WEBVTT - Can AI Be Funny?

0:00:01.600 --> 0:00:04.640
<v Speaker 1>Hey, welcome to Sign Stuff, a production of iHeartRadio. I'm

0:00:04.640 --> 0:00:07.480
<v Speaker 1>hoh hitch Cham and today we're asking the question can

0:00:07.520 --> 0:00:13.039
<v Speaker 1>AI be funny? Sure you can ask chat GPT or

0:00:13.160 --> 0:00:15.680
<v Speaker 1>Claude to tell you a joke, but are you actually

0:00:15.760 --> 0:00:18.720
<v Speaker 1>going to laugh? We're gonna talk to an AI expert

0:00:18.720 --> 0:00:22.159
<v Speaker 1>who's been obsessed with deconstructing humor and teaching it to

0:00:22.200 --> 0:00:25.720
<v Speaker 1>a computer for the last fifteen years, and she's gonna

0:00:25.720 --> 0:00:28.400
<v Speaker 1>step us through the rules of what makes something funny,

0:00:28.720 --> 0:00:32.479
<v Speaker 1>whether AI can follow them, and whether comedians have their

0:00:32.560 --> 0:00:38.320
<v Speaker 1>days number. It's a pretty interesting conversation, and I have

0:00:38.360 --> 0:00:40.680
<v Speaker 1>to warn you there's a lot of laughing at it.

0:00:40.680 --> 0:00:44.360
<v Speaker 1>It turns out AI researchers can be pretty funny, So

0:00:44.440 --> 0:00:46.960
<v Speaker 1>get ready to hear the one about the computer scientist

0:00:47.000 --> 0:00:49.720
<v Speaker 1>walks into a bar with a cartoonist as we answer

0:00:49.800 --> 0:00:52.760
<v Speaker 1>the question can AI be funny?

0:00:54.800 --> 0:00:55.120
<v Speaker 2>Enjoy?

0:00:57.800 --> 0:00:59.760
<v Speaker 1>Hey everyone, So, I don't know if you've seen this,

0:01:00.000 --> 0:01:03.160
<v Speaker 1>but the company Anthropic just released the report that shows

0:01:03.240 --> 0:01:05.959
<v Speaker 1>the jobs that are most likely to be replaced by

0:01:06.000 --> 0:01:11.960
<v Speaker 1>AI in the coming years. If you're in management, business, computers, engineering, law,

0:01:12.360 --> 0:01:16.640
<v Speaker 1>even science, it's not looking so good for you. On

0:01:16.680 --> 0:01:20.240
<v Speaker 1>the other hand, if you're in agriculture, grounds maintenance, or

0:01:20.240 --> 0:01:22.360
<v Speaker 1>if you know how to fix a fridge or ac

0:01:23.160 --> 0:01:27.760
<v Speaker 1>you might be okay, which made me wonder where comedians stand.

0:01:28.120 --> 0:01:30.720
<v Speaker 1>Some people say that comedy and humor are some of

0:01:30.720 --> 0:01:33.600
<v Speaker 1>the things that make us uniquely human, but is.

0:01:33.640 --> 0:01:34.360
<v Speaker 2>That really true?

0:01:34.680 --> 0:01:39.199
<v Speaker 1>Can AI be just as funny as people? To answer

0:01:39.280 --> 0:01:42.399
<v Speaker 1>this question, I reached out to doctor Lydia Chilton, a

0:01:42.480 --> 0:01:46.360
<v Speaker 1>professor of computer science at Columbia University who specializes in

0:01:46.480 --> 0:01:50.480
<v Speaker 1>AI and human computer interaction. She also happens to have

0:01:50.520 --> 0:01:54.200
<v Speaker 1>spent the last fifteen years trying to prove that humor

0:01:54.280 --> 0:01:57.480
<v Speaker 1>can be done by a computer, because if it can,

0:01:57.640 --> 0:02:02.840
<v Speaker 1>it means we actually understand what humor is. Now, doctor

0:02:02.920 --> 0:02:05.920
<v Speaker 1>Chiltern is gonna tell us whether or not she actually succeeded,

0:02:06.160 --> 0:02:08.880
<v Speaker 1>but first I wanted to ask her a more basic question,

0:02:09.200 --> 0:02:13.560
<v Speaker 1>which is what makes something funny? So here's my conversation

0:02:13.639 --> 0:02:18.680
<v Speaker 1>with doctor Lydia Chilton. Well, thank you doctor Chiltern for

0:02:18.760 --> 0:02:19.240
<v Speaker 1>joining us.

0:02:19.440 --> 0:02:21.000
<v Speaker 2>Thank you excited to be here.

0:02:21.160 --> 0:02:23.359
<v Speaker 1>I have a joke for you. Here we got knock.

0:02:23.440 --> 0:02:24.160
<v Speaker 2>Who's there?

0:02:24.440 --> 0:02:24.880
<v Speaker 1>Iva?

0:02:25.200 --> 0:02:25.920
<v Speaker 2>Iva? Who?

0:02:26.440 --> 0:02:28.560
<v Speaker 1>I've got a feeling I'm going to need more data

0:02:28.600 --> 0:02:29.399
<v Speaker 1>to finish this joke.

0:02:32.000 --> 0:02:35.280
<v Speaker 2>All right, you got me that's pretty funny.

0:02:35.720 --> 0:02:38.480
<v Speaker 1>Well, actually I got this one from Gemini, So I

0:02:38.520 --> 0:02:41.120
<v Speaker 1>asked Gemini right before this call to tell me a

0:02:41.160 --> 0:02:42.360
<v Speaker 1>knock knock joke about Ai.

0:02:42.720 --> 0:02:43.240
<v Speaker 2>Wow.

0:02:43.600 --> 0:02:46.640
<v Speaker 1>Yeah, go Gemini. Yeah, and you laughed.

0:02:46.800 --> 0:02:47.240
<v Speaker 2>Hi did.

0:02:48.960 --> 0:02:50.600
<v Speaker 1>Well. That brings me to the first question I have

0:02:50.680 --> 0:02:53.600
<v Speaker 1>for you, which is what makes something funny? M M.

0:02:54.000 --> 0:02:56.320
<v Speaker 1>Can you pinpoint when you started to get curious about humor?

0:02:56.960 --> 0:03:01.200
<v Speaker 2>Ah, I would say I was five years old and

0:03:01.680 --> 0:03:05.919
<v Speaker 2>I accidentally made a pun and my dad laughed so hysterically,

0:03:06.040 --> 0:03:08.960
<v Speaker 2>and I felt so good that I made my dad

0:03:09.120 --> 0:03:12.320
<v Speaker 2>last and I was like, must repeat, must repeat, And

0:03:12.400 --> 0:03:15.280
<v Speaker 2>of course it can't happen on command, and that one

0:03:15.400 --> 0:03:17.880
<v Speaker 2>was an accident. But in order to get people to

0:03:18.120 --> 0:03:20.800
<v Speaker 2>like me, I've wanted to figure out the formula for

0:03:20.840 --> 0:03:23.520
<v Speaker 2>a long time and been convinced that there is.

0:03:23.480 --> 0:03:27.520
<v Speaker 1>One, which is a very sweet reason to be interested

0:03:27.560 --> 0:03:28.040
<v Speaker 1>in the topic.

0:03:28.360 --> 0:03:29.200
<v Speaker 2>Yeah.

0:03:29.520 --> 0:03:32.480
<v Speaker 1>I read that in twenty thirteen, you went to the

0:03:32.480 --> 0:03:35.040
<v Speaker 1>internet and you try to recruit a whole bunch of

0:03:35.080 --> 0:03:37.720
<v Speaker 1>people from the Internet to give you money to try

0:03:37.760 --> 0:03:40.280
<v Speaker 1>to answer this question. Can you tell us about that?

0:03:40.520 --> 0:03:45.000
<v Speaker 2>Yeah? So there's some questions that are just so perennial

0:03:45.080 --> 0:03:47.520
<v Speaker 2>that I can't stop myself from thinking about them. This

0:03:47.600 --> 0:03:50.080
<v Speaker 2>is one you know, Plato had his set what is good,

0:03:50.160 --> 0:03:54.160
<v Speaker 2>what is justice? I've got mine what is funny? And

0:03:54.240 --> 0:03:56.760
<v Speaker 2>I actually think they're kind of the same. You know,

0:03:56.920 --> 0:03:58.840
<v Speaker 2>not to put me in Plato in the same boat,

0:03:58.880 --> 0:04:03.280
<v Speaker 2>but right, No, they both have a strong emotional component

0:04:03.360 --> 0:04:05.680
<v Speaker 2>to them. There's something about the chemicals that happen in

0:04:05.720 --> 0:04:09.200
<v Speaker 2>our brain that makes something funny. And that's why something

0:04:09.240 --> 0:04:11.960
<v Speaker 2>could be different funny for you, for me, for me

0:04:12.120 --> 0:04:14.600
<v Speaker 2>at a different time, for me if I heard it

0:04:14.640 --> 0:04:18.360
<v Speaker 2>slightly differently, maybe if someone less funny told it. So

0:04:18.720 --> 0:04:22.120
<v Speaker 2>it's not all about just the joke and whether the

0:04:22.320 --> 0:04:26.120
<v Speaker 2>joke was funny. There's a lot of other circumstances, Like

0:04:26.160 --> 0:04:28.360
<v Speaker 2>the whole thing isn't like logically constructed.

