Hacker News new | ask | show | jobs
by version_five 1071 days ago
Why is it automatically a bad thing if they use the data to train llms? Who is losing under this situation?
2 comments

> Why is it automatically a bad thing if they use the data to train llms? Who is losing under this situation?

At a glance of the post and only a few seconds thought, I would guess the 'catchers' - the people who are being paid to answer the questions.

If they are being brought to the platform with the promise of regular work, building up a profile of expertise & reliability so they can generate more revenue on the platform, but under the hood their responses are just training LLMs so their work can be automated and they're eventually replaced by a catcher-bot trained on their responses... then I would argue they are losing in this situation, as they've invested time building up their profile only to risk having it stripped away.

It's a bit like building a cool new physical product and then going to a lot of effort to build a shopfront and presence on Amazon, only to have an AmazonBasics version of your product appear after a couple of months.

edit: not to say that this is what is happening with this site or that their intent is malicious! Just gaming out a potential case where there are some losers on the user side here.

It's not that it's automatically a bad thing. It's that it's a thing many people will object to, so if that's what's being done, it should be disclosed.