WEBVTT - Flatus Ex Machina, Part 1

0:00:03.040 --> 0:00:06.119
<v Speaker 1>Welcome to Stuff to Blow Your Mind from how Stuffworks

0:00:06.120 --> 0:00:14.120
<v Speaker 1>dot com. Hey are you welcome to Stuff to Blow

0:00:14.160 --> 0:00:17.040
<v Speaker 1>your Mind? My name is Robert Lamb and I'm Joe McCormick,

0:00:17.079 --> 0:00:18.919
<v Speaker 1>and today we're gonna be talking about one of my

0:00:19.000 --> 0:00:23.960
<v Speaker 1>favorite comedy subjects, What's funny about the way that machines fail?

0:00:24.600 --> 0:00:26.040
<v Speaker 1>And here's just a heads up. This is gonna be

0:00:26.079 --> 0:00:28.800
<v Speaker 1>a two part episode because we ended up going super long.

0:00:28.920 --> 0:00:31.480
<v Speaker 1>Like the machines, we can't stop. So I want to

0:00:31.520 --> 0:00:35.880
<v Speaker 1>start with a particular example, probably my favorite funny thing

0:00:35.960 --> 0:00:40.919
<v Speaker 1>on the Internet these days, the hilarious almost successes of

0:00:41.240 --> 0:00:47.120
<v Speaker 1>artificial intelligence trying to generate examples of human language almost

0:00:47.200 --> 0:00:50.479
<v Speaker 1>but not quite human. Yes, I don't know why this

0:00:50.640 --> 0:00:53.559
<v Speaker 1>does it for me, but aside from Highlander to The Quickening,

0:00:53.560 --> 0:00:57.880
<v Speaker 1>pretty much nothing makes me laugh harder than language generated

0:00:57.880 --> 0:01:01.280
<v Speaker 1>by artificial neural networks and machine learning. And we'll explain

0:01:01.320 --> 0:01:03.760
<v Speaker 1>a little bit more about exactly how this works in

0:01:03.800 --> 0:01:06.200
<v Speaker 1>a minute, but first I just thought we should look

0:01:06.200 --> 0:01:08.560
<v Speaker 1>at a few examples of what this is like. If

0:01:08.560 --> 0:01:11.600
<v Speaker 1>you're on the Internet, you've probably encountered this at some point,

0:01:12.400 --> 0:01:15.679
<v Speaker 1>because it's become popular in the past few years. Especially

0:01:15.760 --> 0:01:18.280
<v Speaker 1>for its comedic value. If you're not on the internet, boy,

0:01:18.280 --> 0:01:20.960
<v Speaker 1>are you in for a treat um by far? I

0:01:20.959 --> 0:01:23.400
<v Speaker 1>would say the best of this stuff I've come across

0:01:23.560 --> 0:01:27.120
<v Speaker 1>is traceable back to the blog of a person named

0:01:27.240 --> 0:01:30.200
<v Speaker 1>Janelle Shane, who I think her day job as she

0:01:30.240 --> 0:01:33.920
<v Speaker 1>works in industrial researching optics, but as a hobby, she

0:01:34.040 --> 0:01:38.399
<v Speaker 1>trains neural networks with text based data uh, text based

0:01:38.480 --> 0:01:43.240
<v Speaker 1>data sets to spit out these amazing simulations of types

0:01:43.319 --> 0:01:46.280
<v Speaker 1>of human language, and so they'll they'll be in categories

0:01:46.319 --> 0:01:50.280
<v Speaker 1>like it. She gets an AI program to write recipes

0:01:50.320 --> 0:01:53.280
<v Speaker 1>for foods or to come up with the names of

0:01:53.440 --> 0:01:56.080
<v Speaker 1>paint colors or something. And the way, of course you

0:01:56.080 --> 0:01:58.040
<v Speaker 1>would do this is, and we'll explain more of the

0:01:58.080 --> 0:02:00.800
<v Speaker 1>details in a bit, but you'd train it on existing

0:02:00.960 --> 0:02:04.000
<v Speaker 1>names of paint colors, or you'd train it on existing

0:02:04.040 --> 0:02:06.440
<v Speaker 1>recipes of food. Right. So the end result here is

0:02:06.440 --> 0:02:10.079
<v Speaker 1>you essentially have a machine trying to human but not

0:02:10.200 --> 0:02:13.840
<v Speaker 1>quite pulling it off, but but but doing so in

0:02:14.000 --> 0:02:17.520
<v Speaker 1>such hilarious fashion. Yeah, so you should absolutely look up

0:02:17.600 --> 0:02:21.000
<v Speaker 1>Janelle Shane's blog. It's called AI Weirdness dot com. It's

0:02:21.080 --> 0:02:23.840
<v Speaker 1>reliably such a source of joy. But I want to start.

0:02:23.880 --> 0:02:25.720
<v Speaker 1>In fact, I will say that if you were if

0:02:25.760 --> 0:02:27.640
<v Speaker 1>you're scratching your head and you're thinking, oh, didn't I

0:02:27.680 --> 0:02:31.920
<v Speaker 1>see something hilarious in this vein recently, there's there's a

0:02:31.919 --> 0:02:35.280
<v Speaker 1>there's a high probability that it originated from AI weirdness

0:02:35.360 --> 0:02:38.600
<v Speaker 1>dot com. Yes, that that blog is just awesome, but

0:02:38.639 --> 0:02:40.280
<v Speaker 1>I wanted to look at a few examples of what

0:02:40.320 --> 0:02:42.440
<v Speaker 1>this is like. So one thing is consider the work,

0:02:42.680 --> 0:02:45.720
<v Speaker 1>uh that that Jennell Shane did training in neural network

0:02:45.800 --> 0:02:48.200
<v Speaker 1>to come up with names for D and D spells.

0:02:48.960 --> 0:02:51.560
<v Speaker 1>So you take the Dungeons and Dragons manual and you'd

0:02:51.600 --> 0:02:54.359
<v Speaker 1>feed in all the actual names of spells to the

0:02:54.480 --> 0:02:56.400
<v Speaker 1>neural network, so it gets a sense of what these

0:02:56.400 --> 0:02:58.760
<v Speaker 1>things are like, and then it tries to come up

0:02:58.840 --> 0:03:01.440
<v Speaker 1>with similar types of names on its own. Now, to

0:03:01.520 --> 0:03:03.760
<v Speaker 1>give everybody just a quick idea first of what actual

0:03:03.840 --> 0:03:08.280
<v Speaker 1>Dungeons and Dragon spells are named. You have everything from

0:03:08.480 --> 0:03:13.919
<v Speaker 1>Magic Missile or Crown of Madness to Evard's Black Tentacles

0:03:14.080 --> 0:03:16.320
<v Speaker 1>or anytime there's a wizard name in there, you know

0:03:16.360 --> 0:03:19.320
<v Speaker 1>you're in for some good stuff. Um, well, there's you know,

0:03:19.320 --> 0:03:22.799
<v Speaker 1>stuff like Glyph of Warding or what's the what's the

0:03:22.840 --> 0:03:25.440
<v Speaker 1>one I'm trying to think of? Oh, stuff like Leo

0:03:25.520 --> 0:03:31.119
<v Speaker 1>Moon's Secret Chest or Leoman's Tiny Hut. That's another great one. Well,

0:03:31.200 --> 0:03:33.280
<v Speaker 1>so here are the ones that it came up with.

0:03:33.560 --> 0:03:37.920
<v Speaker 1>How about selections like Mister of Light, Confusing, Storm of

0:03:37.960 --> 0:03:44.520
<v Speaker 1>the Giffling, Song of Goom, song of the Darn, Ward

0:03:44.640 --> 0:03:49.000
<v Speaker 1>of Snay to the pooda Beast, primal rear. You've got

0:03:49.000 --> 0:03:51.840
<v Speaker 1>to watch out for primal rear. Someone's storm Bear. Now

0:03:51.880 --> 0:03:54.760
<v Speaker 1>that one sounds legitimate, because that's one of the beauties

0:03:54.840 --> 0:03:58.400
<v Speaker 1>of these exercises is is when there's one that either

0:03:58.640 --> 0:04:02.080
<v Speaker 1>almost works or actually does work. Because I think some

0:04:02.240 --> 0:04:04.800
<v Speaker 1>in storm Bear, I can I can easily imagine I

0:04:04.800 --> 0:04:09.080
<v Speaker 1>can describe the storm Bear blasting out of the portal

0:04:09.480 --> 0:04:12.280
<v Speaker 1>and rushing into combat on behalf of your you know,

0:04:12.320 --> 0:04:16.360
<v Speaker 1>your your storm mage. Yeah, it's almost kind of effortlessly evocative.

0:04:17.240 --> 0:04:19.880
<v Speaker 1>A divine boom. How about that? That one sounds pretty

0:04:19.880 --> 0:04:23.239
<v Speaker 1>good too. Soul of the Bill. Now now we're flying

0:04:23.279 --> 0:04:27.760
<v Speaker 1>back down the hill. Farc mate about Charm of the CODs,

0:04:30.279 --> 0:04:33.520
<v Speaker 1>Death of the Sun. Okay, that that's that would have

0:04:33.560 --> 0:04:35.720
<v Speaker 1>to be a high level spell. But okay, okay, three

0:04:35.720 --> 0:04:39.880
<v Speaker 1>more greater flick. Okay, that's that's just a can trip there,

0:04:39.960 --> 0:04:42.760
<v Speaker 1>that's just a flick, a magical flick of the ear

0:04:43.400 --> 0:04:48.920
<v Speaker 1>curse clam. I like it. Daving Fire's confusing. Now. While

0:04:49.000 --> 0:04:52.359
<v Speaker 1>these phrases I think are mostly funny on their own,

0:04:52.440 --> 0:04:55.039
<v Speaker 1>I think they're probably even funnier if you're an actual

0:04:55.120 --> 0:04:57.440
<v Speaker 1>D and D player, because you not only get the

0:04:57.440 --> 0:05:00.480
<v Speaker 1>pleasure of the nonsense words and the you know, the

0:05:01.240 --> 0:05:04.360
<v Speaker 1>syllables that seem out of place in their context, like

0:05:04.880 --> 0:05:07.120
<v Speaker 1>Dave does not really seem to go very well with

0:05:07.200 --> 0:05:10.320
<v Speaker 1>some kind of magical fire spell, But if you actually

0:05:10.360 --> 0:05:12.599
<v Speaker 1>play D and D you probably also get some humor

0:05:12.680 --> 0:05:17.040
<v Speaker 1>from just like seeing the little resonances that these spells

0:05:17.080 --> 0:05:20.560
<v Speaker 1>have with actual spells that you would recognize. I now

0:05:20.680 --> 0:05:23.360
<v Speaker 1>want to run a gaming session where there's some sort

0:05:23.400 --> 0:05:27.080
<v Speaker 1>of warp effect in place where suddenly all the magic

0:05:27.200 --> 0:05:31.520
<v Speaker 1>users are forced to use of spells from this list,

0:05:31.960 --> 0:05:33.960
<v Speaker 1>and they don't know what they're going to exactly do

0:05:34.080 --> 0:05:37.200
<v Speaker 1>until they cast them. Even better, what if they had

0:05:37.240 --> 0:05:40.520
<v Speaker 1>to What if your character suddenly had amnesia and then

0:05:40.600 --> 0:05:43.480
<v Speaker 1>had to act as if they had bios that were

0:05:43.520 --> 0:05:46.480
<v Speaker 1>also generated by an artificial neural network. That's right, because

0:05:46.480 --> 0:05:49.480
<v Speaker 1>AI weirdness dot Com also has a wonderful piece titled

0:05:49.520 --> 0:05:52.880
<v Speaker 1>D and D Character bios now making slightly more sense.