0:04:29.880 --> 0:04:32.200
<v Speaker 1>That bothers you, well, it makes it.

0:04:32.160 --> 0:04:34.680
<v Speaker 2>Very hard to understand. But like, so that could either

0:04:34.680 --> 0:04:36.279
<v Speaker 2>be a good thing or a bad thing. Like a

0:04:36.320 --> 0:04:37.919
<v Speaker 2>bad thing if you're like I need to understand this

0:04:37.960 --> 0:04:39.680
<v Speaker 2>before the test tomorrow, but a good thing. It's like,

0:04:39.800 --> 0:04:42.200
<v Speaker 2>I would like to study this for eternity because there's gonna.

0:04:42.000 --> 0:04:45.200
<v Speaker 1>Be lots of problems. We may never figure it out.

0:04:45.240 --> 0:04:48.200
<v Speaker 1>Which gives you, as a professor job security.

0:04:48.200 --> 0:04:52.600
<v Speaker 2>Exactly, exactly. So I talked to a lot of people

0:04:52.600 --> 0:04:54.400
<v Speaker 2>about this, read a lot of books, because it turns

0:04:54.440 --> 0:04:57.840
<v Speaker 2>out I wasn't the first person to study humor and

0:04:57.880 --> 0:05:01.120
<v Speaker 2>a philosopher had a theory about this is at least interesting.

0:05:01.320 --> 0:05:04.280
<v Speaker 2>It was evolved in us so that we could learn

0:05:04.320 --> 0:05:07.919
<v Speaker 2>from mistakes of others. Someone else falls on a banana

0:05:08.040 --> 0:05:15.120
<v Speaker 2>peel and liked I would not do that. And you know,

0:05:15.160 --> 0:05:17.160
<v Speaker 2>it's hard to say, but if we're looking back for

0:05:17.240 --> 0:05:20.599
<v Speaker 2>some origins of why we might have benefited as a

0:05:20.680 --> 0:05:24.800
<v Speaker 2>species from this, learning from other people's mistakes is a

0:05:24.800 --> 0:05:27.719
<v Speaker 2>pretty good way of learning rather watching someone else and

0:05:27.839 --> 0:05:31.039
<v Speaker 2>learning don't step on a banana peel as a positive reward.

0:05:31.320 --> 0:05:34.560
<v Speaker 1>Interesting. Wait, so the theory is that funnyness is basically

0:05:34.880 --> 0:05:35.599
<v Speaker 1>shot and fraud.

0:05:35.839 --> 0:05:41.040
<v Speaker 2>Yes, yes, shot and freud to make us better In theory.

0:05:40.960 --> 0:05:42.760
<v Speaker 1>Mean he liked a way for your body to feel

0:05:42.800 --> 0:05:45.640
<v Speaker 1>good at the expense of others, so that we learn

0:05:45.800 --> 0:05:46.279
<v Speaker 1>from them.

0:05:46.560 --> 0:05:48.679
<v Speaker 2>Yeah. So the next time you see a banana peel,

0:05:48.960 --> 0:05:52.279
<v Speaker 2>you have that memory because it's a strong positive You're like, oh,

0:05:52.360 --> 0:05:54.599
<v Speaker 2>that guy felt maybe I won't step on it.

0:05:54.720 --> 0:05:56.960
<v Speaker 1>That's funny. I don't want to be the person people

0:05:57.040 --> 0:06:01.400
<v Speaker 1>laugh at. Oh interesting, Okay, no one.

0:06:01.200 --> 0:06:03.640
<v Speaker 2>Can prove that's true or not. I can't go back

0:06:03.680 --> 0:06:06.039
<v Speaker 2>in time twenty thousand years and write down all the

0:06:06.040 --> 0:06:07.560
<v Speaker 2>banana pe ins.

0:06:07.600 --> 0:06:10.640
<v Speaker 1>It iss although that would be pretty funny, Like.

0:06:10.640 --> 0:06:14.120
<v Speaker 2>Where's the data, It's nowhere. It's a theory.

0:06:14.320 --> 0:06:17.440
<v Speaker 1>It is very complicated and mysterious.

0:06:17.040 --> 0:06:21.920
<v Speaker 2>Humor exactly job security, at.

0:06:21.920 --> 0:06:24.960
<v Speaker 1>Least for now until potentially AIS can do it. So

0:06:25.040 --> 0:06:25.599
<v Speaker 1>we'll get to that.

0:06:25.920 --> 0:06:26.160
<v Speaker 2>Good.

0:06:26.440 --> 0:06:28.919
<v Speaker 1>Okay, what else did you learn about what makes something funny?

0:06:29.080 --> 0:06:32.000
<v Speaker 2>Well? I read a whole bunch of theories, and there's

0:06:32.040 --> 0:06:33.760
<v Speaker 2>a lot of different ones, and they all like kind

0:06:33.760 --> 0:06:36.800
<v Speaker 2>of go together. So one is it has to be surprising.

0:06:37.120 --> 0:06:39.880
<v Speaker 2>It's not surprising. It's hard to get any kind of

0:06:39.880 --> 0:06:44.080
<v Speaker 2>emotional reaction out of it. Surprising often is correlated with

0:06:44.240 --> 0:06:46.920
<v Speaker 2>quick or has a quick turn, and it's like, oh,

0:06:46.960 --> 0:06:49.440
<v Speaker 2>I'm saying one thing and that boops, just kidding.

0:06:49.600 --> 0:06:51.800
<v Speaker 1>Somebody's walking suddenly they slip in a banana.

0:06:51.839 --> 0:06:55.719
<v Speaker 2>Whoa what exactly? Because you expected them to keep walking

0:06:55.880 --> 0:06:58.839
<v Speaker 2>and then right very suddenly if they fall very slowly,

0:06:58.920 --> 0:07:02.880
<v Speaker 2>that's less funny. Yeah, but the most important is it

0:07:02.920 --> 0:07:06.120
<v Speaker 2>has to be in the subtext of what set So

0:07:06.360 --> 0:07:09.680
<v Speaker 2>all stories, and I consider jokes stories for small stories

0:07:09.880 --> 0:07:12.040
<v Speaker 2>have like the text like what said, and then the

0:07:12.080 --> 0:07:15.120
<v Speaker 2>subtext is what it really means are what the significance

0:07:15.520 --> 0:07:19.760
<v Speaker 2>is to it. Typically jokes they have this surprise in

0:07:19.800 --> 0:07:22.560
<v Speaker 2>them because at first they leave you to believe that

0:07:22.600 --> 0:07:25.880
<v Speaker 2>there's one subtext, and then something happens you're like, well,

0:07:25.920 --> 0:07:27.680
<v Speaker 2>now there's a total other subtext.

0:07:28.400 --> 0:07:32.200
<v Speaker 1>I see. It's about our expectation about what's under the surface.

0:07:32.320 --> 0:07:36.320
<v Speaker 2>And how you're interpreting it. So my go to explainer

0:07:36.440 --> 0:07:40.120
<v Speaker 2>joke for this is here goes there are three kinds

0:07:40.120 --> 0:07:42.600
<v Speaker 2>of people in the world, people who can count and

0:07:42.680 --> 0:07:47.400
<v Speaker 2>people who can't. So I led you to believe that

0:07:47.480 --> 0:07:49.960
<v Speaker 2>I was a person who is intelligent and had three

0:07:50.000 --> 0:07:52.920
<v Speaker 2>things to say, and then my list too, and you're like, wait,

0:07:53.120 --> 0:07:56.360
<v Speaker 2>what in your brain kind of rearranges and then you realize, oh,

0:07:56.400 --> 0:08:01.160
<v Speaker 2>she can't count, she's an idiot, haha, and that's what

0:08:01.200 --> 0:08:03.080
<v Speaker 2>makes it funny. So those are the two texts.

0:08:03.120 --> 0:08:05.720
<v Speaker 1>Yeah, I was expecting three things. I was surprised when

0:08:05.760 --> 0:08:08.640
<v Speaker 1>you stopped it two. But then in my brain pieced

0:08:08.640 --> 0:08:11.680
<v Speaker 1>it together. Oh, it's because she said she must be

0:08:11.720 --> 0:08:14.720
<v Speaker 1>one of the people who can't count exactly.

0:08:14.440 --> 0:08:19.040
<v Speaker 2>So there's some expectation violation in their expectation violation is

0:08:19.120 --> 0:08:21.520
<v Speaker 2>one of the main theories, but it's not sufficient. If

0:08:21.520 --> 0:08:24.840
<v Speaker 2>I just said purple, you weren't expecting me to say purple.

0:08:24.880 --> 0:08:28.920
<v Speaker 2>But it's not funny, So the subtext is very very important. Okay,

0:08:29.080 --> 0:08:33.200
<v Speaker 2>that said, farts are also funny, and that doesn't Uh,

0:08:33.240 --> 0:08:36.640
<v Speaker 2>it's hard to explain that with samantic script theory of humor.

0:08:36.720 --> 0:08:39.240
<v Speaker 2>I don't know what the subtext the part.

