Some context about this here: https://arstechnica.com/information-technology/2023/08/openai-details-how-to-keep-chatgpt-from-gobbling-up-website-data/
the robots.txt would be updated with this entry
User-agent: GPTBot
Disallow: /
Obviously this is meaningless against non-openai scrapers or anyone who just doesn’t give a shit.
deleted by creator
Idk , I am not all for a private entity scraping comments and then using it , thats basically us doing free labour for them! No cool !
Wouldn’t they theoretically be able to set up their own instance, federate with all the larger ones and scrape the data this way? Not sure if blocking them via the robots.txt file is the most effective barrier in case that they really want the data.
Robots.txt is more of an honor system. If they respect , they won’t do that trick.
Robots.txt is just a notice anyways. Your scraper could just ignore it, no workaround necessary.
I could be wrong but wouldn’t people be able to file class action lawsuits against these companies? because they are literally copying content without obtaining any prior explicit user consent, also I’m pretty sure Europeans have an upper hand with data privacy protection from GDPR (European data being extracted/harvested and transferred to US servers)
I could be wrong though
The possibility exists, though I wouldn’t hold my breath.
If they’ll pay us when they scrape our content, sure.
I think this is a general question and problem for the whole fediverse, and can easily lead to the question of whether, or even when the fediverse is going to embrace having closed or private spaces or even invite only spaces, in order to try to secure some “human interaction only” social media.
That won’t stop OpenAI. We need actual blocking, on the server side. Problem is, with federation and all, it will be really, really difficult to do. And expensive.
I can understand privacy concerns, but I feel like it’s inevitable that LLMs will be used to make lots of decisions, some possibly important, so wouldn’t you want some content included in its training? For instance, would you want an LLM to be ignorant of FOSS because all the FOSS sites blocked it, and then a child asks an LLM for advice on software and gets recommended Microsoft and Apple products only?
No