0:05:53.000 --> 0:05:55.560
<v Speaker 1>In fact, I would say this post from I think

0:05:55.560 --> 0:05:57.440
<v Speaker 1>this was last week or something we were looking at.

0:05:57.480 --> 0:06:00.000
<v Speaker 1>I think this inspired me to want to do this episode.

0:06:00.160 --> 0:06:03.200
<v Speaker 1>We should read a couple of these. I'll read the

0:06:03.200 --> 0:06:08.480
<v Speaker 1>first one here. Quote. Frick found his old family's fortune

0:06:08.760 --> 0:06:12.119
<v Speaker 1>and his curiosity, and he went to a small city

0:06:12.200 --> 0:06:15.680
<v Speaker 1>to see if he could find a work in the goldfish.

0:06:15.760 --> 0:06:19.520
<v Speaker 1>He heard stories of a goldfish, a goldfish, a sea monster,

0:06:19.640 --> 0:06:22.280
<v Speaker 1>and a silver fish, a sea monster, and a ship

0:06:22.360 --> 0:06:25.520
<v Speaker 1>that was a ship of exploration. The ship was full

0:06:25.600 --> 0:06:30.320
<v Speaker 1>of fish and evil some treasure, but it was not

0:06:30.440 --> 0:06:33.360
<v Speaker 1>to be When Frick found the ship. He rushed back

0:06:33.960 --> 0:06:38.080
<v Speaker 1>and found the ship full of treasure and full of fish.

0:06:38.120 --> 0:06:41.800
<v Speaker 1>He wanted to be a pirate and fight it. Now.

0:06:41.800 --> 0:06:43.719
<v Speaker 1>I like that because there's a there's a lot of

0:06:43.760 --> 0:06:47.440
<v Speaker 1>silliness in there, but it does have the basic shape

0:06:47.480 --> 0:06:50.040
<v Speaker 1>of a bio and and and and and it's weirdness,

0:06:50.040 --> 0:06:54.360
<v Speaker 1>and it's the stuff that is more nonsensical actually feels

0:06:54.400 --> 0:06:58.480
<v Speaker 1>suitably magical and fantastic. You know that there's this this

0:06:58.640 --> 0:07:00.880
<v Speaker 1>fish that's not a fish, that's also a ship that's

0:07:00.880 --> 0:07:03.640
<v Speaker 1>full of treasure and fish, and he's going to become

0:07:03.640 --> 0:07:06.479
<v Speaker 1>a pirate and fighted, and so I think you could

0:07:06.560 --> 0:07:08.960
<v Speaker 1>run with that. This is one of the fascinating things

0:07:08.960 --> 0:07:12.520
<v Speaker 1>about neural net generated text is that it often has

0:07:12.680 --> 0:07:16.560
<v Speaker 1>the format of what you're going for correct, It just

0:07:16.640 --> 0:07:20.800
<v Speaker 1>doesn't have the sense of it correct like it will

0:07:20.960 --> 0:07:24.200
<v Speaker 1>esthetically and in shape be very much like what you're

0:07:24.200 --> 0:07:28.000
<v Speaker 1>looking for, but keywords do not make any sense at all.

0:07:28.520 --> 0:07:30.800
<v Speaker 1>Then again, another one that I thought was kind of

0:07:30.840 --> 0:07:34.240
<v Speaker 1>interesting in what it showed about common themes in in

0:07:34.320 --> 0:07:36.720
<v Speaker 1>D and D character bios, or I guess maybe fantasy

0:07:36.760 --> 0:07:41.000
<v Speaker 1>more generally, was one bio that included the lines um

0:07:41.040 --> 0:07:43.559
<v Speaker 1>the orc Warlock was captured and killed by a group

0:07:43.600 --> 0:07:46.120
<v Speaker 1>of Orcs. He was imprisoned and forced to work for

0:07:46.160 --> 0:07:48.960
<v Speaker 1>a giant tome. He was imprisoned and imprisoned for a

0:07:48.960 --> 0:07:51.480
<v Speaker 1>while until he was rescued by a group of adventurers.

0:07:51.720 --> 0:07:53.880
<v Speaker 1>He was imprisoned and imprisoned for a while until he

0:07:53.920 --> 0:07:56.080
<v Speaker 1>was rescued by a group of adventurers who were looking

0:07:56.120 --> 0:07:59.800
<v Speaker 1>for a group of adventurers to help eradicate the orchest ferocity. See,

0:07:59.800 --> 0:08:02.040
<v Speaker 1>now that's wonderful. I actually really like that when it's

0:08:02.040 --> 0:08:05.080
<v Speaker 1>a little you know, nonsensical. But uh, first of all,

0:08:05.160 --> 0:08:08.440
<v Speaker 1>you know, being an orc is hard. It's it's it's

0:08:08.440 --> 0:08:13.080
<v Speaker 1>a warlike world for the orc. But but then the

0:08:13.120 --> 0:08:15.559
<v Speaker 1>idea that he's working for a giant tone, the idea

0:08:15.600 --> 0:08:18.600
<v Speaker 1>that there's like a magical book employing him, and the

0:08:19.200 --> 0:08:23.120
<v Speaker 1>repetitive imprisonment. Yeah, it's a hard knock life. You're always

0:08:23.160 --> 0:08:27.480
<v Speaker 1>getting imprisoned and then escaping and encountering a band of adventurers. Yeah,

0:08:27.640 --> 0:08:30.160
<v Speaker 1>and working for weird magic books. So that one, that

0:08:30.200 --> 0:08:32.320
<v Speaker 1>one works, I think. Of course. Another great one on

0:08:32.360 --> 0:08:35.199
<v Speaker 1>this page is an injury that just seems to devolve

0:08:35.320 --> 0:08:39.200
<v Speaker 1>into endless like dozens of repetitions of variations of the

0:08:39.240 --> 0:08:43.080
<v Speaker 1>phrase big cat. Yes, little did he know. During the

0:08:43.080 --> 0:08:47.080
<v Speaker 1>monastery's course of time, when the monastery's training and growth

0:08:47.160 --> 0:08:50.200
<v Speaker 1>was complete, his mother told big cat, big cat and

0:08:50.280 --> 0:08:52.640
<v Speaker 1>big cat to drive their own path and test big

0:08:52.679 --> 0:08:55.560
<v Speaker 1>cat with big cats, Army of big cats, Army of

0:08:55.600 --> 0:08:58.760
<v Speaker 1>big cats and big cats and big cats. Big get

0:08:58.880 --> 0:09:01.480
<v Speaker 1>is big cat, big cat and goes on, yes something

0:09:01.679 --> 0:09:06.480
<v Speaker 1>for for many many lines. I'm trying to picture big cat. Uh,

0:09:06.800 --> 0:09:09.480
<v Speaker 1>though I can't. I don't actually go to like tiger

0:09:09.600 --> 0:09:12.680
<v Speaker 1>or lie in there. I think more of a kind

0:09:12.720 --> 0:09:16.640
<v Speaker 1>of mental power great basilisk type of fat house cat.

0:09:16.920 --> 0:09:18.800
<v Speaker 1>I think that would work. I also can't help but

0:09:18.880 --> 0:09:22.160
<v Speaker 1>think of cheshire cats, and of course cat Bus from

0:09:22.160 --> 0:09:25.360
<v Speaker 1>Totoro is being sort of in the vein of big Cat,

0:09:25.400 --> 0:09:30.240
<v Speaker 1>Big Cat, big Cat. Um. How about a neural networks

0:09:30.280 --> 0:09:33.079
<v Speaker 1>trained on a corpus of more than forty three thousand

0:09:33.200 --> 0:09:37.600
<v Speaker 1>questions and answer style jokes. I'll just read one of them.

0:09:37.679 --> 0:09:39.880
<v Speaker 1>Why do you call a pastor a cross the road?

0:09:40.600 --> 0:09:47.880
<v Speaker 1>He take the chicken? Yeah. Another great one from AI

0:09:47.920 --> 0:09:52.480
<v Speaker 1>Weirdness dot Com was a neural network designs Halloween costumes. Okay,

0:09:52.480 --> 0:09:54.840
<v Speaker 1>so you just feeded a bunch of Halloween costume names. Yeah,

0:09:54.880 --> 0:09:57.079
<v Speaker 1>and this one, this one was tremendous fun. I remember

0:09:57.080 --> 0:09:59.000
<v Speaker 1>when it first came. I just made me laugh so hard.

0:09:59.600 --> 0:10:07.199
<v Speaker 1>Highlight It's include uh, Saxy Pumpkins, Um, Disco Monster, Spartan Gandalf.

0:10:07.200 --> 0:10:10.480
<v Speaker 1>I really like that one. Starfleet Shark. Yeah, that's good. Um,

0:10:11.120 --> 0:10:15.120
<v Speaker 1>let's see Martian Devil. That was too believable. Panda clam though,

0:10:15.880 --> 0:10:20.760
<v Speaker 1>that's that's pretty good shark cow. That's very interesting given

0:10:21.240 --> 0:10:24.120
<v Speaker 1>some of our past discussions on was it a shark

0:10:24.120 --> 0:10:25.960
<v Speaker 1>cow that we were discussing. I believe it was. Oh,

0:10:26.080 --> 0:10:29.640
<v Speaker 1>that was That's Daniel Dennett's thought experiment exposing what he

0:10:29.679 --> 0:10:34.439
<v Speaker 1>believes to be flaws in Donald David's Donald Davidson's Swampman

0:10:34.920 --> 0:10:37.880
<v Speaker 1>thought experiment. It was the cow shark where he says,

0:10:38.520 --> 0:10:41.240
<v Speaker 1>is it actually meaningful if you posit a cow shark

0:10:41.360 --> 0:10:43.520
<v Speaker 1>to ask whether it's a cow or a shark? A

0:10:43.640 --> 0:10:48.480
<v Speaker 1>little accidental thought experiment here from this list. Uh, you'll

0:10:48.520 --> 0:10:52.000
<v Speaker 1>find less thought experimentation in such entries though, as Snape

0:10:52.080 --> 0:11:02.840
<v Speaker 1>scarce snape scarecrow, or or how about Lee garbage Lady garbage.

