4 ms·
A lot of people were only willing to contribute to StackOverflow because of the CC licensing, trusting the knowledge wouldn't be locked up. As a business that
by jefftk 3y ago
A lot of people were only willing to contribute to StackOverflow because of the CC licensing, trusting the knowledge wouldn't be locked up. As a business that depends on vast amounts of volunteer effort they need to balance providing a site where people are willing to contribute against making as much money as they can.
- urbandw311er 3y agoI wonder how many of those contributors, if re-consulted, would sign up to having their contributions used to train a for-profit LLM though? I certainly didn’t sweat it out helping people on SO to pay for Sam Altman’s fucking swimming pool.
- whatyesaid 3y agoI mean they would just scrape it if there's no data dump. It just makes it harder for the small guys. They probably scraped and are scraping HackerNews. Generative AI doesn't follow copyright or even explicit software licenses as we have seen in AI art with human signatures and Microsoft Copilot.
- blihp 3y agoThere was always the possibility of some sort of aggregator/other front end sitting on top of the SO data. We just didn't know exactly what a successful one would look like until relatively recently. I always limited how much I contributed based on that as likely outcome. Discontinuing the data dump is a much bigger deal to me and completely changes the value proposition of their various sites.
- jefftk 3y agoFor what it's worth, as someone who has put a lot of writing online, I'm not bothered by having my writing including in the training sets of these LLMs. I write because I want to share knowledge, and it isn't important whether people get the knowledge directly from me versus mediated by friends, LLMs, etc.