12
$\begingroup$

When we started contributing to CV we did not foresee the content would be used to train chat bots and that chat bots would be as profitable as they are.

Did we consent to the content to be used to train language models?

$\endgroup$
9
  • 10
  • 2
    $\begingroup$ Your concern was tangentially addressed here. $\endgroup$ Jan 8 at 18:09
  • 1
    $\begingroup$ thanks, these links are interesting. I don't see enough discussion of the idea that AI training is different kind of use that could not have been foreseen before, and that's why it has to be licenses explicitly, and can not be covered by fair use concept. $\endgroup$
    – Aksakal
    Jan 8 at 19:15
  • 1
    $\begingroup$ 1) Training sounds fine. There are a lot of humans as well that use StackExchange for their training. 2) A problem would be the case when entire posts or large parts of it are literally quoted without reference, or when original/recent/non-trivial ideas can not be traced back to their original source. Q) Which of these situations are you talking about in this question? If it is the first case, why do you think it is bad? If it is the second case, are there examples of language models actually doing this? $\endgroup$ Jan 8 at 21:08
  • 7
    $\begingroup$ It is not that I think the case #1 is bad, but I contend that human learning is fundamentally different from machine learning. Suppose, you share your family album with me and I learned something about your relatives. Now, imagine an ML algorithm is trained on the same album, now it can crawl the internet and find all your relatives everywhere in the world in every picture ever taken. $\endgroup$
    – Aksakal
    Jan 8 at 22:11
  • $\begingroup$ *ChatGPT $\endgroup$ Jan 18 at 22:24
  • $\begingroup$ @Aksakal it is not the algorithm that crawls the internet though $\endgroup$
    – Firebug
    Jan 19 at 12:39
  • $\begingroup$ Another thing is, did we have to consent? Is a model trained on data considered to be a derivative according to CC? $\endgroup$
    – Firebug
    Jan 19 at 12:39
  • $\begingroup$ @Firebug many people think we did not have to consent, but I think we did. $\endgroup$
    – Aksakal
    Jan 20 at 0:19

0

You must log in to answer this question.

Browse other questions tagged .