0:11:02.960 --> 0:11:05.000
<v Speaker 1>Lady garbage is good. So yeah, there's there are a

0:11:05.040 --> 0:11:09.600
<v Speaker 1>lot of great, great entries on that list. Okay, one

0:11:09.679 --> 0:11:12.280
<v Speaker 1>last thing from General Shane's work, I just must mention

0:11:12.440 --> 0:11:15.520
<v Speaker 1>some some titles for recipes that she wrote a script

0:11:15.520 --> 0:11:19.760
<v Speaker 1>to come up with. Um. These include chocolate pickle sauce,

0:11:19.920 --> 0:11:28.600
<v Speaker 1>whole chicken cookies, salmon beef style chicken bottom, artichoke, gelatin dogs,

0:11:28.640 --> 0:11:34.560
<v Speaker 1>and crock pot cold water. Where any of these attempted

0:11:34.640 --> 0:11:38.080
<v Speaker 1>these recipes well, actually at one point she so those

0:11:38.080 --> 0:11:41.240
<v Speaker 1>were just names at some point. I don't have these selected,

0:11:41.240 --> 0:11:43.920
<v Speaker 1>but she did have it based on a corpus of

0:11:44.040 --> 0:11:48.280
<v Speaker 1>recipes generate new recipes which are just nightmares, you know,

0:11:48.400 --> 0:11:51.640
<v Speaker 1>like huge lists of ingredients that would be like, you know,

0:11:52.320 --> 0:11:58.480
<v Speaker 1>a third cup s'more goals, you know, a cup of Horseradishuh.

0:11:58.600 --> 0:12:01.360
<v Speaker 1>Then there was one website I don't remember who actually

0:12:01.360 --> 0:12:03.880
<v Speaker 1>did it, might have been Super Deluxe. Somebody made a

0:12:03.960 --> 0:12:07.479
<v Speaker 1>video where they took one of these AI generated recipes

0:12:07.520 --> 0:12:10.440
<v Speaker 1>and just literally made it. Now, they had some trouble

0:12:10.520 --> 0:12:13.160
<v Speaker 1>because some of the ingredients were not real words. They're

0:12:13.160 --> 0:12:16.200
<v Speaker 1>just you know, like so they had to I guess

0:12:16.200 --> 0:12:19.319
<v Speaker 1>substitute something or they put in instead of smart goals

0:12:19.360 --> 0:12:21.679
<v Speaker 1>it was coco powder or something. But they ended up

0:12:21.679 --> 0:12:25.760
<v Speaker 1>with these like, uh, pasta shells. I think that had

0:12:25.800 --> 0:12:28.679
<v Speaker 1>like chocolate and stuff all over them. But anyway, so

0:12:29.080 --> 0:12:31.560
<v Speaker 1>that I think that's currently one of the best websites

0:12:31.559 --> 0:12:33.920
<v Speaker 1>on the internet. Go to AI weirdness dot com if

0:12:33.920 --> 0:12:36.959
<v Speaker 1>you want more joy and humor. But um, but I

0:12:37.320 --> 0:12:41.080
<v Speaker 1>was wondering, why is this the funniest thing out there

0:12:41.160 --> 0:12:44.080
<v Speaker 1>for me right now? What? What is so inherently funny

0:12:44.120 --> 0:12:49.520
<v Speaker 1>about the ways machines create things that are sort of

0:12:49.679 --> 0:12:52.920
<v Speaker 1>like real language, just close enough to be in the

0:12:53.000 --> 0:12:55.160
<v Speaker 1>zone where they are funny, but at the same time

0:12:55.240 --> 0:12:58.240
<v Speaker 1>far enough off that they're they're totally hilarious. And I

0:12:58.280 --> 0:13:01.240
<v Speaker 1>think that they're at least two parts it's uh that

0:13:01.360 --> 0:13:04.400
<v Speaker 1>make this this stuff so golden. One is that there's

0:13:04.440 --> 0:13:08.840
<v Speaker 1>something inherently funny about machines trying to behave like humans

0:13:08.840 --> 0:13:12.040
<v Speaker 1>and failing. Specifically, it's the ways that they fail, and

0:13:12.040 --> 0:13:15.480
<v Speaker 1>we'll we'll definitely explore this more throughout the episode. But

0:13:16.080 --> 0:13:19.559
<v Speaker 1>the way that they failed demonstrates a kind of pristine,

0:13:19.840 --> 0:13:25.000
<v Speaker 1>oblivious quality of stupidity. It's like a kind of platonic stupidity,

0:13:25.040 --> 0:13:29.600
<v Speaker 1>isolated from the ability to appreciate itself. It's like the

0:13:29.640 --> 0:13:33.240
<v Speaker 1>funniness of, you know, watching automatic doors repeatedly trying to

0:13:33.280 --> 0:13:36.200
<v Speaker 1>close on something that's blocking them like that that's not

0:13:36.320 --> 0:13:39.120
<v Speaker 1>itself so funny, but you see an inkling of the

0:13:39.160 --> 0:13:42.920
<v Speaker 1>same thing there that comes through in these AI generated texts.

0:13:43.440 --> 0:13:45.800
<v Speaker 1>But then we sort of mentioned this earlier. It's also

0:13:45.920 --> 0:13:49.960
<v Speaker 1>funny because it tends to reveal something interesting about the

0:13:50.040 --> 0:13:53.040
<v Speaker 1>human culture game that it's trying to play like it

0:13:53.120 --> 0:13:56.640
<v Speaker 1>brings this sort of cold objectivity to phenomenon that we

0:13:56.679 --> 0:14:00.560
<v Speaker 1>don't necessarily always bring, and it can identify a quardly

0:14:00.600 --> 0:14:03.800
<v Speaker 1>replicate trends and behaviors that we might fail to notice

0:14:03.840 --> 0:14:06.400
<v Speaker 1>in the same way that like you might notice funny

0:14:06.480 --> 0:14:10.000
<v Speaker 1>things that are actually present in the names of D

0:14:10.120 --> 0:14:14.920
<v Speaker 1>and D spells by watching a computer try and replicate them. Yeah, yeah,

0:14:14.920 --> 0:14:17.520
<v Speaker 1>I think these are these are these are two strong reasons. Okay,

0:14:17.559 --> 0:14:19.400
<v Speaker 1>we're going to take a quick break. We'll be right

0:14:19.440 --> 0:14:24.600
<v Speaker 1>back with more than thank Alright, we're back now with

0:14:24.680 --> 0:14:26.360
<v Speaker 1>all things humor, And of course we'll get into this

0:14:26.400 --> 0:14:29.320
<v Speaker 1>in this episode as we continue to discuss what humor

0:14:29.520 --> 0:14:33.000
<v Speaker 1>is and then why machines in some cases achieve it,

0:14:33.240 --> 0:14:39.840
<v Speaker 1>Like the absurd is is funny, Like the absurdity is hilarious,

0:14:39.960 --> 0:14:43.040
<v Speaker 1>And it seems it seems like we're in a situation

0:14:43.080 --> 0:14:45.040
<v Speaker 1>where a lot of times, some of these, especially these

0:14:45.120 --> 0:14:50.840
<v Speaker 1>neural net situations, were we were accidentally creating absurdity engines.

0:14:50.880 --> 0:14:56.800
<v Speaker 1>We're creating machines that that produce absurdity, and uh, you know,

0:14:58.160 --> 0:15:00.040
<v Speaker 1>you know, what can you say, like why is I

0:15:00.240 --> 0:15:04.080
<v Speaker 1>is something that is absurd funny? You know? Because we'll

0:15:04.120 --> 0:15:05.800
<v Speaker 1>get into all that in a bit. But but sometimes

0:15:05.840 --> 0:15:08.440
<v Speaker 1>I feel like the answer might be it's funny because

0:15:08.440 --> 0:15:11.640
<v Speaker 1>it's funny. Well, yeah, I mean, there's definitely an ineffable

0:15:11.720 --> 0:15:15.080
<v Speaker 1>quality of of of humor. That is one of the

0:15:15.120 --> 0:15:18.040
<v Speaker 1>reasons there's so many different theories of humor. Then, as

0:15:18.080 --> 0:15:20.160
<v Speaker 1>I said, we'll explore them more as the episode goes on.

0:15:20.240 --> 0:15:24.280
<v Speaker 1>But um, yeah, it's obviously something that's really hard to

0:15:24.280 --> 0:15:26.560
<v Speaker 1>to narrow down and put your finger on. It's that

0:15:26.640 --> 0:15:29.920
<v Speaker 1>there seemed to be all these strange, conflicting, overlapping reasons

0:15:29.960 --> 0:15:33.040
<v Speaker 1>we find things funny. But I do feel like this

0:15:33.160 --> 0:15:38.360
<v Speaker 1>strain of machine humor machine failure humor being one of

0:15:38.360 --> 0:15:42.440
<v Speaker 1>the funniest types of humor UH is bigger than just

0:15:42.600 --> 0:15:46.320
<v Speaker 1>the AI text. Because it got me I started thinking

0:15:46.360 --> 0:15:49.480
<v Speaker 1>about what are some of the funniest scenes in movies

0:15:49.560 --> 0:15:51.760
<v Speaker 1>I can think of? And when I tried to think

0:15:51.800 --> 0:15:53.920
<v Speaker 1>about that, I can't help but think of the dark

0:15:54.000 --> 0:15:58.880
<v Speaker 1>humor of the hyper violent board room scene in Robo Cop.

0:15:59.040 --> 0:16:02.200
<v Speaker 1>With ed to own mind, I would argue one of

0:16:02.240 --> 0:16:04.200
<v Speaker 1>the I mean, it's not for children, this is like

0:16:04.240 --> 0:16:08.000
<v Speaker 1>a hyper violent, horrible UH scene, but it's also in

0:16:08.000 --> 0:16:10.240
<v Speaker 1>in a in a morbid way. One of the funniest

0:16:10.280 --> 0:16:13.560
<v Speaker 1>scenes I think in film history, Like and so, why

0:16:13.640 --> 0:16:16.440
<v Speaker 1>is it so funny? I think it hinges on parallels

0:16:16.840 --> 0:16:19.960
<v Speaker 1>between humans and machines and the scene and the similarities

0:16:20.000 --> 0:16:23.160
<v Speaker 1>in the ways they fail. So brief recap of the

0:16:23.200 --> 0:16:26.760
<v Speaker 1>scene is, uh, you know, Ronnie Cox plays this you

0:16:26.800 --> 0:16:31.040
<v Speaker 1>know self, serious bloviating businessman who's proudly proclaiming, you know,

0:16:31.080 --> 0:16:34.480
<v Speaker 1>I've got the technology that's the future of policing. And

0:16:34.520 --> 0:16:36.720
<v Speaker 1>he brings out this robot called ED two oh nine

0:16:36.800 --> 0:16:39.040
<v Speaker 1>that's got these big guns on it, and they're saying

0:16:39.040 --> 0:16:41.320
<v Speaker 1>that it's going to take over the police force and

0:16:41.320 --> 0:16:44.520
<v Speaker 1>it's this great new technology. And then they demonstrate it

0:16:44.600 --> 0:16:47.920
<v Speaker 1>on a guy and it malfunctions, and uh, it tells

0:16:48.000 --> 0:16:50.680
<v Speaker 1>him to drop his weapon in the demonstration, and he does,

0:16:50.760 --> 0:16:52.680
<v Speaker 1>and it doesn't seem to notice he has and then

0:16:52.720 --> 0:16:55.640
<v Speaker 1>it shoots him like a hundred times, just a ridiculous

0:16:55.680 --> 0:16:59.359
<v Speaker 1>amount of times. And but it's something about the way

0:16:59.600 --> 0:17:02.360
<v Speaker 1>that the the people in the room fail in the

0:17:02.400 --> 0:17:04.920
<v Speaker 1>same way the machine does. Like he just plows through

0:17:05.000 --> 0:17:08.639
<v Speaker 1>this this horrible, violent encounter, and then afterwards somebody in

