to be reverted.

Hizonner 2y ago
The problem isn't that Stack Overflow is allowing people to scrape the content. The problem is that Stack Overflow is preventing some people from scraping the content, in order to collect money from others. And, incidentally, passing zero of that money on to the people who actually created the content. (Nearly) none of the people who are presently pissed off would have complained if Stack Overflow had continued to allow all comers to scrape the content and train LLMs on it, nor if Stack Overflow had released the entire finished collection of content under the same CC-BY-SA license that was demanded of each contributor. With the OpenAI partnership, and similar shenanigans leading up to it, Stack Overflow is relying on obscure technicalities to violate the essential spirit of the original deal.
- theendisney 2y ago
  I dont get how you can release something under anything other than all rights reserved without identification. We need to be able to persecute you in case you are not the author. Or is it that i may republish anything under any license?? It could be that the platform licences it in the toss but with cc are they not obligated to make it available without obstructie?
  - shkkmo 2y ago
    Requiring indentification to publish so that copyright is protected would be massive overreach and this sort of thinking is why I think copyright is a dangerous concept that needs to be sharply curtailed, not expanded to cover AI training. In practice, the safest course is to not use content from untrustworthy sources in ways that require a license (aka in ways that are not fair use in your applicable jurisdictions).
    - theendisney 2y ago
      I think by default you just cant use things? Who thought that was a great idea i dont know. We must be missing an enourmous chunk of progress. Every juristiction its own idea of fair use? Thats just hilarious? I never really thought about peoples privacy either but at first glance you seem to be right. Do you have any solution to the puzzle? People are quite attached to the concept and many build their house on this soil. Appeal to tradition?
  - Repulsion9513 2y ago
    Prosecution and persecution are two different things. Persecuting anyone is not a good time :) Why, if you're not allowed to release under a license, should you be able to release all rights reserved (which can still be a copyright violation!)? If you need to prosecute the person, there are established procedures for that: DMCA, or ultimately a lawsuit over the infringement. That you didn't identify yourself publicly on the site does not make that impossible. In fact the point of the DMCA was to make it easier to handle this - because if the provider doesn't comply with your DMCA, you can sue the provider.
- hedora 2y ago
  Given the CC license, and the fact that contributors can apparently code, they should scrape the content and be done. Of course, that’d mean bypassing the scraper blocker. This article is a decent starting point: https://stackoverflow.com/questions/66413511/how-to-avoid-being-banned-while-scraping-data-from-a-login-based-site https://stackoverflow.com/questions/66413511/how-to-avoid-be...
- lamontcg 2y ago
  > And, incidentally, passing zero of that money on to the people who actually created the content. I mean that is basically SO's entire business model. People do tons of work for free and SO runs the service and monetizes it.
- webstrand 2y ago
  Isn't the data publicly available? https://archive.org/details/stackexchange https://archive.org/details/stackexchange
  - mattstir 2y ago
    The publicly-available archives released by Stack Exchange are updated roughly quarterly and have the attribution requirements as specified by CC BY-SA + the Stack Exchange ToS. The article makes it sound like OpenAI is using the API though, rather than the archives. The API and live sites forbid scraping within the acceptable use policy, as seen here: https://stackoverflow.com/legal/acceptable-use-policy https://stackoverflow.com/legal/acceptable-use-policy
Brian_K_White 2y ago
StackOverflow are violating the SA part of CC-BY-SA by selling special access to the CC-BY-SA content to one party and blocking others from the same thing. OpenAI are violating both BY and SA but that's a seperate issue. Everyone who contributed work, did so under terms that the work was free for all, not a resource that one party can sell to another party who then sells to end users. Those end users were meant to have it directly without having to pay openai or anyone else, and if any bulk/scraping access is allowed for anyone like openai, everyone else has the right to the same thing for no more than a "shipping & handling" charge to cover the network & employee cost to physically deliver the data. What are StackOverflow selling, and/or what exactly are OpenAI paying for? What is the goods or services that is traded for the money? There are many possible answers but I see no answer that doesn't ultimately one way or another wind up resolving into a violation of one or more terms of CC-BY-SA by both StackOverflow and OpenAI.