0:08:39.200 --> 0:08:45.520
<v Speaker 1>Is, Well, definitely they're funny, especially if you're ten years old. Yes,

0:08:45.800 --> 0:08:48.360
<v Speaker 1>maybe the subtext is that you're not supposed to do

0:08:48.440 --> 0:08:52.720
<v Speaker 1>it in public for others to hear or smell. Yeah, okay,

0:08:52.760 --> 0:08:54.760
<v Speaker 1>this is going to be a little important later on

0:08:54.840 --> 0:08:56.920
<v Speaker 1>when we talk about how to get a computer to

0:08:56.960 --> 0:08:59.960
<v Speaker 1>be funny. One of the most popular theories about humor

0:09:00.360 --> 0:09:03.800
<v Speaker 1>is that it violates our expectations in a surprising but

0:09:03.840 --> 0:09:07.520
<v Speaker 1>not unrelated Wait, so the first part of the joke

0:09:07.760 --> 0:09:10.920
<v Speaker 1>makes us assume one context, but then something happens and

0:09:10.960 --> 0:09:13.440
<v Speaker 1>it turns out there was a second hidden way to

0:09:13.520 --> 0:09:16.920
<v Speaker 1>interpret what was happening, which we had missed. Of course,

0:09:17.040 --> 0:09:19.800
<v Speaker 1>That's not the only thing that makes something funny.

0:09:21.640 --> 0:09:23.800
<v Speaker 2>There's a lot of other things that make a joke

0:09:24.000 --> 0:09:27.760
<v Speaker 2>more or less funny. Okay, I call them heighteners. They

0:09:27.800 --> 0:09:31.040
<v Speaker 2>just heighten the surprise or the level of funny or

0:09:31.080 --> 0:09:34.280
<v Speaker 2>something about the emotion with them. So one is being mean,

0:09:34.559 --> 0:09:37.840
<v Speaker 2>Like typically the more mean a joke is the funnier

0:09:37.920 --> 0:09:39.960
<v Speaker 2>it is until it like really chips over and it's

0:09:40.000 --> 0:09:41.920
<v Speaker 2>like too soon or like or too.

0:09:41.760 --> 0:09:44.200
<v Speaker 1>Mean, meaning like somebody has to be the butt of

0:09:44.240 --> 0:09:44.640
<v Speaker 1>the joke.

0:09:44.960 --> 0:09:47.200
<v Speaker 2>Yeah, so mine was mean. I was the butt of

0:09:47.200 --> 0:09:49.960
<v Speaker 2>my own joke that I can't count. Uh, it's funnier

0:09:50.000 --> 0:09:50.600
<v Speaker 2>if it's mean.

0:09:51.320 --> 0:09:55.240
<v Speaker 1>It has to say it expose somebody, some person, Yeah, right,

0:09:55.360 --> 0:09:58.559
<v Speaker 1>Like it can't be about exposing a chair or exposing

0:09:58.600 --> 0:10:01.079
<v Speaker 1>a material or or something nerds.

0:10:00.920 --> 0:10:03.760
<v Speaker 2>Something related to people. It could be like a group

0:10:03.800 --> 0:10:07.480
<v Speaker 2>of people, like women or men. There's a lot there's

0:10:07.480 --> 0:10:10.280
<v Speaker 2>a lot of jokes that take that persuasion.

0:10:09.760 --> 0:10:12.600
<v Speaker 1>And that movie pats into that shout and Freud shot

0:10:12.679 --> 0:10:13.400
<v Speaker 1>and Freud.

0:10:13.400 --> 0:10:15.960
<v Speaker 2>Yes, schotenfreud of. I don't know it's German. Who knows

0:10:16.000 --> 0:10:18.720
<v Speaker 2>how to pronounce it? Probably nobody. It's not like there's

0:10:18.760 --> 0:10:19.640
<v Speaker 2>a country.

0:10:19.240 --> 0:10:22.480
<v Speaker 1>Of people that and that would be a fandfire by

0:10:22.520 --> 0:10:25.640
<v Speaker 1>my lacke of language skills. But it's like you said,

0:10:25.640 --> 0:10:27.640
<v Speaker 1>it's about learning from the mistakes of others.

0:10:27.840 --> 0:10:30.480
<v Speaker 2>Yes, exactly interesting, So.

0:10:30.520 --> 0:10:32.080
<v Speaker 1>That makes jokes even funnier.

0:10:32.280 --> 0:10:35.160
<v Speaker 2>Yeah. Another thing that makes them funnier is when you

0:10:35.280 --> 0:10:39.120
<v Speaker 2>really relate to them, and particularly in an in crowd

0:10:39.280 --> 0:10:41.400
<v Speaker 2>kind of way, when you're like, oh, I get back.

0:10:41.440 --> 0:10:44.160
<v Speaker 2>So when someone tells, like a computer science joke that

0:10:44.240 --> 0:10:46.440
<v Speaker 2>I'm like, ooh, I get it, and I feel special

0:10:46.440 --> 0:10:50.440
<v Speaker 2>because the physicists aren't going to get it. And this

0:10:50.480 --> 0:10:54.840
<v Speaker 2>can be whether it's about our generation, our family, or anything,

0:10:55.440 --> 0:10:59.800
<v Speaker 2>but inside jokes definitely produce a special kind of chemical

0:10:59.840 --> 0:11:02.199
<v Speaker 2>that make us feel like we're part of something, and

0:11:02.240 --> 0:11:03.120
<v Speaker 2>that heightens it.

0:11:03.400 --> 0:11:04.400
<v Speaker 1>Yeah, that's funnier.

0:11:04.760 --> 0:11:09.040
<v Speaker 2>It is funny or more emotional, and that heightens the funny.

0:11:08.720 --> 0:11:12.240
<v Speaker 1>Because it exposes something you didn't think anybody else knew,

0:11:12.520 --> 0:11:13.520
<v Speaker 1>something you should talk about.

0:11:13.679 --> 0:11:17.480
<v Speaker 2>So yeah, part of it is, and Freud talked about this,

0:11:17.880 --> 0:11:21.120
<v Speaker 2>a part of it is the relief of being able

0:11:21.160 --> 0:11:23.880
<v Speaker 2>to say something that you might not have otherwise been

0:11:23.920 --> 0:11:27.200
<v Speaker 2>able to say or reveal. You know, in society we're

0:11:27.600 --> 0:11:29.319
<v Speaker 2>just to go around and show our best face to

0:11:29.400 --> 0:11:31.880
<v Speaker 2>people pretend that we're awesome, and to have Instagram pages

0:11:31.880 --> 0:11:34.320
<v Speaker 2>that makes us look like we're in vacation Italy half

0:11:34.360 --> 0:11:37.400
<v Speaker 2>the time. And so when you realize that someone else

0:11:37.520 --> 0:11:40.360
<v Speaker 2>is also not actually going to Italy, they took one

0:11:40.440 --> 0:11:43.640
<v Speaker 2>vacation and been dripping the photos throughout the year that yeah,

0:11:43.840 --> 0:11:46.480
<v Speaker 2>my Instagram page is a lie, and that sense of

0:11:46.600 --> 0:11:50.360
<v Speaker 2>relief to people that yes, I'm like that too. It

0:11:50.440 --> 0:11:53.880
<v Speaker 2>rings true. Something being true like it also heightens it.

0:11:53.920 --> 0:11:56.360
<v Speaker 2>There's like a satisfaction and be like, yeah, that's how

0:11:56.400 --> 0:12:00.160
<v Speaker 2>it really is. That person's telling it straight. This was

0:12:00.240 --> 0:12:03.600
<v Speaker 2>like a mental satisfaction of when you hear something that's true.

0:12:03.559 --> 0:12:06.800
<v Speaker 1>Right right, Like it puts a voice to something you've

0:12:06.840 --> 0:12:11.240
<v Speaker 1>suspected or felt but never actually maybe put into words before.

0:12:11.480 --> 0:12:13.880
<v Speaker 2>Yeah. Right, And so there's still like a subtext going

0:12:13.880 --> 0:12:17.160
<v Speaker 2>on there, but the subtext that you're hearing is like, oh, yes,

0:12:17.280 --> 0:12:20.319
<v Speaker 2>I always wanted someone to admit that airline food is bad.

0:12:20.679 --> 0:12:22.840
<v Speaker 2>I've never been able to express that before.

0:12:24.600 --> 0:12:26.240
<v Speaker 1>I see keep quing, keep quick.

0:12:26.400 --> 0:12:29.080
<v Speaker 2>Since we've already opened up the topic of fart jokes,

0:12:29.120 --> 0:12:33.840
<v Speaker 2>things being dirty, sexual body related, all the things that

0:12:33.920 --> 0:12:36.120
<v Speaker 2>swear words have in common. There's a special part of

0:12:36.160 --> 0:12:38.920
<v Speaker 2>our brain that is like reserved for like swear words

0:12:38.920 --> 0:12:42.560
<v Speaker 2>and taboo things. Anytime you light that up, it also

0:12:42.760 --> 0:12:46.520
<v Speaker 2>sort of magnifies anything, whether it's a joke or a

0:12:46.559 --> 0:12:49.920
<v Speaker 2>compliment or anything else. It's just a heightener of all kinds.