0:17:08.680 --> 0:17:11.520
<v Speaker 1>the background is like, can we get a paramedic after

0:17:11.640 --> 0:17:15.240
<v Speaker 1>this guy has been shot like a hundred times, as if,

0:17:15.359 --> 0:17:17.959
<v Speaker 1>like the machine, they're just sort of like carrying on

0:17:18.160 --> 0:17:22.280
<v Speaker 1>their like route behaviors without understanding what they're doing or

0:17:22.320 --> 0:17:26.000
<v Speaker 1>thinking about them. Yeah. The edd to oh nine is

0:17:26.040 --> 0:17:31.320
<v Speaker 1>such a great design in the original RoboCop because it

0:17:31.320 --> 0:17:34.760
<v Speaker 1>it resembles it has animal qualities to it. It looks

0:17:34.880 --> 0:17:39.400
<v Speaker 1>kind of like a bipedal dinosaur. Uh, and yet it's

0:17:39.400 --> 0:17:42.680
<v Speaker 1>also smooth and abstract in so many ways that it

0:17:42.720 --> 0:17:45.800
<v Speaker 1>looks like either a highly designed piece of technology, which

0:17:45.800 --> 0:17:47.359
<v Speaker 1>of course it is. It has you know, it looks

0:17:47.400 --> 0:17:50.840
<v Speaker 1>like it's a nice piece of stereo equipment, but but

0:17:50.920 --> 0:17:54.719
<v Speaker 1>then it also lacks any additional details, like it's almost

0:17:54.760 --> 0:17:58.440
<v Speaker 1>a silhouette of a lie of an animal. Yes, um,

0:17:59.119 --> 0:18:01.760
<v Speaker 1>and the growls and it growls as well. But then

0:18:01.840 --> 0:18:04.119
<v Speaker 1>also later a really funny scene in the movie is

0:18:04.160 --> 0:18:08.680
<v Speaker 1>the discovery that this horrible violent killer robot is can

0:18:08.720 --> 0:18:11.600
<v Speaker 1>be defeated by stairs. Like it can't use stairs. It

0:18:11.600 --> 0:18:15.399
<v Speaker 1>has these wonderful like all terrain looking um legs that

0:18:15.480 --> 0:18:18.760
<v Speaker 1>it walks on, and that it can't manipulate stairs. It

0:18:18.800 --> 0:18:21.560
<v Speaker 1>falls down them, and then it's it's like a you know,

0:18:21.600 --> 0:18:25.239
<v Speaker 1>a ridiculous upside down duckling. I mean it's so hilarious too,

0:18:25.280 --> 0:18:27.119
<v Speaker 1>because I mean it dad to a nine is highly

0:18:27.119 --> 0:18:31.320
<v Speaker 1>effective in other situations if it's just filling a guy

0:18:31.359 --> 0:18:37.000
<v Speaker 1>with bullets, highly effective. Um, just battling RoboCop also highly

0:18:37.000 --> 0:18:40.280
<v Speaker 1>effective for the most part um. And this falls in

0:18:40.359 --> 0:18:43.440
<v Speaker 1>line I think pretty well with our experiences of machines

0:18:43.840 --> 0:18:47.240
<v Speaker 1>and of AI, we can create highly effective specialist in

0:18:47.280 --> 0:18:50.879
<v Speaker 1>many areas of AI and robotics. So you create a

0:18:50.920 --> 0:18:53.800
<v Speaker 1>machine that just puts bullets into this guy, uh, you know,

0:18:53.800 --> 0:18:55.879
<v Speaker 1>it does a great job. But when it comes to

0:18:55.960 --> 0:19:00.000
<v Speaker 1>creating a general AI or a machine that can navigate

0:19:00.080 --> 0:19:03.240
<v Speaker 1>the complex natural or even or the human created world

0:19:03.520 --> 0:19:07.639
<v Speaker 1>such as the stairs, Uh, there's continual challenge there. Like

0:19:07.720 --> 0:19:09.800
<v Speaker 1>that's kind of that's that's what a lot of of

0:19:09.840 --> 0:19:12.600
<v Speaker 1>what's going on in robotics and AI is about. Yeah,

0:19:12.720 --> 0:19:15.600
<v Speaker 1>or the just recognizing that it's not actually supposed to

0:19:15.600 --> 0:19:19.359
<v Speaker 1>shoot the board executive during the demonstration exactly. Yeah. But

0:19:19.400 --> 0:19:23.280
<v Speaker 1>then also just the idea that it's it's wonderful, it's

0:19:23.320 --> 0:19:26.920
<v Speaker 1>something and terrible at another thing. That imbalance is often

0:19:26.920 --> 0:19:28.840
<v Speaker 1>where we find a lot a lot of hilarity in

0:19:28.960 --> 0:19:33.159
<v Speaker 1>other um, you know, another comedic stories and bits of

0:19:33.160 --> 0:19:37.119
<v Speaker 1>fiction or just situations like one of my favorite cut

0:19:37.160 --> 0:19:40.600
<v Speaker 1>scenes of all times from Conan the Barbarian. So there's

0:19:40.600 --> 0:19:43.159
<v Speaker 1>a scene in it where this is the Arnold Schwarzenegger original.

0:19:43.240 --> 0:19:45.439
<v Speaker 1>There's it's like a blooper out take. That's a blooper

0:19:45.440 --> 0:19:46.840
<v Speaker 1>Outare you find it? I don't think most of the

0:19:46.880 --> 0:19:50.879
<v Speaker 1>special DVDs. It's available on YouTube as well, but basically

0:19:50.960 --> 0:19:55.800
<v Speaker 1>Conan has just escaped or been released from servitude and

0:19:55.840 --> 0:19:59.320
<v Speaker 1>he's running across the you know, the waste land essentially,

0:19:59.600 --> 0:20:02.679
<v Speaker 1>while dogs are chasing him in order to to to

0:20:02.800 --> 0:20:05.480
<v Speaker 1>eat his flesh. He manages in the finished film to

0:20:05.520 --> 0:20:08.440
<v Speaker 1>scramble up some rocks and that begins this new story

0:20:08.560 --> 0:20:11.359
<v Speaker 1>arc him discovering this great old sword. But this is

0:20:11.440 --> 0:20:15.240
<v Speaker 1>like young corporal beef body giant Arnold Schwartz, and this

0:20:15.320 --> 0:20:17.520
<v Speaker 1>is the this is the Arnold Schwarzenegger that was so

0:20:17.640 --> 0:20:20.520
<v Speaker 1>ripped he had to like lose muscle so that he

0:20:20.520 --> 0:20:24.359
<v Speaker 1>could actually hold the sword correctly. Uh So there's this

0:20:24.680 --> 0:20:28.400
<v Speaker 1>out take though, where he's running in chains, the dogs

0:20:28.400 --> 0:20:31.320
<v Speaker 1>are chasing him in the movie Dogs, and then he's

0:20:31.359 --> 0:20:34.399
<v Speaker 1>scrambling up the stones and the dogs catch him and

0:20:34.480 --> 0:20:38.120
<v Speaker 1>drag him back down, and he's just screamed a and

0:20:38.119 --> 0:20:41.240
<v Speaker 1>and cursing the whole time. Yeah, it's I watched it

0:20:41.480 --> 0:20:43.920
<v Speaker 1>after you linked it very very funny, and you see

0:20:43.960 --> 0:20:45.920
<v Speaker 1>the p as come in to get the dogs off

0:20:45.960 --> 0:20:50.440
<v Speaker 1>the dog, and he's like, yeah, it's it's funny because

0:20:50.480 --> 0:20:53.760
<v Speaker 1>the idea of Conan the barbarian, or even just Prime

0:20:53.920 --> 0:20:58.240
<v Speaker 1>Arnold failing like this, it's just start contrast to actual

0:20:58.359 --> 0:21:02.560
<v Speaker 1>or perceived strength of a character and or individual. And

0:21:02.560 --> 0:21:05.600
<v Speaker 1>it's also funny because he wasn't actually mauled to death

0:21:05.720 --> 0:21:08.560
<v Speaker 1>by movie dogs, right, of course, he wasn't actually seriously

0:21:08.560 --> 0:21:12.960
<v Speaker 1>harmed in this incident, but he was apparently inconvenienced and

0:21:13.000 --> 0:21:15.080
<v Speaker 1>had his pride wounded. Right, And it will come back

0:21:15.119 --> 0:21:18.879
<v Speaker 1>to that idea to the degree to which you know,

0:21:18.920 --> 0:21:23.120
<v Speaker 1>the severity the outcome um comes into play and determining

0:21:23.160 --> 0:21:25.399
<v Speaker 1>if something is funny or not. Yeah, I think I

0:21:25.440 --> 0:21:28.040
<v Speaker 1>can definitely see what you're talking about that the failures

0:21:28.040 --> 0:21:32.119
<v Speaker 1>of technology are especially funny when there are other ways

0:21:32.240 --> 0:21:37.000
<v Speaker 1>that the technology is highly advanced or presented as highly advanced,

0:21:37.359 --> 0:21:41.000
<v Speaker 1>and as long as like nobody of course dies, you know, um,

0:21:41.560 --> 0:21:43.320
<v Speaker 1>but but I do wonder too if there's a darker

0:21:43.320 --> 0:21:45.520
<v Speaker 1>streak and all of this too, you know, something that

0:21:45.640 --> 0:21:49.680
<v Speaker 1>ties into a deeply rooted human disdain for the other,

0:21:49.960 --> 0:21:53.840
<v Speaker 1>especially for others of species. There is a quote from C. S.

0:21:53.920 --> 0:21:56.879
<v Speaker 1>Lewis's The Lion, The Which, the Wardrobe and the Four Children.

