from none of that works Department
We already wrote about the biggest headline-grabbing moment yesterday with Elon Musk and Andrew Ross Sorkin’s Dealbook interview, but there’s another crazy Techdirt-related thing involving copyright and AI. There was one. As we have explained many times, copyright is the wrong tool to use to regulate AI, and its use can lead to bad outcomes.
But there was absolutely nothing in this part of the interview (from either side) that made any sense at all.
It starts with a completely ignorant and incorrect question from Aaron Ross Sorkin. He doesn’t seem ready for this at all.
Ars: So one of the things about training on data is the idea that you’re not going to train on data. So these things aren’t trained based on people’s copyrighted information. Historically. That was the concept.
Elon: Yes, that’s a big lie.
Ars: please say it again.
Elon: All of these AIs are trained on copyrighted data. clearly.
Ars: So when OpenAI says that, you think it’s a lie…None of them say they train on copyrighted data.
Elon: Yes, that’s a lie.
Ars: it’s a lie. Straight?
Elon:A straight lie.
So…there’s a lie there, but it’s Andrew Ross Sorkin who claims that no AI company trains on copyrighted data. Everyone recognizes that. They say doing so is fair use (because it is). So the entire premise of this argument is wrong. OpenAI admitted in court that, of course: Train with copyrighted material. It’s just that you believe fair use allows it (because it actually does).
So in a sense, Elon is right to push back on Sorkin’s argument here. But Musk appears to accept the false premise of Sorkin’s question, which is that AI companies claim they don’t train on copyrighted data, which is misleading. If Mr. Musk knew what he was doing, he would have told Mr. Sorkin that his premise was wrong and that no company would deny training using such materials.
From there, Sorkin goes on to make an even more confusing argument, saying that snippets of articles on ExTwitter are fair use, but because people may post entire articles, when combined they may not be much more. There is…but it claims…it’s not like this. It works anyway. This is embarrassing, someone please give him a lesson in Copyright 101.
Either way, the mask then made the whole thing…umm…weird as hell. Because Musk did this when Sorkin kept pressuring him about the copyright case.
Musk: I don’t know, except that by the time these cases are decided, digital gods will have been born. So at that point you can ask the digital gods. Hmm. These cases are never decided within a reasonable time frame.
If someone I know starts saying things like that, I’ll get them checked.
Whether or not you believe that AGI (artificial general intelligence) is on the horizon, or that it might create a “digital god,” this will be the case before these cases are decided. The idea of it happening is…well…unrealistic. But even if we somehow arrive at an AGI within the next few years as these cases progress, the idea that such an AGI will render courts and copyright law obsolete is similarly wishful thinking.
Heck, we’ve been claiming for years that the internet itself has already made copyright law obsolete, but copyright law is still here to stay, and is getting stupider all the time. I wish it were true that technology would make copyright laws even more obsolete and move things to a better overall system…but that won’t happen because of the “digital gods”.
Of course, if Elon really believes that a digital god will arrive in the next few years, which probably explains why he doesn’t care about ExTwitter advertisers anymore.
Filed Under: Andrew Ross Sorkin, Copyright, Digital Gods, Elon Musk, Fair Use
Companies: openai, twitter, x, xai