Mind twist: AI leaders ask for regulation, 'which should give us pause,' and should Exxon pay for climate change?

Christy Strawser (0:00)

The day's biggest stories, the people behind the headlines, the conversations you won't hear anywhere else.

I'm Christy Strawser. You're listening to Audacy's On Deadline. There are plenty of adages about wolves guarding hen houses, and most agree it's not a good idea. So, why then are we turning to tech companies to protect us from AI overreach? Here's Newell Normand in New Orleans.

Newell Normand (0:23)

Joining us to talk about it is Andrew Schwartz, Professor of IT at the EJR, so College of Business at LSU.

Andrew, welcome to the show. This was kind of an interesting happenstance. One has to wonder, first off, if we're running violative of the FTC provisions, where you have these companies kind of colluding together, signing these accords together about how they're gonna move forward, and where does it go from here?

Andrew Schwartz (0:52)

Yeah, I guess a great question is a great way to begin a conversation. And really, this is an interesting time, because we started thinking about regulation. And I think the question at the heart of it is, who's gonna regulate it?

Who is gonna regulate it, and how are they gonna regulate it? Is this a global regulation? If we saw in the interaction with China and Xi about, are we gonna start to look at almost like a nuclear arms race around AI? Is it gonna happen at a federal level?

Obviously, the president has weighed in and decided to do it at more of a legislative approach, but is that the right approach or not? So I think we need to have those conversations about, where should we have some sort of regulation and where shouldn't we?

Newell Normand (1:31)

And what's driving this issue relative to these top tech leaders? The skeptic in me wants to say one of it is cost containment. It's gonna cost a lot of money to self-police this issue, right? I mean, you're talking about an incredibly robust system. There's gonna have to be a lot of investment, you know, as it relates to audit trail, this, that, and the list just goes on and on and on and on.

And then whether or not they trust themselves, right? Under the kind of the capitalistic structure, if my competitor's doing some things, that they have a competitive advantage, and maybe I'm recommending to my board not to do it, I'm not gonna live very long in that position that I have moving forward.

Andrew Schwartz (2:21)

Absolutely. Now, look, I'm not a political scientist person, so I can't really speak to the politics of this, and I'm not gonna try to. But what I can say is we do have to start to think about self-policing if the efficacy is going to be there. And you raised a couple of really good points about pricing, but we also think about competitiveness. So if we think about you as a business owner in self-policing, if you have a breakthrough, how are you gonna react if you're a chat GBT or Google? Are you gonna wanna be policed if you're really pushing that frontier model forward? Maybe, maybe not. And if we look at the history of the country, really, if we look at monopolistic patterns, if we look at breaking up companies, it's the federal government that had to step in to do that. It wasn't self-policing. So really, we have to question the efficacy of self-policing as a tactic to an industry that really is in need of and is asking for regulation. And let's keep that in mind. You have an industry actively asking for regulation. I don't know if you know, but this is kind of an unusual circumstance. I can't recall a similar circumstance in my lifetime when I've seen an industry raise their hand and say, please regulate us. That should cause us a little bit of pause, right?

Newell Normand (3:31)

Yes, it does. And that's why I keep asking these questions. But when I look at one of the first items that they talk about in this accord, and I'll just read it, implement robust internal controls to monitor capabilities and alignment of its models during training and deployment around areas like cybersecurity, biosecurity, chemical threats, and to ensure that its models do not hack or access technical systems in unintended ways.

Andrew Schwartz (4:03)

Let's pick up on some words there, let's look at the second to the last word that you just described. Unintended, right? Now, the question is, who is the one who it's being unintended for?

So we're assuming that we're all actors that are all operating with the same moral compass. You and I both know that that is not true. So maybe if I view the use of AI in a nefarious way, that is 100% how I intended to use it. So under that, this definition, the self-regulation is not going to work because I'm intending to use it in a nefarious way.