0:21:56.960 --> 0:22:00.080
<v Speaker 1>That's my son suggested alternate title for it. He's okay, like,

0:22:00.080 --> 0:22:02.640
<v Speaker 1>why don't they call it the Land the Witch Wardrobe

0:22:02.720 --> 0:22:05.359
<v Speaker 1>the Four Children? Like they're in it too, and they're

0:22:05.400 --> 0:22:08.159
<v Speaker 1>not in the title. Seems like a missed opportunity. But anyway,

0:22:08.280 --> 0:22:10.439
<v Speaker 1>there's this wonderful quote from it that I find I

0:22:10.440 --> 0:22:15.240
<v Speaker 1>found kind of creepy on a recent reread of the book. Quote,

0:22:15.560 --> 0:22:18.440
<v Speaker 1>but in general, take my advice when you meet anything

0:22:18.760 --> 0:22:21.520
<v Speaker 1>that is going to be human and isn't yet or

0:22:21.720 --> 0:22:24.600
<v Speaker 1>used to be human once and isn't now or ought

0:22:24.640 --> 0:22:27.600
<v Speaker 1>to be human, and isn't you keep your eyes on

0:22:27.640 --> 0:22:32.160
<v Speaker 1>it and feel for your hatchet. Well, that is very creepy. Now,

0:22:32.200 --> 0:22:34.040
<v Speaker 1>I think in the context of the book, this is

0:22:34.040 --> 0:22:37.960
<v Speaker 1>going to be referring to, like, you know, possessed objects

0:22:37.960 --> 0:22:40.879
<v Speaker 1>and stuff like creepy and magical. Basically, yeah, basically, the

0:22:40.880 --> 0:22:44.359
<v Speaker 1>message here was talking. Animals are cool, you know, you

0:22:44.359 --> 0:22:46.920
<v Speaker 1>can hang out with them in Narnia, But there are

0:22:46.920 --> 0:22:49.440
<v Speaker 1>other things in Narnia that are dangerous, and you can

0:22:49.440 --> 0:22:52.440
<v Speaker 1>tell if they're dangerous based on how human they seem,

0:22:52.600 --> 0:22:54.720
<v Speaker 1>they're they're trying to be or used to be. Well,

0:22:54.760 --> 0:22:57.520
<v Speaker 1>that's one thing in a in a fantasy context, in

0:22:57.520 --> 0:22:59.720
<v Speaker 1>a in a science fiction or even just a real

0:22:59.720 --> 0:23:02.960
<v Speaker 1>technology context, that's that's a different thing entirely, and starts

0:23:03.000 --> 0:23:06.280
<v Speaker 1>making you think about uh, well, you know, fear of

0:23:06.320 --> 0:23:12.280
<v Speaker 1>advancing technology mimicking human behavior, fears of AI. Yeah, yeah, yeah,

0:23:12.280 --> 0:23:14.760
<v Speaker 1>To what extent do we delight in the falls and

0:23:14.960 --> 0:23:17.840
<v Speaker 1>errors of inhuman entities because we don't wish to see

0:23:17.840 --> 0:23:21.399
<v Speaker 1>them succeed? You know, we we celebrate that the telltale

0:23:21.440 --> 0:23:24.359
<v Speaker 1>signs of their otherness because we kind of dread the

0:23:24.440 --> 0:23:26.520
<v Speaker 1>day when there will be no way to tell. And

0:23:27.080 --> 0:23:29.440
<v Speaker 1>I think there's a lot there's a strong argument to

0:23:29.480 --> 0:23:31.600
<v Speaker 1>be made that that day will be here sooner than

0:23:31.640 --> 0:23:34.360
<v Speaker 1>we think, and in many respects it already is. We've

0:23:34.359 --> 0:23:38.160
<v Speaker 1>talked about, for instance, robocalls on the show before um

0:23:38.359 --> 0:23:42.560
<v Speaker 1>and and just how and also chat bots and to

0:23:43.320 --> 0:23:45.159
<v Speaker 1>it becomes frightening when you look at where we are

0:23:45.200 --> 0:23:48.400
<v Speaker 1>with the technology. Now, UM, certainly you get a robocall,

0:23:48.520 --> 0:23:52.040
<v Speaker 1>it's not going hopefully not going to um deceive you

0:23:52.200 --> 0:23:54.320
<v Speaker 1>long term. But I think a lot of us are

0:23:54.320 --> 0:23:56.880
<v Speaker 1>having that experience where you pick one of these up

0:23:56.960 --> 0:23:58.520
<v Speaker 1>and at first you think it is a human you

0:23:58.560 --> 0:24:02.400
<v Speaker 1>are talking to, and then you realize that it is not. Well.

0:24:02.440 --> 0:24:05.600
<v Speaker 1>I think one of the funny things about stuff like chatbots,

0:24:05.600 --> 0:24:08.520
<v Speaker 1>which also deal in language but can be much much

0:24:08.560 --> 0:24:12.040
<v Speaker 1>more convincing than these, uh, these neural networks that generate

0:24:12.160 --> 0:24:15.919
<v Speaker 1>you know, lists of uh, lists of character bios or something,

0:24:16.320 --> 0:24:19.120
<v Speaker 1>is that the things that are generally more convincing these

0:24:19.200 --> 0:24:22.720
<v Speaker 1>days are programmed. I think with more explicit rules, they

0:24:22.720 --> 0:24:25.560
<v Speaker 1>tend to have more kind of human meddling and exactly

0:24:25.560 --> 0:24:28.359
<v Speaker 1>what they're going to be doing, and by having less

0:24:28.359 --> 0:24:32.200
<v Speaker 1>freedom to be creative and all that, and ultimately having

0:24:32.280 --> 0:24:35.200
<v Speaker 1>less potential, they can actually be more kind of narrowly

0:24:35.320 --> 0:24:38.679
<v Speaker 1>convincing the I think one of the things that's interesting

0:24:38.720 --> 0:24:41.880
<v Speaker 1>about the the neural network generated text is that it's

0:24:41.920 --> 0:24:45.080
<v Speaker 1>not anywhere close yet. You know, you can't really use

0:24:45.119 --> 0:24:47.160
<v Speaker 1>it yet to make things where you go like, yeah,

0:24:47.240 --> 0:24:49.760
<v Speaker 1>that's definitely real human. I mean, maybe you can in

0:24:49.880 --> 0:24:53.439
<v Speaker 1>some again narrow scenarios, but we like, for instance, if

0:24:53.480 --> 0:24:55.480
<v Speaker 1>you have something that will tell you what your Wu

0:24:55.560 --> 0:24:58.399
<v Speaker 1>Tang clan name will be. You know, they sort of

0:24:58.920 --> 0:25:03.000
<v Speaker 1>random generation and uh systems which are totally different, different thing. Uh.

0:25:03.000 --> 0:25:05.439
<v Speaker 1>And we'll get into the distinction here. But but you know,

0:25:05.480 --> 0:25:09.920
<v Speaker 1>systems like that, just through sheer random um matching of words,

0:25:10.480 --> 0:25:13.960
<v Speaker 1>those can be effective. Yeah, but it doesn't mean that

0:25:14.359 --> 0:25:16.240
<v Speaker 1>it's you know, it's an entirely different kettle of fish

0:25:16.280 --> 0:25:19.280
<v Speaker 1>in terms of of what's going on with neural networks

0:25:19.280 --> 0:25:21.680
<v Speaker 1>and where they seem to be going. Well, maybe we

0:25:21.680 --> 0:25:23.440
<v Speaker 1>should take a quick break and then we come back.

0:25:23.480 --> 0:25:26.400
<v Speaker 1>We can explain just the basics of how this these

0:25:26.480 --> 0:25:31.320
<v Speaker 1>kind of things actually work. Alright, we're back. So I'm

0:25:31.320 --> 0:25:33.800
<v Speaker 1>gonna try to do the simple version of how a

0:25:33.840 --> 0:25:35.879
<v Speaker 1>neural network works, because if you get in the weeds,

0:25:35.880 --> 0:25:40.240
<v Speaker 1>obviously neural networks become extremely complicated. I know, I I

0:25:40.280 --> 0:25:41.919
<v Speaker 1>spent a lot of time deep in a bunch of

0:25:41.960 --> 0:25:44.720
<v Speaker 1>articles trying to understand technical details that I'm not actually

0:25:44.760 --> 0:25:48.639
<v Speaker 1>gonna end up using here. Um. So the simple version

0:25:48.760 --> 0:25:51.800
<v Speaker 1>is think of a neural network as a machine that

0:25:51.880 --> 0:25:55.840
<v Speaker 1>transforms values. That's it. You know, it has values that

0:25:55.920 --> 0:25:58.600
<v Speaker 1>come in, like number variables, and then it puts out

0:25:58.680 --> 0:26:00.960
<v Speaker 1>values at the end. It's like it's kind of like

0:26:01.000 --> 0:26:03.520
<v Speaker 1>the toaster conveyor belt at quiz Nos or one of

0:26:03.520 --> 0:26:06.880
<v Speaker 1>those sandwich shops. You know, you're untoasted. Sandwich comes in,

0:26:07.119 --> 0:26:10.800
<v Speaker 1>toasted sandwich comes out. If everything goes according to plan. Now,

0:26:10.840 --> 0:26:12.960
<v Speaker 1>if it doesn't go according to plan, maybe it splits

0:26:12.960 --> 0:26:15.800
<v Speaker 1>out something that's on fire or something that you know,

0:26:15.960 --> 0:26:18.120
<v Speaker 1>who knows what goes on in now, or nothing comes

0:26:18.160 --> 0:26:20.840
<v Speaker 1>out because the bagels are building up inside of it,

0:26:20.960 --> 0:26:23.639
<v Speaker 1>right exactly. But a neural networks core job is to

0:26:23.800 --> 0:26:27.840
<v Speaker 1>just accept inputs and produce outputs. Okay. An example would

0:26:27.880 --> 0:26:31.600
<v Speaker 1>be image recognition. Their neural networks designed to look at

0:26:31.640 --> 0:26:34.639
<v Speaker 1>an image, a digital image, and come up with a

0:26:34.720 --> 0:26:37.239
<v Speaker 1>text string that says this is what that is. So

0:26:37.320 --> 0:26:39.240
<v Speaker 1>look at a picture of a dog and say that's

0:26:39.240 --> 0:26:41.639
<v Speaker 1>a dog. And you've probably seen examples of this on

0:26:41.680 --> 0:26:44.520
<v Speaker 1>the web. So you'd have numerical values going in. It

0:26:44.640 --> 0:26:46.480
<v Speaker 1>might be things like, you know, be a field of

0:26:46.480 --> 0:26:50.680
<v Speaker 1>pixels with numerical values for their color and placement, and

0:26:50.720 --> 0:26:53.639
<v Speaker 1>then it would have an output that's, uh, say a

0:26:53.760 --> 0:26:56.560
<v Speaker 1>string of text, which would actually be like a ranking,

0:26:56.720 --> 0:27:00.280
<v Speaker 1>like the top ranked string of text that matches with

0:27:00.520 --> 0:27:03.760
<v Speaker 1>those pixels in that configuration. But so the question is

0:27:03.760 --> 0:27:06.040
<v Speaker 1>what's going on inside the machine? How does it turn

0:27:06.160 --> 0:27:09.560
<v Speaker 1>that input into the correct matching output, and then and

0:27:09.600 --> 0:27:11.960
<v Speaker 1>then of course, how does it fail? Because I've also

0:27:12.000 --> 0:27:17.920
<v Speaker 1>found this tremendously amusing. Tumbler Um recently changed their guidelines

0:27:17.920 --> 0:27:21.000
<v Speaker 1>about what's acceptable content and what's not and stuff to

0:27:21.000 --> 0:27:24.960
<v Speaker 1>blow your mind has has Slash had a tumbler account

0:27:25.520 --> 0:27:28.800
<v Speaker 1>recently just rebranded it as the Transgenesis tumbler account, but

0:27:28.920 --> 0:27:30.199
<v Speaker 1>it had a lot of old stuff to put your

0:27:30.240 --> 0:27:32.520
<v Speaker 1>mind content on there. And suddenly I got a page

0:27:32.520 --> 0:27:35.120
<v Speaker 1>of all the things that have been flagged for for

0:27:35.280 --> 0:27:38.440
<v Speaker 1>potentially violating the new terms. And it was amusing because

0:27:38.440 --> 0:27:40.159
<v Speaker 1>some of it was like, okay, well that has a

0:27:40.480 --> 0:27:43.159
<v Speaker 1>classical painting on it. It's got like a you know,

0:27:43.200 --> 0:27:45.800
<v Speaker 1>it's something there's something that might look like nudity, and

0:27:45.840 --> 0:27:49.000
<v Speaker 1>so it got flagged, makes sense, or the topic is

0:27:49.040 --> 0:27:51.840
<v Speaker 1>something that is a little too sketchy for them, and

0:27:51.880 --> 0:27:54.240
<v Speaker 1>they're like, okay, that's been flagged by the machine. But