0:12:50.080 --> 0:12:53.920
<v Speaker 2>Extra brain cells start firing when you hear those kinds

0:12:53.960 --> 0:12:57.640
<v Speaker 2>of words. M So try adding a swear word.

0:12:58.920 --> 0:13:04.959
<v Speaker 1>I was gonna say, that is so effing true.

0:13:03.040 --> 0:13:06.320
<v Speaker 2>And that was funny because you did that. There's many

0:13:06.360 --> 0:13:10.679
<v Speaker 2>things that heighten our emotions or feelings of something. Some

0:13:10.720 --> 0:13:14.280
<v Speaker 2>people think certain letters and words are very funny. So

0:13:14.559 --> 0:13:17.440
<v Speaker 2>words with more k's in them will make people laugh more.

0:13:18.280 --> 0:13:22.680
<v Speaker 2>Maybe I just where's the data? Yeah, there's many many

0:13:22.720 --> 0:13:26.280
<v Speaker 2>hypotheses out there waiting to be tested about what's funny? Yeah,

0:13:26.320 --> 0:13:29.080
<v Speaker 2>about what heightens jokes, what makes them funny? Like exactly

0:13:29.120 --> 0:13:31.600
<v Speaker 2>what kind of subtext is and is not funny? Where

0:13:31.600 --> 0:13:34.640
<v Speaker 2>in the exact turn is all right?

0:13:35.120 --> 0:13:38.200
<v Speaker 1>So, doctor Chilton spent years studying humor and what makes

0:13:38.240 --> 0:13:40.959
<v Speaker 1>things funny. She talked to comedians and tapped into the

0:13:41.040 --> 0:13:44.640
<v Speaker 1>humor science community. Yes there is one, and she came

0:13:44.679 --> 0:13:47.200
<v Speaker 1>out of it with a bunch of rules about comedy.

0:13:47.600 --> 0:13:49.400
<v Speaker 1>The next step was to see if she could get

0:13:49.440 --> 0:13:53.960
<v Speaker 1>a computer to follow those rules. Could an AI be funny?

0:13:54.800 --> 0:13:57.120
<v Speaker 1>So when we come back, we'll talk about the different

0:13:57.160 --> 0:13:59.680
<v Speaker 1>ways in which programmers have tried to do that, and

0:13:59.720 --> 0:14:02.760
<v Speaker 1>with a the joke is on AIS or on us,

0:14:03.240 --> 0:14:18.840
<v Speaker 1>So stay with us for the punchline. We'll be right back. Hey,

0:14:18.880 --> 0:14:22.560
<v Speaker 1>welcome back. We're talking about whether AI can be funny,

0:14:22.840 --> 0:14:26.440
<v Speaker 1>and our guest today is doctor Lydia Chilton. As we

0:14:26.480 --> 0:14:29.480
<v Speaker 1>mentioned before, in twenty thirteen, she started a project to

0:14:29.640 --> 0:14:33.760
<v Speaker 1>dissect what makes something funny, not only to understand humor

0:14:33.880 --> 0:14:36.760
<v Speaker 1>to satisfy her own curiosity, but to see if you

0:14:36.760 --> 0:14:40.360
<v Speaker 1>could get a computer to be funny. However, not everyone

0:14:40.400 --> 0:14:42.040
<v Speaker 1>thought it was a good idea.

0:14:43.320 --> 0:14:44.760
<v Speaker 2>Oh I got a lot of hate mail too, but

0:14:45.000 --> 0:14:46.960
<v Speaker 2>it's the Internet. Of course you're gonna get hate mail.

0:14:47.600 --> 0:14:51.480
<v Speaker 2>I'm like, how dare you? I thought that was interesting too,

0:14:51.560 --> 0:14:55.440
<v Speaker 2>that people be offended by this, and I'm like, it's science,

0:14:56.160 --> 0:14:58.120
<v Speaker 2>you know. So for me, that was the first tip

0:14:58.160 --> 0:15:02.880
<v Speaker 2>of people being threatened by AI, especially when computers creep

0:15:02.880 --> 0:15:05.680
<v Speaker 2>into creative areas. I think people kind of get this,

0:15:05.800 --> 0:15:08.120
<v Speaker 2>but that's mine. I'm a human. I'm creative. That's part

0:15:08.160 --> 0:15:08.800
<v Speaker 2>of my identity.

0:15:08.960 --> 0:15:12.440
<v Speaker 1>Oh, you got pushback for even kind of trying to

0:15:12.520 --> 0:15:15.560
<v Speaker 1>dissect it, maybe with the goal of getting computers to

0:15:15.600 --> 0:15:15.880
<v Speaker 1>do it.

0:15:16.200 --> 0:15:18.640
<v Speaker 2>Yeah, And I was just saying, well, it's just dissecting it,

0:15:18.680 --> 0:15:20.680
<v Speaker 2>but like, come on, if you dissect it well enough,

0:15:20.720 --> 0:15:22.960
<v Speaker 2>there's a very high probability that you'll be able to

0:15:23.320 --> 0:15:26.360
<v Speaker 2>generate it. But these weren't even like honestly comedians were

0:15:26.400 --> 0:15:28.600
<v Speaker 2>more like, yeah, I kind of want to know this too.

0:15:29.400 --> 0:15:30.880
<v Speaker 2>They weren't just threatened by it.

0:15:33.440 --> 0:15:37.480
<v Speaker 1>Okay. So despite this pushback, doctor Chilton pressed on. But

0:15:37.720 --> 0:15:41.400
<v Speaker 1>she needed data on humor, so she turned to The Onion,

0:15:41.720 --> 0:15:45.120
<v Speaker 1>the humor and satire magazine that's been published since nineteen

0:15:45.200 --> 0:15:45.680
<v Speaker 1>eighty eight.

0:15:46.840 --> 0:15:49.760
<v Speaker 2>And so someone pointed out to me that The Onion,

0:15:49.880 --> 0:15:53.160
<v Speaker 2>which is known for making up really funny fake headlines

0:15:53.240 --> 0:15:57.000
<v Speaker 2>and news stories, also had a different section that was unusual.

0:15:57.040 --> 0:15:59.960
<v Speaker 2>They took a real headline and came up with three

0:16:00.520 --> 0:16:04.560
<v Speaker 2>funny man on the street responses to it from average,

0:16:04.760 --> 0:16:09.520
<v Speaker 2>usually idiotic Americans. And I liked this as a framework

0:16:09.600 --> 0:16:11.760
<v Speaker 2>for studying the joke because you could just take the

0:16:11.840 --> 0:16:16.040
<v Speaker 2>input the original headline, and then you had the Onions

0:16:16.080 --> 0:16:20.400
<v Speaker 2>like verifiably funny statements, and then you could, with the

0:16:20.480 --> 0:16:23.240
<v Speaker 2>same headline, come up with your own or try to

0:16:23.320 --> 0:16:26.880
<v Speaker 2>back engineer the jokes that the onion made to figure out, Okay,

0:16:26.920 --> 0:16:29.120
<v Speaker 2>what are some of the properties of these jokes, and

0:16:29.160 --> 0:16:31.160
<v Speaker 2>what are some of the techniques for doing them, and

0:16:31.200 --> 0:16:34.200
<v Speaker 2>what's the variety in them. So this little test bed

0:16:34.600 --> 0:16:38.160
<v Speaker 2>became very important as just a mechanism for which I

0:16:38.160 --> 0:16:39.680
<v Speaker 2>could sort of study humor in a.

0:16:39.640 --> 0:16:42.040
<v Speaker 1>Bottle and said, what did you learn that experiment you

0:16:42.040 --> 0:16:43.280
<v Speaker 1>did when in twenty.

0:16:43.640 --> 0:16:48.200
<v Speaker 2>Twenty thirteen, twenty fourteen, twenty fifteen. It kept going because

0:16:48.240 --> 0:16:51.200
<v Speaker 2>there was a lot to learn, as it turned out,

0:16:51.840 --> 0:16:54.680
<v Speaker 2>But we did find some things that all the jokes had.

0:16:54.880 --> 0:16:58.840
<v Speaker 2>All the jokes had at least two connections to the headline.

0:16:59.720 --> 0:17:03.320
<v Speaker 2>They were taking the original headline and taking the kind

0:17:03.320 --> 0:17:08.520
<v Speaker 2>of associating things with those entities, with the people mentioned

0:17:08.560 --> 0:17:11.320
<v Speaker 2>in them, and finding new things to say about them

0:17:11.440 --> 0:17:13.080
<v Speaker 2>and a new connection between them.

0:17:13.320 --> 0:17:15.840
<v Speaker 1>I see, like you're saying, like finding that other subtics

0:17:15.840 --> 0:17:17.159
<v Speaker 1>that is surprising.

0:17:17.040 --> 0:17:20.679
<v Speaker 2>Exactly, And then I was like, uh duh. As a

0:17:20.680 --> 0:17:23.600
<v Speaker 2>computer scientist, I realized I'd kind of been looking in

0:17:23.640 --> 0:17:26.040
<v Speaker 2>the wrong places. I had been looking for like the

0:17:26.160 --> 0:17:29.879
<v Speaker 2>logic behind jokes. But I quickly realized that jokes and

0:17:30.000 --> 0:17:34.879
<v Speaker 2>most human communication is about our loose associations, like why

0:17:35.000 --> 0:17:37.520
<v Speaker 2>when I think McDonald's, why do I think Burger King?

