One of the things that frustrated me as a writer was when writers would then steal a photo for their blog and be shocked to discover that photos have copyrights and use, when, like of course they do. (Also, if you've been blogging as long as I have, you've watched this discovery occur a few times.)
But with this AI stealing all the text it can find, I realized I had had my own thing that I had missed. I blogged a little about the town hall with Pandora I went to way back when: http://www.talkapedia.com/2007/09/pandora.html?m=1 Interestingly I did not mention in my blog post that part of the reason the Pandora team was hanging out in DC was to convince Congress they shouldn't have to pay the same rates radio stations did to play songs. But looking back this was music streamers basically setting up a service that used music and sold music access to it's users, and then didn't want to pay for said music access.
And the AI/Large Language models need text written by actual humans in order to appropriately mimic human speech, but then don't want to pay for the very thing they need. I mean, they could of course write their own text. But that would take longer than stealing other people's text. So they don't want to do that. And then the AI people pretend that this roadblock suddenly appeared. As if the text they stole had just been lying out in the break room and they had no way to know it belonged to someone when they took it.
So now of course our music industry is impacted by the rules the streamers retroactively forced through. And I hope we won't see the same with LLMs. Not just because some of my text has been stolen. But also, I hope we can learn from our past mistakes.