0:27:54.320 --> 0:27:57.320
<v Speaker 1>the most hilarious one was a picture of a baby

0:27:57.400 --> 0:28:01.760
<v Speaker 1>bat on on somebody's pall them and that was flagged

0:28:01.960 --> 0:28:04.960
<v Speaker 1>as as as likely inappropriate. And so I was just

0:28:04.960 --> 0:28:06.600
<v Speaker 1>trying to figure that one out, Like what is it

0:28:06.600 --> 0:28:09.240
<v Speaker 1>about a baby bat that it? I mean, did it

0:28:09.320 --> 0:28:13.200
<v Speaker 1>think this was genitalia or like, like what because because

0:28:13.200 --> 0:28:15.680
<v Speaker 1>I know it's somehow clicked off a number of boxes

0:28:16.080 --> 0:28:19.119
<v Speaker 1>and when and then the automated result was no baby

0:28:19.160 --> 0:28:21.399
<v Speaker 1>bats on tumbler. Well, I guess I don't know if

0:28:21.440 --> 0:28:23.280
<v Speaker 1>you know, but I don't know what the mechanism for

0:28:23.359 --> 0:28:26.080
<v Speaker 1>identifying that is. It might be something like this, but

0:28:26.320 --> 0:28:28.560
<v Speaker 1>yeah that I love seeing that kind of failure. And

0:28:28.840 --> 0:28:31.119
<v Speaker 1>notice that we are laughing now like it or we

0:28:31.119 --> 0:28:33.639
<v Speaker 1>were laughing. It is funny that it looks at that

0:28:33.720 --> 0:28:35.880
<v Speaker 1>and says, I don't know about this bad I think

0:28:35.880 --> 0:28:38.120
<v Speaker 1>this might be porno. Yeah, I mean then the stakes

0:28:38.120 --> 0:28:40.920
<v Speaker 1>are pretty low. I ultimately, one picture of a baby

0:28:40.920 --> 0:28:43.320
<v Speaker 1>bat no longer on a tumbler page that we don't

0:28:43.360 --> 0:28:45.920
<v Speaker 1>really use. It doesn't really affect me personally, but you

0:28:45.920 --> 0:28:49.520
<v Speaker 1>could see where this could lead to, uh too far

0:28:49.600 --> 0:28:53.719
<v Speaker 1>worse problems if given the right scenario. Okay, so so

0:28:53.760 --> 0:28:55.960
<v Speaker 1>back to the neural network. So you've got this machine.

0:28:56.080 --> 0:28:59.520
<v Speaker 1>Inside the machine, values are being transformed. You have inputs

0:28:59.560 --> 0:29:02.960
<v Speaker 1>and output and so inside on the inside, and neural

0:29:03.000 --> 0:29:07.040
<v Speaker 1>network consists of layers of sort of stations of value

0:29:07.040 --> 0:29:10.840
<v Speaker 1>transformation that are called referred to as neurons. And each

0:29:10.920 --> 0:29:15.040
<v Speaker 1>neuron essentially accepts a series of numerical values as input

0:29:15.520 --> 0:29:19.000
<v Speaker 1>and then it just performs some kind of mathematical transformation

0:29:19.120 --> 0:29:22.560
<v Speaker 1>of those values based on what's known as the weights

0:29:22.600 --> 0:29:25.880
<v Speaker 1>of its connections with the sources that received the inputs from.

0:29:26.160 --> 0:29:29.880
<v Speaker 1>So you've got these interlinking sources of information, inputs and

0:29:29.920 --> 0:29:34.200
<v Speaker 1>outputs throughout, and each neuron takes inputs, sums them, does

0:29:34.240 --> 0:29:37.520
<v Speaker 1>some transformation, and produces an output. So for each neuron,

0:29:37.560 --> 0:29:40.040
<v Speaker 1>you've got a bunch of numbers coming from different sources.

0:29:40.040 --> 0:29:42.720
<v Speaker 1>Each one gets treated with a certain bias based on

0:29:42.760 --> 0:29:45.720
<v Speaker 1>where it came from, and then the neuron spits out

0:29:45.720 --> 0:29:48.600
<v Speaker 1>a new value. And these neurons exist in layers, so

0:29:48.640 --> 0:29:52.320
<v Speaker 1>there are these waves of inputs, say the pixels and

0:29:52.480 --> 0:29:57.040
<v Speaker 1>image getting passed and transformed through one layer after another

0:29:57.360 --> 0:30:00.520
<v Speaker 1>of these neurons, until finally the network produces numbers that

0:30:00.600 --> 0:30:03.680
<v Speaker 1>constitute its final outputs. In this case, this would be

0:30:03.720 --> 0:30:06.480
<v Speaker 1>something like it's top guesses at the word string that

0:30:06.520 --> 0:30:09.320
<v Speaker 1>describes the image you put in at the beginning. Now

0:30:09.320 --> 0:30:11.640
<v Speaker 1>you'll see from this that the value of neural network

0:30:11.720 --> 0:30:16.520
<v Speaker 1>depends completely on how well those connections between neurons are

0:30:16.600 --> 0:30:19.280
<v Speaker 1>weighted to produce the correct results. If they just have

0:30:19.440 --> 0:30:22.600
<v Speaker 1>random weights, then the network will just produce random output.

0:30:22.680 --> 0:30:24.920
<v Speaker 1>It won't be any better than making up numbers at random.

0:30:24.960 --> 0:30:28.320
<v Speaker 1>So the network has to be calibrated or trained somehow

0:30:28.360 --> 0:30:31.000
<v Speaker 1>to produce outputs that are correct, and there are multiple

0:30:31.000 --> 0:30:33.280
<v Speaker 1>ways to do this. Uh. It of course could be

0:30:33.320 --> 0:30:35.720
<v Speaker 1>programmed to some degree by hand, right, you could have

0:30:35.720 --> 0:30:39.560
<v Speaker 1>a program or explicitly uh going in and tinkering with

0:30:39.640 --> 0:30:42.480
<v Speaker 1>waiting rules to try to get the outcomes to be better.

0:30:42.760 --> 0:30:45.200
<v Speaker 1>But can it can also be trained through machine learning,

0:30:45.280 --> 0:30:49.080
<v Speaker 1>which is a process where inputs are already associated with

0:30:49.160 --> 0:30:53.000
<v Speaker 1>correct outputs, Like you've already got a text string associated

0:30:53.040 --> 0:30:55.280
<v Speaker 1>with the image that you put in and you say

0:30:55.320 --> 0:30:57.320
<v Speaker 1>this is what you should say when it comes out,

0:30:57.640 --> 0:31:00.080
<v Speaker 1>And each time it runs the process, it checks to

0:31:00.120 --> 0:31:02.840
<v Speaker 1>see how far off from the correct known output that

0:31:02.920 --> 0:31:06.000
<v Speaker 1>it was, and then tries to change the internal waiting

0:31:06.000 --> 0:31:08.600
<v Speaker 1>to get closer to the correct answer. And of course,

0:31:08.680 --> 0:31:11.280
<v Speaker 1>with automatic machine learning, you can do this at scale.

0:31:11.600 --> 0:31:14.520
<v Speaker 1>You can do it thousands of times. You could potentially

0:31:14.560 --> 0:31:17.440
<v Speaker 1>do it millions of times just training over and over,

0:31:17.680 --> 0:31:19.400
<v Speaker 1>and you might be able to see a parallel here

0:31:19.440 --> 0:31:21.840
<v Speaker 1>with one of the ways that we actually learn. Uh,

0:31:22.080 --> 0:31:24.520
<v Speaker 1>you know, we we learned in multiple ways. Sometimes we

0:31:24.640 --> 0:31:27.840
<v Speaker 1>learned by being taught explicit rules to follow, like if

0:31:27.880 --> 0:31:30.640
<v Speaker 1>we're learning in school what an insect is, we might

0:31:30.720 --> 0:31:33.360
<v Speaker 1>learn that an insect is a small animal with an

0:31:33.360 --> 0:31:38.200
<v Speaker 1>exoskeleton that has six legs. Or sometimes, on the other hand,

0:31:38.240 --> 0:31:41.640
<v Speaker 1>we learned to generalize from particulars. We might see lots

0:31:41.640 --> 0:31:44.480
<v Speaker 1>of animals pictures of lots of animals and notice that

0:31:44.560 --> 0:31:47.600
<v Speaker 1>the ones that are called insects all happen to have

0:31:47.720 --> 0:31:51.640
<v Speaker 1>six legs and exo skeletons, and therefore we derive this

0:31:51.680 --> 0:31:55.440
<v Speaker 1>category called insect from that survey. And in logic, of course,

0:31:55.480 --> 0:31:57.920
<v Speaker 1>this this process where we come up with general rules

0:31:57.960 --> 0:32:01.240
<v Speaker 1>from lots of individual examples, is known as induction. So

0:32:01.360 --> 0:32:03.760
<v Speaker 1>machine learning to train your oal networks is kind of

0:32:03.800 --> 0:32:08.320
<v Speaker 1>like allowing computers to learn categories by induction, kind of

0:32:08.360 --> 0:32:10.240
<v Speaker 1>like we do when we just go out and look

0:32:10.280 --> 0:32:12.680
<v Speaker 1>at the world and see what we find. But one

0:32:12.720 --> 0:32:15.960
<v Speaker 1>of the things that really sets humans apart from computers

0:32:16.560 --> 0:32:20.800
<v Speaker 1>is that humans seem to have this amazing, remarkable ability

0:32:21.200 --> 0:32:24.320
<v Speaker 1>to generalize from particulars. We can often get the gist

0:32:24.400 --> 0:32:28.160
<v Speaker 1>of a category from just a tiny handful of examples.

0:32:28.200 --> 0:32:31.680
<v Speaker 1>You know, when you're giving somebody examples of something to like,

0:32:31.800 --> 0:32:34.000
<v Speaker 1>give them the gist of what you're talking about. You

0:32:34.040 --> 0:32:37.000
<v Speaker 1>don't usually need to list a million examples. You can

0:32:37.080 --> 0:32:40.680
<v Speaker 1>list two or three maybe, or sometimes even just one. Yeah.

0:32:40.800 --> 0:32:43.280
<v Speaker 1>This gets into the idea of judging a book by

0:32:43.280 --> 0:32:46.840
<v Speaker 1>its cover, right, not supposed to, but we often do,

0:32:46.960 --> 0:32:49.160
<v Speaker 1>and sometimes you you can if you pick up on

0:32:49.200 --> 0:32:52.240
<v Speaker 1>particular things on you know, specifically to use the book, example,

0:32:52.720 --> 0:32:57.200
<v Speaker 1>specific things about the design or the era of the cover. Yeah. Yeah,

0:32:57.240 --> 0:33:00.520
<v Speaker 1>And in fact, sometimes you can you can judge things

0:33:00.560 --> 0:33:03.720
<v Speaker 1>about the contents of a book just by knowing certain

0:33:03.760 --> 0:33:06.880
<v Speaker 1>things about how certain types of books end up with

0:33:06.920 --> 0:33:10.960
<v Speaker 1>certain types of covers. Like you might think I tend

0:33:11.080 --> 0:33:14.400
<v Speaker 1>to like books that have hand drawn illustrations on the

0:33:14.400 --> 0:33:16.880
<v Speaker 1>cover more than I like books that have sort of

0:33:16.920 --> 0:33:21.320
<v Speaker 1>like c g I stock image cover photos, which means

0:33:21.360 --> 0:33:24.320
<v Speaker 1>you probably like books from like, you know, at least

0:33:24.360 --> 0:33:28.040
<v Speaker 1>the nineteen eighties and before. Yeah, because it seems like

0:33:28.040 --> 0:33:31.080
<v Speaker 1>we have far fewer hand drawn, uh you know, covers

0:33:31.120 --> 0:33:34.520
<v Speaker 1>on books these days. Yeah, why are people putting stock

0:33:34.600 --> 0:33:36.800
<v Speaker 1>photos on the covers of books? I do not get it.