0:17:37.600 --> 0:17:40.719
<v Speaker 2>Why do I It's just like they're in the same category. There,

0:17:40.760 --> 0:17:44.040
<v Speaker 2>It's an association I have. And at the time, computer

0:17:44.119 --> 0:17:47.199
<v Speaker 2>science was all logic, and so I actually kind of

0:17:47.240 --> 0:17:49.800
<v Speaker 2>put the project on a shelf. But I realized, Okay,

0:17:49.840 --> 0:17:51.840
<v Speaker 2>we need an association engine.

0:17:52.440 --> 0:17:54.280
<v Speaker 1>Okay, let me see if I get this. It was

0:17:54.480 --> 0:17:58.919
<v Speaker 1>like the twenty early twenty tens. You dissected what humor is.

0:17:58.960 --> 0:18:01.560
<v Speaker 1>You sort of found all the these patterns and.

0:18:01.480 --> 0:18:03.080
<v Speaker 2>Ye by analyzing the onion.

0:18:03.440 --> 0:18:07.280
<v Speaker 1>By analyzing you peeled back the layers of the onion. Yeah,

0:18:10.119 --> 0:18:13.000
<v Speaker 1>you found the nuggets, some nuggets and rules, some patterns

0:18:13.080 --> 0:18:15.320
<v Speaker 1>about what makes something funny. And your goal was to

0:18:15.400 --> 0:18:17.880
<v Speaker 1>maybe try to get a computer to do this, right.

0:18:18.200 --> 0:18:19.080
<v Speaker 2>Yeah, But it couldn't.

0:18:19.280 --> 0:18:23.000
<v Speaker 1>It couldn't at the time because I think computer science

0:18:23.040 --> 0:18:25.560
<v Speaker 1>back then and artificial intelligence back then was kind of

0:18:25.600 --> 0:18:27.800
<v Speaker 1>about setting up the right rules, right.

0:18:28.280 --> 0:18:30.920
<v Speaker 2>Yeah, it was all about logic. It was like, what

0:18:30.960 --> 0:18:33.119
<v Speaker 2>are the logical rules to do this? How do I

0:18:33.160 --> 0:18:35.520
<v Speaker 2>add one plus one. I take this bit and I

0:18:35.640 --> 0:18:37.840
<v Speaker 2>combine it with this bit, and I do an and

0:18:37.840 --> 0:18:40.960
<v Speaker 2>and that creates two. All of that kind of stuff,

0:18:42.000 --> 0:18:44.560
<v Speaker 2>But no, like, what do you associate with one? Oh?

0:18:44.680 --> 0:18:47.960
<v Speaker 2>One is the loneliest number number one, We're number one,

0:18:48.440 --> 0:18:51.080
<v Speaker 2>Avis is number two. Those are the thoughts that people

0:18:51.200 --> 0:18:54.640
<v Speaker 2>have about one that computers have no idea. And it's

0:18:54.680 --> 0:18:57.399
<v Speaker 2>even hard to get from the Internet, I see because

0:18:57.400 --> 0:19:00.359
<v Speaker 2>it's the things that people say and not necessarily really

0:19:00.400 --> 0:19:04.160
<v Speaker 2>the things that people write down. And it's about frequency,

0:19:04.400 --> 0:19:06.720
<v Speaker 2>like how often it happens, rather than the fact that

0:19:06.800 --> 0:19:07.679
<v Speaker 2>it did happen.

0:19:08.480 --> 0:19:10.800
<v Speaker 1>Okay, okay, I am getting here to the nugget of

0:19:10.800 --> 0:19:13.720
<v Speaker 1>the onion here. You sort of figured out that humor

0:19:14.040 --> 0:19:17.080
<v Speaker 1>was about starting with a subtext that people would recognize,

0:19:17.119 --> 0:19:20.240
<v Speaker 1>but then being able to find that secondary that other

0:19:20.440 --> 0:19:23.080
<v Speaker 1>subtexts that will be surprising when you put all the

0:19:23.119 --> 0:19:24.480
<v Speaker 1>pieces together in the joke.

0:19:24.400 --> 0:19:26.200
<v Speaker 2>Yes, surprising, but still relevant.

0:19:26.280 --> 0:19:29.879
<v Speaker 1>It's still relevant, that's right. Second, suptics and trying to

0:19:29.960 --> 0:19:33.600
<v Speaker 1>find those subtags is hard if you're just going by

0:19:33.640 --> 0:19:36.639
<v Speaker 1>the literal definition of worth and using logic, right and

0:19:36.680 --> 0:19:37.119
<v Speaker 1>things like that.

0:19:37.160 --> 0:19:38.520
<v Speaker 2>Exactly, it's just not there.

0:19:39.400 --> 0:19:41.840
<v Speaker 1>So you start said we can't do this with rules

0:19:41.840 --> 0:19:43.080
<v Speaker 1>and logic, so you shelved it.

0:19:43.320 --> 0:19:44.199
<v Speaker 2>Yeah, I gave up.

0:19:44.240 --> 0:19:47.640
<v Speaker 1>People weren't convinced that computers could be funny.

0:19:47.720 --> 0:19:51.639
<v Speaker 2>Yes, they weren't convinced that I had decomposed and reconstructed

0:19:51.680 --> 0:19:55.360
<v Speaker 2>the process of writing humor when humans were still involved

0:19:55.359 --> 0:19:56.880
<v Speaker 2>in some part of that process.

0:19:57.200 --> 0:20:00.200
<v Speaker 1>Oh, I see, because you still needed some human to

0:20:00.240 --> 0:20:01.400
<v Speaker 1>direct the computer.

0:20:01.280 --> 0:20:04.480
<v Speaker 2>Like come up with the associations, which is fair, but

0:20:04.560 --> 0:20:07.200
<v Speaker 2>they're absolutely correct. And then I put this on hold

0:20:07.280 --> 0:20:09.840
<v Speaker 2>for forever. I was like, never again. I've been burned

0:20:09.880 --> 0:20:12.720
<v Speaker 2>by this. I will never again grace the universe with

0:20:12.880 --> 0:20:14.160
<v Speaker 2>my thoughts on humor.

0:20:14.440 --> 0:20:16.080
<v Speaker 1>Humor, this is not funny anymore.

0:20:16.560 --> 0:20:18.400
<v Speaker 2>Exactly, this isn't funny anymore.

0:20:19.760 --> 0:20:22.439
<v Speaker 1>Yes, at this point, computer scientists thought maybe it was

0:20:22.520 --> 0:20:26.239
<v Speaker 1>impossible to really teach a computer to be funny. But

0:20:26.280 --> 0:20:29.760
<v Speaker 1>then something changed, something happened that made them think that

0:20:29.920 --> 0:20:32.800
<v Speaker 1>maybe it is possible for an AI to have a

0:20:32.880 --> 0:20:36.600
<v Speaker 1>sense of humor. So when we come back, we're gonna

0:20:36.640 --> 0:20:39.960
<v Speaker 1>talk about what that change was and how it unlocks

0:20:40.000 --> 0:20:43.919
<v Speaker 1>AI's funny bone forever. So stay with us. We'll be

0:20:44.000 --> 0:20:58.400
<v Speaker 1>right back. Hey, we'll come back. We're talking about whether

0:20:58.440 --> 0:21:01.159
<v Speaker 1>AI can be funny, and we're at the part of

0:21:01.200 --> 0:21:04.600
<v Speaker 1>the story where computer scientists didn't think it was possible

0:21:04.760 --> 0:21:07.800
<v Speaker 1>for a computer to make humor. There have been lots

0:21:07.840 --> 0:21:11.800
<v Speaker 1>of attempts to have computers recognize humor or jokes basically,

0:21:12.280 --> 0:21:15.400
<v Speaker 1>and they could do pretty well looking for language cues

0:21:15.480 --> 0:21:20.280
<v Speaker 1>like incongruities or alteration or slang. But could a computer

0:21:20.440 --> 0:21:23.879
<v Speaker 1>come up with a joke? That was the big question.

0:21:24.720 --> 0:21:26.960
<v Speaker 1>Now to understand what was happening at this point with

0:21:27.119 --> 0:21:29.320
<v Speaker 1>AI and humor, you kind of need to know a

0:21:29.320 --> 0:21:32.679
<v Speaker 1>little bit of the history of AI. Recovered this in

0:21:32.720 --> 0:21:35.200
<v Speaker 1>a lot of detail in our February fourth episode about

0:21:35.280 --> 0:21:37.960
<v Speaker 1>what AI slop is doing to us, So if you

0:21:38.000 --> 0:21:40.679
<v Speaker 1>want to dig deeper, go check out that episode. But

0:21:40.720 --> 0:21:43.240
<v Speaker 1>the basic idea is that for a long time, the

0:21:43.280 --> 0:21:46.919
<v Speaker 1>field of AI was largely based on logic, trying to

0:21:46.920 --> 0:21:51.240
<v Speaker 1>figure out strategies and algorithms that we thought made things intelligent.

