Nilay Patel

And that's just, I've been operating under that assumption for some time.

We need to stop here for another quick break.

We're back with Bridget McCormick, the head of the American Arbitration Association.

We were just discussing the AI arbitrator tool and the way she believes tools like it can actually increase trust in the judicial system at a time when trust in institutions continues to drop.

Now, I want to focus on the ways in which the system could go wrong through hallucinations, biased outcomes, and the many ways you could obviously see a tool like this having adverse effects.

And I wanted to ask Bridget how she and her team are trying to protect against all this, especially now as the AI arbitrator is starting to hear real cases.

So the flip side of this, and this is us covering AI at the Verge for years now, it'll just talk to you.

And some people are very happy with that.

And that is far more trust than they have in even the other people who live in their houses, right?

Like we see that play out all the time.

The downside of that is that these systems hallucinate at high rates.

that they are tuned to please you, and we can see that all over the place, that they work differently, they layer differently.

In fact, the chances of them getting something wrong kind of exponentially increase as you stack them up in these ways.

How have you protected against that here?

Because it feels like, yes, you can increase this sense of agency and trust, but the downside is this thing might just be making it up as it goes along.

I don't know that I have a lot of trust in public justice system.

What I know about those guys is that they get old and sometimes they go away.

I wish more of them would go away at higher rates, but sometimes they go away.

Sometimes they get replaced by newer, different people with different biases.

And the system replenishes at some rate that feels like accountability, right?