0:33:37.120 --> 0:33:39.920
<v Speaker 1>But you you didn't need to read a million books

0:33:39.960 --> 0:33:41.960
<v Speaker 1>to come to that conclusion. You could probably come to

0:33:42.000 --> 0:33:45.959
<v Speaker 1>that conclusion after reading I don't know three books like you,

0:33:46.200 --> 0:33:49.480
<v Speaker 1>really we get the gist of things really fast. And

0:33:50.080 --> 0:33:54.400
<v Speaker 1>that's in contrast to computers, which really really don't at all.

0:33:54.840 --> 0:33:58.080
<v Speaker 1>This is one of those strange and amazing things about neuroscience,

0:33:58.120 --> 0:34:00.240
<v Speaker 1>about the human brain. How do we solve such a

0:34:00.280 --> 0:34:04.719
<v Speaker 1>difficult problem as generalizing from particulars with so few examples

0:34:04.720 --> 0:34:07.320
<v Speaker 1>to draw from? And of course, another example of the

0:34:07.320 --> 0:34:10.040
<v Speaker 1>generalizing power of the human brain is in language, Like

0:34:10.080 --> 0:34:11.680
<v Speaker 1>we've been talking about it, like, how is it that

0:34:11.960 --> 0:34:15.000
<v Speaker 1>most of the time kids learn how to speak a

0:34:15.080 --> 0:34:19.280
<v Speaker 1>language without being taught explicit rules of syntax and grammar

0:34:19.320 --> 0:34:22.120
<v Speaker 1>and the definitions and usages of all the common words,

0:34:22.480 --> 0:34:26.480
<v Speaker 1>and without hearing billions and billions of examples of sentences.

0:34:26.560 --> 0:34:29.120
<v Speaker 1>They just pick it up. Well, there's there there are

0:34:29.160 --> 0:34:31.600
<v Speaker 1>some specific answers to that. If we've talked about that

0:34:31.600 --> 0:34:33.839
<v Speaker 1>on the show before. Yeah, well, I mean I think

0:34:33.880 --> 0:34:35.839
<v Speaker 1>that there there is a good case to be made

0:34:35.880 --> 0:34:40.000
<v Speaker 1>that the human brain is specially geared towards language acquisition

0:34:40.040 --> 0:34:44.280
<v Speaker 1>in childhood. That's sort of like one of our species superpowers.

0:34:44.320 --> 0:34:47.800
<v Speaker 1>And then those windows close, or they don't completely close,

0:34:47.880 --> 0:34:53.320
<v Speaker 1>but they the windows become much smaller later on in life.

0:34:53.960 --> 0:34:56.319
<v Speaker 1>You know. Speaking of children, they are you know, they

0:34:56.360 --> 0:34:59.759
<v Speaker 1>are also frequently a font of weirdness and beauty as

0:34:59.840 --> 0:35:03.000
<v Speaker 1>they two are learning to function in the human world,

0:35:03.080 --> 0:35:05.600
<v Speaker 1>in the in the adult human world, and then they

0:35:05.640 --> 0:35:08.959
<v Speaker 1>say and do things that sometimes hit a weird zone

0:35:09.000 --> 0:35:12.799
<v Speaker 1>that is either hilarious or sometimes a little frightening, or

0:35:12.920 --> 0:35:15.600
<v Speaker 1>or even a little bit elegant. Yes, I know exactly

0:35:15.640 --> 0:35:19.480
<v Speaker 1>what you mean. Like, kids often do the same funny

0:35:19.560 --> 0:35:24.279
<v Speaker 1>outputs based on induction that these machine learning algorithms do.

0:35:24.920 --> 0:35:27.000
<v Speaker 1>Like you can see it's funny in the same way

0:35:27.000 --> 0:35:29.120
<v Speaker 1>that they might say something that's a little bit off

0:35:29.160 --> 0:35:32.839
<v Speaker 1>and kind of absurd, but you can sort of understand

0:35:33.000 --> 0:35:36.760
<v Speaker 1>the rules that got them there, right. Like one example

0:35:36.880 --> 0:35:39.160
<v Speaker 1>I always refer to is from years and years ago,

0:35:39.400 --> 0:35:41.960
<v Speaker 1>I went to a children's puppet shows before I had

0:35:42.040 --> 0:35:43.920
<v Speaker 1>a kid, my my wife and I went to check

0:35:43.960 --> 0:35:47.760
<v Speaker 1>this out because it was actors from a local improv group,

0:35:48.040 --> 0:35:52.799
<v Speaker 1>Dad's Garage, Atlanta. They put on the show Uncle Grandpa's

0:35:52.800 --> 0:35:54.960
<v Speaker 1>Who Deal His story Time? I think it was, and

0:35:55.000 --> 0:35:57.239
<v Speaker 1>so you had these sort of seasoned, uh you know,

0:35:57.280 --> 0:35:59.960
<v Speaker 1>improv vets, and they were doing a puppet show for kids,

0:36:00.320 --> 0:36:03.799
<v Speaker 1>and they're taking ideas from the audience and they said, like,

0:36:03.800 --> 0:36:06.439
<v Speaker 1>who should our main character be? And so they hand

0:36:06.440 --> 0:36:08.560
<v Speaker 1>the mic to some little little girl in the front

0:36:08.640 --> 0:36:12.040
<v Speaker 1>row and she says Batman the Girl, which which is

0:36:12.080 --> 0:36:14.680
<v Speaker 1>so hilarious and I don't think an adult would be

0:36:14.719 --> 0:36:16.839
<v Speaker 1>able to come up with that, but you can sort

0:36:16.840 --> 0:36:18.839
<v Speaker 1>of tease it apart and figure out how she got

0:36:18.880 --> 0:36:21.359
<v Speaker 1>there and you know, with it. But uh, but it's

0:36:21.440 --> 0:36:23.959
<v Speaker 1>it's just one of you know, many examples I'm sure

0:36:24.040 --> 0:36:26.839
<v Speaker 1>that that anyone with children in their life can can

0:36:26.880 --> 0:36:28.799
<v Speaker 1>turn to where they come with something that is just

0:36:29.360 --> 0:36:33.120
<v Speaker 1>so goofy or weird or or sometimes terrifying. Well, I

0:36:33.160 --> 0:36:36.480
<v Speaker 1>think the real funniness and pleasure in that is that

0:36:36.520 --> 0:36:40.239
<v Speaker 1>it's Batman the Girl and not bat Girl. Right, Yeah,

0:36:40.280 --> 0:36:43.319
<v Speaker 1>Like it's almost there, and but by not being there,

0:36:43.960 --> 0:36:47.759
<v Speaker 1>it's it's also like it's it's even better like it's

0:36:47.760 --> 0:36:51.239
<v Speaker 1>not it's not Backgirl, it's Batman the girl. It's just

0:36:51.280 --> 0:36:55.400
<v Speaker 1>so nonsensical, uh and beautiful at the same time. You know.

0:36:55.480 --> 0:36:59.840
<v Speaker 1>Another one of the great AI text generation experiments that

0:37:00.000 --> 0:37:03.360
<v Speaker 1>in El Shane did on her blog was was generating

0:37:03.800 --> 0:37:07.399
<v Speaker 1>that you know those Valentine's Day candy hearts. Yes, yes,

0:37:07.760 --> 0:37:09.920
<v Speaker 1>She had a programmed sample those and then try to

0:37:09.920 --> 0:37:12.600
<v Speaker 1>come up with examples and ended up saying things I

0:37:12.600 --> 0:37:18.840
<v Speaker 1>think like like sweat, pooh and hole and time hug.

0:37:19.120 --> 0:37:22.680
<v Speaker 1>Time hug sounds good time hug time like time cop. Yeah,

0:37:22.719 --> 0:37:25.359
<v Speaker 1>but it also sounds like something that like it might

0:37:25.360 --> 0:37:28.479
<v Speaker 1>be a term that aliens come up with for human love.

0:37:28.840 --> 0:37:30.840
<v Speaker 1>It's like they engage not in a hug, but in

0:37:30.880 --> 0:37:33.640
<v Speaker 1>a time hug. It is as if they are hugging

0:37:33.920 --> 0:37:36.480
<v Speaker 1>for the rest of their lives, you know, or something

0:37:36.520 --> 0:37:41.799
<v Speaker 1>like that. Another one all hover and then finally bog love. UM.

0:37:42.239 --> 0:37:45.320
<v Speaker 1>My son, who as of this recording is is six

0:37:45.400 --> 0:37:49.120
<v Speaker 1>almost seven, he drew a picture for my wife and

0:37:49.160 --> 0:37:53.080
<v Speaker 1>I for for Valentine's Valentine's Day CARDI did his school

0:37:53.480 --> 0:37:57.160
<v Speaker 1>and it depicted dinosaurs as these want to do um

0:37:57.239 --> 0:38:00.759
<v Speaker 1>and they're there are some herbivores walking about, there are

0:38:00.840 --> 0:38:05.600
<v Speaker 1>some carnivores eating the flash of fallen dinosaurs, and then,

0:38:05.680 --> 0:38:08.200
<v Speaker 1>as is typical in paleo art, there may be a

0:38:08.239 --> 0:38:11.280
<v Speaker 1>volcano in the background. But then there is a meteor

0:38:11.640 --> 0:38:15.880
<v Speaker 1>coming in hard and fastid Yeah, and he writes, uh,

0:38:16.160 --> 0:38:20.399
<v Speaker 1>I love you on the on the meteor, which which

0:38:20.600 --> 0:38:23.359
<v Speaker 1>is absolutely wonderful because it's like at once it is

0:38:23.520 --> 0:38:26.600
<v Speaker 1>like like this is beautiful, Like he totally means all

0:38:26.640 --> 0:38:28.680
<v Speaker 1>of this, and it's like the most beautiful Valentine I've

0:38:28.680 --> 0:38:31.319
<v Speaker 1>ever received and yet to put it on the the

0:38:31.400 --> 0:38:35.480
<v Speaker 1>instrument of of the of the extinction events. It's just

0:38:35.560 --> 0:38:41.120
<v Speaker 1>so weird and like it it's accidentally brilliant, you know. Yeah,

0:38:41.200 --> 0:38:42.760
<v Speaker 1>I saw that when you put it on the internet.

0:38:42.800 --> 0:38:45.480
<v Speaker 1>It was the sweetest thing I've ever seen. It was

0:38:45.520 --> 0:38:47.680
<v Speaker 1>so good, and it's like, yeah, this is that. This

0:38:47.719 --> 0:38:50.160
<v Speaker 1>is how love works. Love is a is a destroyer

0:38:50.360 --> 0:38:52.719
<v Speaker 1>as well as the creator. Now, we do want to

0:38:52.719 --> 0:38:55.680
<v Speaker 1>stress it with the especially with these text based situations.

