Welcome back to the podcast, bribe, swindle, or steel.

I'm Alexandra Rogge.

Today we're talking about artificial intelligence and what it means and doesn't mean for the compliance world.

There's been a lot of fanfare around AI as the answer to large volumes of compliance information

that can be difficult and expensive to sift through.

Certainly, technology is critically important to managing this volume,

but increasingly we hear from companies that the benefits of AI have been overstated and underdelivered.

My guest today is Merritt Smith.

Merritt is on the research team here at Trace with expertise in data science, public policy, and the intersection between those two fields.

Merritt, thank you for joining me. I'm overjoyed to be here, and I am surprised that I have not

been on one of these podcasts accidentally already because I make so much noise around the corner.

It is true that you sit there with just enormous amounts of data scrolling up your screen.

Why don't you start by helping us with some definitions?

What is the difference, for example, between AI, machine learning, deep learning, neural

networks, help us sort all that out?

These terms should be disambiguated and really put simply,

deep learning is a subcategory of machine learning, which is itself a subcategory of artificial

intelligence. And more deeply, artificial intelligence is this big, gnarly, overarching discipline

with tendrils into neuroscience, tendrils into psychology, philosophy,

computer science, mathematics, all of these different things that it draws on. But what it really

boils down to is the attempt to endow machines with cognitive abilities like humans. A lot of

folks have a Roomba. It's a fairly common household item.

...