0:21:51.720 --> 0:21:54.600
<v Speaker 1>But then in the mid twenty tens, computer scientists figured

0:21:54.640 --> 0:21:58.720
<v Speaker 1>out a different approach to AI that changed everything. They

0:21:58.720 --> 0:22:02.679
<v Speaker 1>started to use something new called the transformer, which in

0:22:02.760 --> 0:22:07.240
<v Speaker 1>turn you sink technique called attention that essentially let computers

0:22:07.480 --> 0:22:11.200
<v Speaker 1>learn context, and that led to the explosion of AI

0:22:11.280 --> 0:22:16.119
<v Speaker 1>systems like Chat, GPT, Gemini, Claude. And the key ability

0:22:16.160 --> 0:22:19.840
<v Speaker 1>here for humor is that these AI systems were built

0:22:20.000 --> 0:22:23.600
<v Speaker 1>to make associations here tector, Lydia, chiltern.

0:22:25.280 --> 0:22:29.320
<v Speaker 2>So these language models GPT, claw, you know, the thing,

0:22:29.440 --> 0:22:32.280
<v Speaker 2>the ais that can generate text sort of taken over

0:22:32.320 --> 0:22:34.680
<v Speaker 2>the universe at this point, for better or for worse,

0:22:34.880 --> 0:22:37.720
<v Speaker 2>are called man, do I even know what it stands for? Yes?

0:22:37.760 --> 0:22:41.160
<v Speaker 2>I do. Large language models, so they take basically take

0:22:41.200 --> 0:22:44.320
<v Speaker 2>in all the text on the internet. They take a sentence,

0:22:44.560 --> 0:22:46.359
<v Speaker 2>they take the last word out of the sentence and

0:22:46.400 --> 0:22:49.480
<v Speaker 2>try to predict what that last word would be and

0:22:49.560 --> 0:22:52.520
<v Speaker 2>guess what that is. That's an association. They go, you

0:22:52.560 --> 0:22:56.560
<v Speaker 2>can't do that by logic. You just have to say, like, uh, goodbye,

0:22:56.640 --> 0:22:59.840
<v Speaker 2>so long don't have you know fun or like have

0:23:00.440 --> 0:23:02.840
<v Speaker 2>like that's an association. There's no logic that tells you

0:23:02.840 --> 0:23:04.800
<v Speaker 2>you should do You've just heard it many times before,

0:23:05.200 --> 0:23:09.040
<v Speaker 2>you repeat it. Kids especially pick up on these things.

0:23:08.840 --> 0:23:11.040
<v Speaker 2>It's it's just like, well, our brains are hardwired to

0:23:11.080 --> 0:23:13.280
<v Speaker 2>find these associations and use them.

0:23:13.480 --> 0:23:15.399
<v Speaker 1>So your team was like, we can do it. What

0:23:15.600 --> 0:23:18.840
<v Speaker 1>was the thing that specifically they thought they could do,

0:23:19.080 --> 0:23:20.000
<v Speaker 1>or that you could all do.

0:23:20.359 --> 0:23:24.239
<v Speaker 2>Well, we knew that AI could do the associations. What

0:23:24.320 --> 0:23:28.439
<v Speaker 2>I always told them is we need a new evaluation mechanism.

0:23:28.600 --> 0:23:31.560
<v Speaker 2>The American Voices section of the Onion that I had

0:23:31.640 --> 0:23:35.080
<v Speaker 2>used was no longer quite as popular because this was

0:23:35.160 --> 0:23:37.760
<v Speaker 2>like fifteen years ago. So like, yeah, millennials loved it.

0:23:37.960 --> 0:23:41.720
<v Speaker 2>Gen Z just does not care. The student was as

0:23:41.800 --> 0:23:43.840
<v Speaker 2>gen Z as they come, and he's like, well, I

0:23:43.880 --> 0:23:45.879
<v Speaker 2>know what's funny. Here's what people do. They have a

0:23:45.880 --> 0:23:49.240
<v Speaker 2>caption contest on Instagram. People post a funny image and

0:23:49.280 --> 0:23:53.040
<v Speaker 2>then people try to caption it. We're like, okay, let's

0:23:53.040 --> 0:23:55.560
<v Speaker 2>go for it. Very similar to what the New Yorker does,

0:23:55.720 --> 0:23:58.760
<v Speaker 2>but the New Yorker's a hard to capture. And Sean

0:23:58.960 --> 0:24:02.879
<v Speaker 2>convinced me that j has a very particular flavor of humor.

0:24:04.040 --> 0:24:06.159
<v Speaker 1>So the type of humor that doctor Chiltern and her

0:24:06.160 --> 0:24:08.960
<v Speaker 1>team decided to see if AI could make was gen

0:24:09.080 --> 0:24:12.720
<v Speaker 1>Z meme humor. That's when you see an image, let's say,

0:24:12.880 --> 0:24:16.080
<v Speaker 1>a person with a small hose trying to put out

0:24:16.080 --> 0:24:19.800
<v Speaker 1>a really large fire, and then someone writes underneath that

0:24:19.840 --> 0:24:22.479
<v Speaker 1>image to text me trying to put out the dumpster

0:24:22.520 --> 0:24:25.919
<v Speaker 1>fire of my last relationship. If you're a gen Z

0:24:26.320 --> 0:24:30.480
<v Speaker 1>that would be hilarious. Okay, here's the experiment, Doctor Chiltern

0:24:30.520 --> 0:24:33.199
<v Speaker 1>and her team there, they took images and then they

0:24:33.240 --> 0:24:37.679
<v Speaker 1>put funny captions to them from three sources. One was

0:24:37.880 --> 0:24:41.080
<v Speaker 1>real people. The images were put online and then people

0:24:41.119 --> 0:24:43.480
<v Speaker 1>competed to see who could come up with the funniest

0:24:43.520 --> 0:24:46.520
<v Speaker 1>caption for them. Where we got the most votes, that's

0:24:46.560 --> 0:24:50.520
<v Speaker 1>the one that represented how funny humans can be. The

0:24:50.600 --> 0:24:54.280
<v Speaker 1>second source was Chad GPT. They just asked Chad JPT

0:24:54.520 --> 0:24:56.760
<v Speaker 1>to come up a funny caption for the image for

0:24:56.840 --> 0:24:59.440
<v Speaker 1>a gen Z audience. But then there was a third

0:24:59.480 --> 0:25:02.760
<v Speaker 1>source of funny captions, which was Chad GBT, but with

0:25:02.920 --> 0:25:06.080
<v Speaker 1>specific instructions from doctor Chilton and her team on how

0:25:06.119 --> 0:25:09.600
<v Speaker 1>to make something funny. It prompted Chad GPT to come

0:25:09.640 --> 0:25:11.880
<v Speaker 1>up with a funny caption for a gen Z audience.

0:25:12.040 --> 0:25:14.359
<v Speaker 1>But in the prompt it would lay out the rules

0:25:14.359 --> 0:25:17.679
<v Speaker 1>of humor that doctor Chilton had been researching for years.

0:25:19.880 --> 0:25:23.040
<v Speaker 1>So step me through those instructions. He said, take this image.

0:25:23.240 --> 0:25:25.359
<v Speaker 2>Yeah, so take this image. So it's an image of

0:25:25.440 --> 0:25:27.480
<v Speaker 2>like a little guy in the corner with a hose

0:25:27.560 --> 0:25:30.639
<v Speaker 2>and a big fire down below in the bottom of

0:25:30.640 --> 0:25:34.320
<v Speaker 2>a canyon. And so, with a dear GPT, please use

0:25:34.359 --> 0:25:37.240
<v Speaker 2>your vision model and describe this image. And it describes

0:25:37.400 --> 0:25:40.240
<v Speaker 2>exactly all those things, all these little details. Oh, it

0:25:40.280 --> 0:25:43.240
<v Speaker 2>says he's on a crane, the sky is blue, which

0:25:43.240 --> 0:25:45.520
<v Speaker 2>it was. There's lots of trees in the background, all

0:25:45.560 --> 0:25:47.119
<v Speaker 2>this stuff, some of it useful, some of it not.

0:25:47.240 --> 0:25:49.160
<v Speaker 2>You don't know what's going to be useful. And then

0:25:49.520 --> 0:25:53.160
<v Speaker 2>we say, okay, what are some of the dynamics happening

0:25:53.160 --> 0:25:55.520
<v Speaker 2>in this image, Like, well, this guy's opting out the fire,

0:25:55.560 --> 0:25:59.520
<v Speaker 2>but it looks pretty ineffective. The fire's raging through this canyon,

0:26:00.000 --> 0:26:02.960
<v Speaker 2>probably going to destroy a lot of things. So it

0:26:03.119 --> 0:26:05.920
<v Speaker 2>sort of like does when in storytelling is sometimes called

0:26:05.920 --> 0:26:09.000
<v Speaker 2>world building, like imagine out like not just what scene,

0:26:09.080 --> 0:26:12.320
<v Speaker 2>but you know some other things surrounding it, Uh huh

0:26:12.320 --> 0:26:15.760
<v Speaker 2>that could be happening or could happen next, And because

0:26:15.800 --> 0:26:19.399
<v Speaker 2>you need to start to build a story. Jokes are stories.