0:38:56.320 --> 0:39:00.319
<v Speaker 1>We're not talking about merror random mashups of text, uh,

0:39:00.400 --> 0:39:03.719
<v Speaker 1>such as like the Wu Tang clan name generator or

0:39:04.760 --> 0:39:07.080
<v Speaker 1>um or more literary example would be the cut up

0:39:07.120 --> 0:39:11.600
<v Speaker 1>technique popularized by author William S. Burrows, where you just

0:39:11.640 --> 0:39:15.759
<v Speaker 1>have like a random um mechanism and play to sort

0:39:15.760 --> 0:39:20.480
<v Speaker 1>of splice words or or sentences together to get something

0:39:21.320 --> 0:39:25.200
<v Speaker 1>that may have sort of accidental meaning to it. No,

0:39:25.360 --> 0:39:29.600
<v Speaker 1>the neural net programs are They are algorithms attempting as

0:39:29.680 --> 0:39:33.880
<v Speaker 1>best they can to approximate the quality of the the

0:39:33.920 --> 0:39:37.399
<v Speaker 1>input texts, the text they're trained on. So they're doing

0:39:37.400 --> 0:39:40.480
<v Speaker 1>their best to make something like this. Uh. And then,

0:39:40.520 --> 0:39:42.440
<v Speaker 1>of course, I mean one of the things is you

0:39:42.520 --> 0:39:45.480
<v Speaker 1>might say, well, why don't they just perfectly spit back

0:39:45.480 --> 0:39:48.239
<v Speaker 1>out the text you've trained them on. In fact, if

0:39:48.280 --> 0:39:50.719
<v Speaker 1>you don't tell them not to, they'll do that. You know,

0:39:50.760 --> 0:39:53.319
<v Speaker 1>they'll just spit back in exactly what you fed in.

0:39:53.680 --> 0:39:55.960
<v Speaker 1>You have to sort of like change some values and

0:39:55.960 --> 0:39:59.239
<v Speaker 1>tinker with it to prevent what's known as overfitting in

0:39:59.280 --> 0:40:02.880
<v Speaker 1>this world. Uh, to sort of force the algorithms to

0:40:02.920 --> 0:40:05.640
<v Speaker 1>be more creative and try to come up with new

0:40:05.760 --> 0:40:08.600
<v Speaker 1>versions of the kind of thing they've seen instead of

0:40:08.640 --> 0:40:11.200
<v Speaker 1>just completely copying what you fed them. Yeah, because we

0:40:11.200 --> 0:40:16.640
<v Speaker 1>want these machines to rule the world, not just Hollywood, right, Um,

0:40:16.680 --> 0:40:18.560
<v Speaker 1>But you know we also have to to distinguish it.

0:40:18.600 --> 0:40:22.160
<v Speaker 1>We're also not talking about fake i AI generated text,

0:40:22.239 --> 0:40:27.439
<v Speaker 1>which can certainly be tremendously entertaining as well. Yes, that's

0:40:27.480 --> 0:40:32.160
<v Speaker 1>such a great genre, people pretending to be neural networks

0:40:32.239 --> 0:40:37.160
<v Speaker 1>creating machine learning generated text, which it's such an amazing

0:40:37.280 --> 0:40:41.920
<v Speaker 1>reversal of principles that humans have intuitively detected what's funny

0:40:42.040 --> 0:40:46.080
<v Speaker 1>about machine learning generated text and then made fake human

0:40:46.160 --> 0:40:50.000
<v Speaker 1>designed versions of it to exploit that inherent humor. Right,

0:40:50.200 --> 0:40:51.759
<v Speaker 1>I don't I don't know how you get deeper on

0:40:51.800 --> 0:40:54.719
<v Speaker 1>the irony pit than that. Yeah, there's a wonderful two

0:40:54.719 --> 0:40:58.520
<v Speaker 1>thousand eighteen tweet by comedian Keaton Patty and this was

0:40:58.560 --> 0:41:01.239
<v Speaker 1>the tweet quote, I worked a boat to watch over

0:41:01.320 --> 0:41:04.160
<v Speaker 1>a thousand hours of Alive Garden commercials and then asked

0:41:04.160 --> 0:41:06.439
<v Speaker 1>it to write an alive garden commercial of its own.

0:41:06.760 --> 0:41:09.879
<v Speaker 1>Here is the first page, uh and uh. And then

0:41:10.080 --> 0:41:14.320
<v Speaker 1>he proceeded to include the images of this text that

0:41:14.480 --> 0:41:18.759
<v Speaker 1>script rather for and alive garden commercial. And and it's

0:41:18.760 --> 0:41:22.960
<v Speaker 1>filled with hilarious nonsense like I shall eat Italian citizens

0:41:23.360 --> 0:41:27.359
<v Speaker 1>and unlimited stick and playing seeming you know, to play

0:41:27.440 --> 0:41:31.480
<v Speaker 1>upon the whole catchphrase of what when you're here? Your home.

0:41:31.520 --> 0:41:33.080
<v Speaker 1>I think when you're here your family and you hear

0:41:33.120 --> 0:41:37.200
<v Speaker 1>your family. There's there's one from this UH, this fake

0:41:37.239 --> 0:41:40.400
<v Speaker 1>script that says leave without me, I'm home, which I

0:41:40.480 --> 0:41:42.960
<v Speaker 1>just I love. I remember laughing so hard at this

0:41:43.000 --> 0:41:44.880
<v Speaker 1>when it came out as well. I love that the

0:41:44.880 --> 0:41:49.959
<v Speaker 1>waitress says lasagna wings with extra italy. Yeah. There's also

0:41:50.040 --> 0:41:53.719
<v Speaker 1>like really funny stage directions in it. Yeah, well, oh yeah,

0:41:53.719 --> 0:41:56.600
<v Speaker 1>you mentioned the infinite uh where it says like we

0:41:56.840 --> 0:42:00.400
<v Speaker 1>the gluten classic O. We believe the wage risk that

0:42:00.480 --> 0:42:02.560
<v Speaker 1>it is from the kitchen. We have no reason not

0:42:02.640 --> 0:42:06.320
<v Speaker 1>to believe. Now, of course, this is just a comedian

0:42:06.440 --> 0:42:09.640
<v Speaker 1>doing this thing, trying to pretend to be an AI. UM.

0:42:09.680 --> 0:42:12.200
<v Speaker 1>But I was reading an article where they quoted Janelle Shane,

0:42:12.280 --> 0:42:14.600
<v Speaker 1>the author of the AI Weirdness blog, who trains all

0:42:14.640 --> 0:42:16.439
<v Speaker 1>these neural networks to come up with all this funny

0:42:16.440 --> 0:42:18.440
<v Speaker 1>stuff we were talking about at the beginning of the episode,

0:42:18.800 --> 0:42:20.680
<v Speaker 1>And you know, she talks about that there are ways

0:42:20.760 --> 0:42:24.759
<v Speaker 1>to notice UH when something was probably written by a

0:42:24.880 --> 0:42:28.160
<v Speaker 1>human instead of by an actual AI. Both can be

0:42:28.239 --> 0:42:31.400
<v Speaker 1>absurd in similar funny ways. But one of the problems

0:42:31.440 --> 0:42:34.520
<v Speaker 1>with this script UH in passing as a real AI

0:42:34.560 --> 0:42:38.600
<v Speaker 1>generated text is that it's actually too coherent, like it's memory.

0:42:38.719 --> 0:42:42.279
<v Speaker 1>Its memory is too long. It remembers characters from many

0:42:42.360 --> 0:42:45.720
<v Speaker 1>lines earlier and still has them appearing and saying things.

0:42:46.440 --> 0:42:49.960
<v Speaker 1>Actual AI generated texts of a much shorter memory. They

0:42:50.040 --> 0:42:53.360
<v Speaker 1>they're not consistent in that way. They don't make Actually,

0:42:53.680 --> 0:42:55.759
<v Speaker 1>they make even less sense than the fake all of

0:42:55.800 --> 0:43:00.200
<v Speaker 1>Garden commercial. So she's saying, don't reach for your match

0:43:00.280 --> 0:43:03.480
<v Speaker 1>it on this one because a real neural net generated

0:43:03.600 --> 0:43:07.839
<v Speaker 1>text only mimics forms, it doesn't mimic meaning. And this

0:43:07.920 --> 0:43:11.680
<v Speaker 1>thing it's it means too much. It's too clever. Okay,

0:43:11.680 --> 0:43:13.919
<v Speaker 1>So we're gonna go ahead and close out this episode now,

0:43:13.960 --> 0:43:16.040
<v Speaker 1>but again there's going to be a part two where

0:43:16.040 --> 0:43:19.160
<v Speaker 1>we continue this discussion and really get more into the

0:43:19.200 --> 0:43:21.759
<v Speaker 1>meat of the topic. In the meantime, check out Stuff

0:43:21.760 --> 0:43:23.480
<v Speaker 1>to Blow your Mind dot com. That's the mothership. That's

0:43:23.480 --> 0:43:25.959
<v Speaker 1>where we'll find all the episodes of the show. There's

0:43:25.960 --> 0:43:27.959
<v Speaker 1>also a little button at the top you can click

0:43:28.000 --> 0:43:30.200
<v Speaker 1>on that go to our T shirt store get t

0:43:30.320 --> 0:43:33.319
<v Speaker 1>shirt stickers with a number of cool designs, designs that

0:43:33.640 --> 0:43:36.160
<v Speaker 1>line up with certain show topics, like the Great Basilisk

0:43:36.320 --> 0:43:39.960
<v Speaker 1>or various Squirrel episodes uh as well as just as

0:43:39.960 --> 0:43:44.200
<v Speaker 1>you know, show logo um material as well, and if

0:43:44.200 --> 0:43:46.200
<v Speaker 1>you want to support the show, the best thing that

0:43:46.239 --> 0:43:48.719
<v Speaker 1>you can do is rate and review stuff to blow

0:43:48.719 --> 0:43:50.239
<v Speaker 1>your mind wherever you have the power to do so.

0:43:50.320 --> 0:43:53.239
<v Speaker 1>Wherever you get this podcast, give us some stars, give

0:43:53.320 --> 0:43:55.719
<v Speaker 1>us a nice review. It really helps us out in

0:43:55.760 --> 0:43:59.680
<v Speaker 1>our war against the almighty algorithms. Big thanks as always

0:43:59.719 --> 0:44:03.360
<v Speaker 1>to our excellent audio producers Alex Williams and Tor Harrison.

0:44:03.640 --> 0:44:05.120
<v Speaker 1>If you would like to get in touch with us

0:44:05.120 --> 0:44:07.520
<v Speaker 1>directly with feedback about this episode or any other, to

0:44:07.560 --> 0:44:10.120
<v Speaker 1>suggest topic for the future, or just to say hi,

0:44:10.360 --> 0:44:13.399
<v Speaker 1>you can email us at blow the Mind at how

0:44:13.480 --> 0:44:27.720
<v Speaker 1>stuff works dot com for moralns and thousands of other topics.

0:44:27.880 --> 0:44:52.120
<v Speaker 1>Does it how stuff works dot com