Live at the National Constitution Center
Sep 22, 2026 · 1 hr · 10 segments
In this episode, Technology and constitutional scholar Jonathan Zittrain and Distinguished Professor of Law at Rutgers Law School Ellen Goodman examine the intersection of the First Amendment and…
Jonathan ZittrainGuestEllen GoodmanGuest
Julie SilverbrookHostSo I think a lot of people are familiar with chatbots, right? That's probably how most people are interacting with AI, but then there is actually AI that can go speak, communicate, act on your behalf, some instances without human intervention.
So Jonathan, let's bring out the dogs to talk about the different types of AI.
Very good, so there's all sorts of types of AI, some of which are just madly categorizing things.
One of my favorite examples is Duchenne versus non-Duchenne smiles, which as psychologists will tell you are real smiles versus not so happy but I gotta smile smiles.
And you train them on enough and pretty soon you've got something you could run against your Facebook or Twitter feed and be like, they're not happy, they're happy, that kind of thing.
There's generative AI, which can generate new stuff, pictures, vocabulary, and such.
Then what you were talking about, Julie, is so-called, and I hate this adjective, but here we are, agentic AI, which just is the noun agent made into an adjective by adding ick to it.
on one axis, we have how powerful it is, which amounts ultimately to its risk potential.
On the other axis, for an agent that is capable of first being given a general goal And using its quote unquote reasoning power, the skeptics would be saying big quotes here, it makes up a plan and is off to the races, maybe just saying like, I'll be back at midnight, don't wait up.
It can go into the world and just do things often because it can pull the same levers humans can.
So for agentic AI, if it's really domesticated and not that powerful, it's like a Pomeranian.
And you know, what's not to like about Pomeranians? I was recently bitten by a Pomeranian-like dog, but we have, we don't teach it at most law schools, but there's a lot of dog law out there.
You know, there's a lot of cases like slip and fall, car accident, dog bite.
So I think a lot of people are familiar with chatbots, right? That's probably how most people are interacting with AI, but then there is actually AI that can go speak, communicate, act on your behalf, some instances without human intervention.
So Jonathan, let's bring out the dogs to talk about the different types of AI.
Very good, so there's all sorts of types of AI, some of which are just madly categorizing things.
One of my favorite examples is Duchenne versus non-Duchenne smiles, which as psychologists will tell you are real smiles versus not so happy but I gotta smile smiles.
And you train them on enough and pretty soon you've got something you could run against your Facebook or Twitter feed and be like, they're not happy, they're happy, that kind of thing.
There's generative AI, which can generate new stuff, pictures, vocabulary, and such.
Then what you were talking about, Julie, is so-called, and I hate this adjective, but here we are, agentic AI, which just is the noun agent made into an adjective by adding ick to it.
on one axis, we have how powerful it is, which amounts ultimately to its risk potential.
On the other axis, for an agent that is capable of first being given a general goal And using its quote unquote reasoning power, the skeptics would be saying big quotes here, it makes up a plan and is off to the races, maybe just saying like, I'll be back at midnight, don't wait up.
It can go into the world and just do things often because it can pull the same levers humans can.
So for agentic AI, if it's really domesticated and not that powerful, it's like a Pomeranian.
And you know, what's not to like about Pomeranians? I was recently bitten by a Pomeranian-like dog, but we have, we don't teach it at most law schools, but there's a lot of dog law out there.
You know, there's a lot of cases like slip and fall, car accident, dog bite.
The rest of this transcript — segmented and speaker-labeled, so you land on the exact moment something was said
Search every transcript — by keyword, by phrase, or by meaning, across every show Radar indexes
Trends — what is surging across podcasts, measured against its own baseline
Alerts — when a name you follow appears in a newly indexed episode
No account is needed to search Radar.