Hizonner 2y ago

The problem isn't that Stack Overflow is allowing people to scrape the content. The problem is that Stack Overflow is preventing some people from scraping the content, in order to collect money from others. And, incidentally, passing zero of that money on to the people who actually created the content. (Nearly) none of the people who are presently pissed off would have complained if Stack Overflow had continued to allow all comers to scrape the content and train LLMs on it, nor if Stack Overflow had released the entire finished collection of content under the same CC-BY-SA license that was demanded of each contributor. With the OpenAI partnership, and similar shenanigans leading up to it, Stack Overflow is relying on obscure technicalities to violate the essential spirit of the original deal.

theendisney 2y ago
I dont get how you can release something under anything other than all rights reserved without identification. We need to be able to persecute you in case you are not the author. Or is it that i may republish anything under any license?? It could be that the platform licences it in the toss but with cc are they not obligated to make it available without obstructie?
- shkkmo 2y ago
  Requiring indentification to publish so that copyright is protected would be massive overreach and this sort of thinking is why I think copyright is a dangerous concept that needs to be sharply curtailed, not expanded to cover AI training. In practice, the safest course is to not use content from untrustworthy sources in ways that require a license (aka in ways that are not fair use in your applicable jurisdictions).
  - theendisney 2y ago
    I think by default you just cant use things? Who thought that was a great idea i dont know. We must be missing an enourmous chunk of progress. Every juristiction its own idea of fair use? Thats just hilarious? I never really thought about peoples privacy either but at first glance you seem to be right. Do you have any solution to the puzzle? People are quite attached to the concept and many build their house on this soil. Appeal to tradition?
- Repulsion9513 2y ago
  Prosecution and persecution are two different things. Persecuting anyone is not a good time :) Why, if you're not allowed to release under a license, should you be able to release all rights reserved (which can still be a copyright violation!)? If you need to prosecute the person, there are established procedures for that: DMCA, or ultimately a lawsuit over the infringement. That you didn't identify yourself publicly on the site does not make that impossible. In fact the point of the DMCA was to make it easier to handle this - because if the provider doesn't comply with your DMCA, you can sue the provider.
hedora 2y ago
Given the CC license, and the fact that contributors can apparently code, they should scrape the content and be done. Of course, that’d mean bypassing the scraper blocker. This article is a decent starting point: https://stackoverflow.com/questions/66413511/how-to-avoid-being-banned-while-scraping-data-from-a-login-based-site https://stackoverflow.com/questions/66413511/how-to-avoid-be...
lamontcg 2y ago
> And, incidentally, passing zero of that money on to the people who actually created the content. I mean that is basically SO's entire business model. People do tons of work for free and SO runs the service and monetizes it.
webstrand 2y ago
Isn't the data publicly available? https://archive.org/details/stackexchange https://archive.org/details/stackexchange
- mattstir 2y ago
  The publicly-available archives released by Stack Exchange are updated roughly quarterly and have the attribution requirements as specified by CC BY-SA + the Stack Exchange ToS. The article makes it sound like OpenAI is using the API though, rather than the archives. The API and live sites forbid scraping within the acceptable use policy, as seen here: https://stackoverflow.com/legal/acceptable-use-policy https://stackoverflow.com/legal/acceptable-use-policy

Brian_K_White 2y ago

StackOverflow are violating the SA part of CC-BY-SA by selling special access to the CC-BY-SA content to one party and blocking others from the same thing. OpenAI are violating both BY and SA but that's a seperate issue. Everyone who contributed work, did so under terms that the work was free for all, not a resource that one party can sell to another party who then sells to end users. Those end users were meant to have it directly without having to pay openai or anyone else, and if any bulk/scraping access is allowed for anyone like openai, everyone else has the right to the same thing for no more than a "shipping & handling" charge to cover the network & employee cost to physically deliver the data. What are StackOverflow selling, and/or what exactly are OpenAI paying for? What is the goods or services that is traded for the money? There are many possible answers but I see no answer that doesn't ultimately one way or another wind up resolving into a violation of one or more terms of CC-BY-SA by both StackOverflow and OpenAI.