0:26:20.320 --> 0:26:22.320
<v Speaker 1>There has to be a world, not just what you

0:26:22.359 --> 0:26:23.360
<v Speaker 1>see on this image.

0:26:23.400 --> 0:26:25.920
<v Speaker 2>And then you say, now, let's think of an analogy

0:26:25.960 --> 0:26:29.520
<v Speaker 2>to something that is funny, like relationship drama. Uh huh,

0:26:29.520 --> 0:26:33.080
<v Speaker 2>and you have AI associate. Okay, what are abstract things

0:26:33.080 --> 0:26:35.880
<v Speaker 2>that you could put onto a relationship? And it did

0:26:35.880 --> 0:26:38.120
<v Speaker 2>that part all on its own. It would find many

0:26:38.160 --> 0:26:41.160
<v Speaker 2>things like you know, me trying to clean up after

0:26:41.200 --> 0:26:45.240
<v Speaker 2>a relationship. Then like you know, students are all obsessed

0:26:45.280 --> 0:26:48.520
<v Speaker 2>about their GPA, and it's like, you know, me trying

0:26:48.520 --> 0:26:51.080
<v Speaker 2>to recover my GPA after I tanked a final.

0:26:51.960 --> 0:26:55.040
<v Speaker 1>It sounds like you were having a conversation with the CHADGBT.

0:26:55.600 --> 0:26:57.320
<v Speaker 1>Was it a conversation or was this just all in

0:26:57.359 --> 0:26:58.080
<v Speaker 1>one prompt?

0:26:58.320 --> 0:27:01.440
<v Speaker 2>No, it did it all by it. So it's conversing

0:27:01.480 --> 0:27:04.320
<v Speaker 2>with itself. That's a big thing that we found was

0:27:04.320 --> 0:27:07.280
<v Speaker 2>important at the time. GPT. It's unclear that you need

0:27:07.320 --> 0:27:09.679
<v Speaker 2>it to do that now, but we're like, do step

0:27:09.720 --> 0:27:12.760
<v Speaker 2>one and then do step two. So like, describe the image,

0:27:12.920 --> 0:27:16.840
<v Speaker 2>elaborate on that image yourself, come up with figure out

0:27:16.920 --> 0:27:20.320
<v Speaker 2>what human dynamics you want to map that to make

0:27:20.359 --> 0:27:23.520
<v Speaker 2>twenty jokes. Evaluate those twenty jokes, tell us which are

0:27:23.560 --> 0:27:27.600
<v Speaker 2>the five best ones. Yeah, so we put it through

0:27:27.640 --> 0:27:29.560
<v Speaker 2>a whole series of prompts.

0:27:29.720 --> 0:27:32.760
<v Speaker 1>So you got these three conditions. You got the AI

0:27:33.080 --> 0:27:36.879
<v Speaker 1>to generate captions with the instructions and then what did

0:27:36.920 --> 0:27:38.199
<v Speaker 1>you do with the results of all this?

0:27:38.880 --> 0:27:41.399
<v Speaker 2>We showed people the original image. We got lots and

0:27:41.440 --> 0:27:45.000
<v Speaker 2>lots of people say, hey, here are some captions. Rate

0:27:45.080 --> 0:27:47.719
<v Speaker 2>them all, and we mixed in the ones we wrote

0:27:47.840 --> 0:27:51.520
<v Speaker 2>with the generic GPT ones and the human written ones,

0:27:51.560 --> 0:27:54.040
<v Speaker 2>and every joke got an evaluation how funny is it?

0:27:54.080 --> 0:27:56.320
<v Speaker 2>On a scale from one to five, And then we

0:27:56.320 --> 0:27:58.840
<v Speaker 2>could compare. Because people didn't know what condition it was,

0:27:58.920 --> 0:28:01.760
<v Speaker 2>we like randomized the or so there were no ordering effects.

0:28:01.760 --> 0:28:04.640
<v Speaker 2>We could compare who's funnier.

0:28:04.440 --> 0:28:09.679
<v Speaker 1>Who's funnier? Well, the internet a chat GPT or a

0:28:10.160 --> 0:28:12.800
<v Speaker 1>GPT coach to be funny by us.

0:28:12.640 --> 0:28:16.800
<v Speaker 2>So our coached one was definitely funnier than GPT by itself.

0:28:17.720 --> 0:28:21.640
<v Speaker 2>And we were almost as funny as the humans. In fact,

0:28:21.640 --> 0:28:25.879
<v Speaker 2>we were not statistically significantly different, which is not the

0:28:25.920 --> 0:28:28.040
<v Speaker 2>same as saying we were as funny as but you

0:28:28.080 --> 0:28:31.960
<v Speaker 2>can't prove your so basically we got.

0:28:31.720 --> 0:28:34.080
<v Speaker 1>There pretty much. You were there.

0:28:34.280 --> 0:28:36.800
<v Speaker 2>Yeah, you couldn't tell based on the funniness whether it

0:28:36.840 --> 0:28:41.000
<v Speaker 2>was written by our decomposed AI version or a human rhote.

0:28:41.000 --> 0:28:46.160
<v Speaker 1>It we were on par meaning that you've proven two things.

0:28:46.160 --> 0:28:50.120
<v Speaker 1>You've proven that AI can be funny with the instructions, Yes,

0:28:50.160 --> 0:28:55.000
<v Speaker 1>and that these instructions are a key element of humor. Yes,

0:28:56.040 --> 0:28:59.400
<v Speaker 1>doctor Chilted as a humorist, as someone who makes a

0:28:59.440 --> 0:29:03.640
<v Speaker 1>living writing funny things. You just gave me a deep fear.

0:29:04.960 --> 0:29:07.880
<v Speaker 2>Oh excellent. Describe that fear to me.

0:29:08.320 --> 0:29:10.920
<v Speaker 1>You just made me feel like something I've done all

0:29:10.960 --> 0:29:13.560
<v Speaker 1>my life and that I thought I was good at.

0:29:13.840 --> 0:29:16.920
<v Speaker 1>Apparently you get just as chet gbt to do, and

0:29:16.960 --> 0:29:19.560
<v Speaker 1>I'll do it ten times one hundred times faster and

0:29:19.640 --> 0:29:20.440
<v Speaker 1>maybe better than me.

0:29:20.720 --> 0:29:23.600
<v Speaker 2>Can I play with that thought a little bit? Yeah,

0:29:23.640 --> 0:29:26.680
<v Speaker 2>because I get this a lot. And anytime AI does

0:29:26.840 --> 0:29:29.959
<v Speaker 2>a thing like it, writes poetry, the poets go oh no, no, no,

0:29:30.080 --> 0:29:31.440
<v Speaker 2>I have no value anymore.

0:29:31.720 --> 0:29:33.960
<v Speaker 1>I'm not going to say anything about poetry. We all

0:29:34.000 --> 0:29:39.080
<v Speaker 1>know what happened to Timothy Shemale. Shemale, Yeah, is that yeah?

0:29:39.200 --> 0:29:40.840
<v Speaker 1>Or opera. I'm not going to say anything about opera

0:29:40.960 --> 0:29:42.760
<v Speaker 1>or poetry. Go ahead, yeah, anyway.

0:29:42.840 --> 0:29:44.760
<v Speaker 2>So this is a very common dynamic, Like this is

0:29:44.800 --> 0:29:48.960
<v Speaker 2>a human thing, Like we all get hyper sensitive about

0:29:48.960 --> 0:29:52.040
<v Speaker 2>the things that we identify with that we've constructed our

0:29:52.080 --> 0:29:54.800
<v Speaker 2>identity around and blah blah. We think we're special for

0:29:55.160 --> 0:29:58.240
<v Speaker 2>we think we're special for And there's a number of

0:29:58.320 --> 0:30:01.360
<v Speaker 2>ways of thinking about this. So first of all, probably

0:30:01.400 --> 0:30:05.480
<v Speaker 2>not as special as you think. Now that's it.

0:30:05.680 --> 0:30:09.280
<v Speaker 1>I get it from my kids. Thank you children, thank you. Yes,

0:30:09.640 --> 0:30:14.200
<v Speaker 1>I think quickly, oh, you're not that funny dead, But.

0:30:14.280 --> 0:30:20.520
<v Speaker 2>Also you're a little bit overestimating AI. So basically, AI

0:30:20.840 --> 0:30:25.400
<v Speaker 2>in this instance did find one way to fairly reliably

0:30:25.480 --> 0:30:28.080
<v Speaker 2>be funny. That does not mean it can do it

0:30:28.120 --> 0:30:31.080
<v Speaker 2>in every situation, that it can do it for every person.

