How is liability assigned when something goes wrong with artificial intelligence (AI)? Who owns a copyright or patent developed by a machine? In this episode of “The AI/ML Show” on Motley Fool Live, recorded on Feb. 16, Fool.com contributors Toby Bordelon, Jose Najarro, and Jason Hall discuss how our current legal framework needs to get up to speed when it comes to AI.
Toby Bordelon: Who’s at fault when something goes wrong with artificial intelligence? It can’t be the artificial intelligence. That doesn’t make any sense. There’s got to be ultimately a person.
Jason Hall: It’s not a legal entity.
Hall: It’s owned by somebody.
Bordelon: I sue the AI, well, good luck with that. What are you going to get? Ultimately there’s got to be a person or a company behind that. But who is it? Is it the programmer? Is it a person who’s using it? Is the company who sold it? The obvious example, we have the self-driving cars. A self-driving car gets into an accident, who is liable for those damages? The driver? The user of the AI, even though they’re not, maybe — let’s fast-forward 20 years — they might not be actively using it? Or is it the company that created the AI? Is it the car manufacturer? But what if the manufacturer wasn’t a developer of the AI? Third parties are helping with this sort of thing. Who pays? Who’s at fault? And the way we’ve structured our system right now, you’ve got to find someone to assign fault. You can assign fault to multiple parties, but you got to assign fault. The best example of a system that tries to avoid that is insurance, where we’re going to assign fault but presumably the person has insurance so a whole bunch of people are bearing that cost because we just decided that’s the way we should operate our streets and our cars. Do we move to something like that with artificial intelligence? There’s going to be some society-level responsibility because we would prefer to have these things than not.
Hall: You’re not even talking about the nuance between criminal law and civil law, right?
Hall: Criminal law, you’ve got to have your judges and your attorneys in your prosecution. They need to know. But then in civil matters, so much more weight gets put on the jury, and the weights of evidence to rule one way or the other is very, laughs] very different. I’ve followed some of the things that have happened with copyright infringement, with music, with sampling over the past 10 years. I’m not sure if you guys have followed it. But I remember the “Ice Ice Baby,” the Vanilla Ice song that came out in the early 1990s that basically sounds like they stole “Under Pressure,” the entire bass line from “Under Pressure,” the Queen and David Bowie song. That song, and then more recent cases where the accuser, or person that claimed, you’re sampling my work, it was far less obvious and the rulings were completely opposite in both cases of what you would expect there. Again, that’s a civil matter where it’s a jury that’s making a judgment call on things. I think that’s one of the most important things for regulators and for legislators to get right, is making sure that that framework is there, so that when a suit happens, when a charge is filed, there’s something to keep it out of the ditches [laughs] because that’s what happens with this stuff. It gets way off the road, and the judge lets it go, and the civil trial and the jury rules. That’s our legal system. Maybe our robot overlords may need to take control of the legal system too, Toby.
Bordelon: Well, I wonder about that.
Hall: I think Jose is in support of that, right?
Bordelon: Yeah. [laughs]
Jose Najarro: I was going to say the only way to make these laws come out faster would probably be to use AI to make these laws work with the current technology. [laughs]
Bordelon: Theoretically, you could do that. You could have an artificial intelligence which is supposed to help Congress get to a certain place faster, right? Analyze this, give me a recommendation, and follow that recommendation into a one-page memo that I can understand, so I don’t have to read the however many thousands of pieces of data to use to analyze this. Do we need AI tools to help legislate? I think eventually we’ll get there, and maybe have some of this. When you look at things like intellectual property rights, a really interesting issue here. We’re quickly running out of time, but I just want to make this comment. Part of what we’re using, and what we want to use artificial intelligence for, is to create things like finding new drugs, right? I did an interview a couple of months ago where we talked about how AI, because they think differently, can find drug combinations or create them from the various molecular combinations that a human wouldn’t get to get to you as fast, if ever. When you make any drug, you file a patent. Who files the patent? In the U.S. system, some other countries are different, but in the U.S. system a corporation cannot file a patent. It must be an individual. Now if you work for a company, you’ve got a contractual obligation, you’re going to assign those rights to them if you’re doing it as part of an employee. But the point is that to file the patent, there’s got to be human to do it. But what if the AI finds that drug, who files the patent? That’s something we have to worry. It’s a minor issue, maybe. It’s a technical issue, but it’s just an example that the current framework, the current regulations, don’t work when the AI is involved. You got to figure something else out, and you can think of any number of things like this where you’re going to have to do something a little bit differently to when you have artificial intelligence creating things, writing music, making art, writing computer code, right? You’re going to have to figure something else out with the current system that incorporates the use of AI in these things. It’s going to be a while. But we certainly have hard work we need to do.