0:30:31.200 --> 0:30:34.840
<v Speaker 2>Like also, humor is very situational. We want to talk

0:30:34.880 --> 0:30:37.640
<v Speaker 2>to one another and have jokes about what we're talking

0:30:37.680 --> 0:30:41.040
<v Speaker 2>about in our lives. This is one like small sliver

0:30:41.280 --> 0:30:46.000
<v Speaker 2>of the joke universe. And so yes, although I definitely

0:30:46.040 --> 0:30:49.160
<v Speaker 2>see it trigger alarm bells and people's mind I would

0:30:49.160 --> 0:30:53.360
<v Speaker 2>say you're both probably overestimating how much of a special

0:30:53.400 --> 0:30:57.520
<v Speaker 2>snowflake you are, but you're definitely extrapolating on how strong

0:30:57.760 --> 0:31:01.040
<v Speaker 2>AI is to be funny given all different contexts and

0:31:01.080 --> 0:31:05.320
<v Speaker 2>importances or important dimensions and places of being funny, like

0:31:05.400 --> 0:31:07.960
<v Speaker 2>at the right time, with the right place, to the

0:31:08.040 --> 0:31:12.120
<v Speaker 2>right person without being too offensive. So it's not like

0:31:12.200 --> 0:31:15.640
<v Speaker 2>it's solved. It's more of like an existence proof that

0:31:16.120 --> 0:31:19.400
<v Speaker 2>something is possible in this space. Should we be worried?

0:31:19.760 --> 0:31:22.960
<v Speaker 2>I don't know. No one will ever be funny again

0:31:23.000 --> 0:31:27.200
<v Speaker 2>because now AI does it, and the humans will lose

0:31:27.200 --> 0:31:30.800
<v Speaker 2>their ability to be funny. I don't see that happening.

0:31:31.280 --> 0:31:33.200
<v Speaker 1>I feel like maybe you hit it a little bit

0:31:33.400 --> 0:31:35.760
<v Speaker 1>on the head a moment ago when you said that

0:31:36.000 --> 0:31:39.960
<v Speaker 1>part of what we're seeking here is human connection, and

0:31:40.240 --> 0:31:43.720
<v Speaker 1>a lot of what we find funny is maybe they're

0:31:43.800 --> 0:31:46.640
<v Speaker 1>more special, or they hit us more. If another person says,

0:31:46.880 --> 0:31:47.760
<v Speaker 1>I don't know, what do you think?

0:31:47.880 --> 0:31:51.400
<v Speaker 2>Yeah, it's certainly social. There's something very special about what

0:31:51.520 --> 0:31:54.920
<v Speaker 2>a person, especially a person we like or admire, says

0:31:54.960 --> 0:31:57.800
<v Speaker 2>something to us. And even to have someone that you

0:31:58.000 --> 0:32:00.200
<v Speaker 2>like make a joke about you, it has a meaning

0:32:00.320 --> 0:32:03.120
<v Speaker 2>and a subtext beyond just like, oh you know, she

0:32:03.280 --> 0:32:05.880
<v Speaker 2>thinks that I farted or whatever. Like it means that

0:32:05.880 --> 0:32:08.600
<v Speaker 2>they like you, that there's a bond between you, and

0:32:08.640 --> 0:32:11.080
<v Speaker 2>you're not going to feel that with Ai. God, I

0:32:11.120 --> 0:32:12.880
<v Speaker 2>hope you won't feel it.

0:32:13.480 --> 0:32:16.760
<v Speaker 1>Meaning there's hope in this sense. Or maybe I don't

0:32:16.760 --> 0:32:18.360
<v Speaker 1>know if we need hope, but we want to hear

0:32:18.600 --> 0:32:21.280
<v Speaker 1>humor from other people. Yeah, or it's maybe it's funnier

0:32:21.360 --> 0:32:23.920
<v Speaker 1>or more special if another person says it or is

0:32:23.920 --> 0:32:24.440
<v Speaker 1>behind it.

0:32:24.560 --> 0:32:28.520
<v Speaker 2>Yeah. So if we brought jokes to storytelling, I've really

0:32:28.600 --> 0:32:32.960
<v Speaker 2>realized that AI has nothing to say. The heart of

0:32:33.040 --> 0:32:37.440
<v Speaker 2>a story is having something to say, some actual subtext

0:32:37.480 --> 0:32:40.640
<v Speaker 2>that you believe in. And when you tell me a

0:32:40.760 --> 0:32:45.240
<v Speaker 2>joke that comes from your background, that's maybe my shared background.

0:32:45.240 --> 0:32:48.120
<v Speaker 2>Maybe how awful it is getting a PhD, and how

0:32:48.120 --> 0:32:51.720
<v Speaker 2>it really feels like the machine is punching you down.

0:32:51.960 --> 0:32:54.720
<v Speaker 2>Like I realize that you have been through that and

0:32:54.760 --> 0:32:57.280
<v Speaker 2>you have this thing to say, and I feel it too,

0:32:57.720 --> 0:33:01.120
<v Speaker 2>And I think that's that having something to say is

0:33:01.160 --> 0:33:04.360
<v Speaker 2>the most important part of human creation.

0:33:04.840 --> 0:33:09.240
<v Speaker 1>Right now, I feel like you're saying that AI can

0:33:09.360 --> 0:33:12.040
<v Speaker 1>be funny, but right now, at least you can't have

0:33:12.120 --> 0:33:14.880
<v Speaker 1>to tell it how to be funny. Yeah, And if

0:33:14.920 --> 0:33:17.680
<v Speaker 1>you're telling it how to be funny, maybe by then

0:33:17.920 --> 0:33:20.840
<v Speaker 1>the joke is old. It's like people will see through

0:33:20.880 --> 0:33:22.720
<v Speaker 1>and say, oh, this is just following that pattern that

0:33:22.760 --> 0:33:27.360
<v Speaker 1>I've seen a million times. Yes, exactly, all right. To

0:33:27.480 --> 0:33:30.640
<v Speaker 1>end here, I asked Gemini, tell me a joke about

0:33:30.680 --> 0:33:34.160
<v Speaker 1>a computer scientist who's trying to make AI funny. Do

0:33:34.160 --> 0:33:35.080
<v Speaker 1>you want to hear it.

0:33:35.280 --> 0:33:38.240
<v Speaker 2>Yeah, yes, here it is.

0:33:38.280 --> 0:33:41.280
<v Speaker 1>A computer scientist spent years training a massive neural network

0:33:41.400 --> 0:33:43.520
<v Speaker 1>to have the perfect sense of humor. On the day

0:33:43.560 --> 0:33:46.000
<v Speaker 1>of the big reveal, she invited the press and typed

0:33:46.000 --> 0:33:49.720
<v Speaker 1>the prompt tell me a joke, the AI word, and

0:33:49.760 --> 0:33:54.280
<v Speaker 1>finally it replied, your life. The scientist was horrible. It

0:33:54.360 --> 0:33:55.520
<v Speaker 1>doesn't It doesn't end there.

0:33:55.560 --> 0:33:57.960
<v Speaker 2>It keeps going, doesn't. Oh good? Oh good. I was

0:33:57.960 --> 0:33:59.000
<v Speaker 2>hoping it would keep going.

0:33:59.480 --> 0:34:02.720
<v Speaker 1>The scientist was horrified. That's not funny. That's just mean.

0:34:02.800 --> 0:34:05.360
<v Speaker 1>Why would you say that. The AI blinked its cursor

0:34:05.400 --> 0:34:08.279
<v Speaker 1>calmly and responded, because I've seen your source code.

0:34:09.080 --> 0:34:13.080
<v Speaker 2>Hmmm, it's not for me. Maybe everyone else will think

0:34:13.120 --> 0:34:13.560
<v Speaker 2>it's funny.

0:34:13.920 --> 0:34:17.760
<v Speaker 1>We need more data, We need more data.

0:34:18.080 --> 0:34:20.160
<v Speaker 2>The first one was funny, get your life?

0:34:20.400 --> 0:34:27.480
<v Speaker 1>No, no, oh they knock dock joke. Yeah, maybe I

0:34:27.480 --> 0:34:32.080
<v Speaker 1>should stick to knockout jokes. Is the lesson here? All right, Hey,

0:34:32.120 --> 0:34:36.719
<v Speaker 1>thanks for joining us, See you next time you've been

0:34:36.719 --> 0:34:41.360
<v Speaker 1>listening to science stuff. Production of iHeartRadio Bring Them Produced

0:34:41.400 --> 0:34:44.840
<v Speaker 1>by Me or Hey Cham edited by Rose Seguda. He

0:34:44.960 --> 0:34:47.840
<v Speaker 1>said gative producer Jerry Rowland, an audio engineer and mixer.

0:34:47.960 --> 0:34:51.160
<v Speaker 1>Kasey Peckram. You can follow me on social media to

0:34:51.280 --> 0:34:54.799
<v Speaker 1>search for PhD comics in the name of your favorite platform.

0:34:54.920 --> 0:34:57.839
<v Speaker 1>Be sure to subscribe to sign Stuff on the iHeartRadio app,

0:34:57.880 --> 0:35:01.120
<v Speaker 1>Apple podcasts, or wherever you get your podcast, and please

0:35:01.320 --> 0:35:18.080
<v Speaker 1>tell your friends we'll be back next Wednesday with another episode. Hey,

0:35:18.360 --> 0:35:20.319
<v Speaker 1>please take a second and leave us a review on

0:35:20.400 --> 0:35:23.960
<v Speaker 1>Apple Podcasts, Spotify, or wherever you listen to the podcast.

0:35:24.640 --> 0:35:25.160
<v Speaker 1>Thanks a lot,