Artificial Intelligence - AI Should Have Known Theme

Andi 0:01

Oh God, that's always really creepy when the AI is like, call me this. It's like, no, I'm not gonna do that. Yeah,

Sups 0:07

this is the one that I'm most skeptical about for the simple reason, if AI were to do it, don't you think this would have been a bigger news?

Tanner 0:16

hallucination is basically a lie But you can't really say it's a lie because it's not. Nefarious. It's not

Andi 0:22

It's not trying to trick you I mean that now we're talking about the philosophy of lying. Yeah. Right. hello and welcome to, I should have known the trivia game show that can't be trusted. Each week, our Quizmaster will present you with four facts about a topic, but one of those facts is a lie, So this month we're doing AI should have known. AI themed and AI generated episodes, and to kick us off, we have our Quizmaster Tanner.

Tanner 0:47

Hi everyone.

Andi 0:48

And he is going to be presenting a traditional style show about artificial intelligence, so he's going to give us four facts about ai, but one of his facts is a lie. So join me, Andy, and our other host Sups in trying to figure out which one it is.

Tanner 1:03

So today we're gonna start from the beginning where AI came from. talk about how it works and how sometimes it doesn't work,

Sups 1:11

Okay.

Andi 1:11

But first we have to start with our pub quiz question. The P Q. Q

Tanner 1:16

The P. Q

Sups 1:17

Q. Q

Andi 1:17

If you've been listening to season four, this is a new thing where at the top of the show you're going to get a pub quiz or bar trivia style question to kind of open up the topic.

Tanner 1:29

Mm-hmm. So what movie is this quote from? Oh, no, and you get a hint. It's spoken by an ai.

Sups 1:36

Okay.

Andi 1:36

Okay.

Tanner 1:37

Okay. The quote is, I've seen things you people wouldn't believe. Attack ships on fire off the shoulder of Orion. I watched sea beams glitter in the dark near the Tan Hauser gate. All those moments will be lost in time, like tears in rain.

Sups 1:53

Is this R two? T two?

Tanner 1:55

no

Andi 1:55

I want it to be, yeah. Is this Wally? Honestly, I don't know that many movies with

Sups 2:00

Is it, is it like her.

Andi 2:01

it's gotta be a Space Odyssey. Ah,

Sups 2:05

Ah, 2001. A

Andi 2:06

one. Yeah.

Sups 2:07

Okay, so now, now give us a hint.

Tanner 2:09

It's from 1982. Oh,

Sups 2:10

Oh no.

Andi 2:11

Oh,

Tanner 2:12

Based on a book.

Andi 2:13

Oh, no. I was gonna say the only other movie I know is the Terminator. That's also not right. Uh I don't know.

Tanner 2:20

It's from Blade Runner, blade

Sups 2:21

A Blade Runner. yes,

Tanner 2:22

yes,

Andi 2:23

Yeah. I guess Androids are ai. Yeah, exactly. That makes sense. And then that's what that creepy dude says. Mm-hmm. Yeah. I saw that movie once maybe. Yes.

Sups 2:32

Which one's better?

Andi 2:33

I, you know, I didn't see the new one. Are you kidding me? I wanna joke, I've barely seen the old one. You know, I haven't seen the new one.

Sups 2:42

Yeah.

Andi 2:42

Anyway,

Sups 2:43

All

Andi 2:43

let's get. Artificial.

Tanner 2:45

Fact number one, the term artificial intelligence was coined in the 1950s.

Andi 2:52

Okay. Right off the bat, as a linguist, I'm very skeptical of that. That does not sound nearly old enough, but, okay. Tell me about it.

Tanner 3:01

So in the sense of the capability of a machine to imitate intelligent human behavior. Okay. This is the first known

Andi 3:09

Okay. Maybe now it's more reasonable.

Tanner 3:12

Okay. So nowadays this means an entire field of computer science and things like that, but in 1950? Alan Touring wrote a paper Called Computing machinery and Intelligence. This didn't actually use the term, but they were definitely talking about intelligent machines. Mm-hmm. And the idea of making a robot that has a working memory mm-hmm. And that can learn from experiments and its mistakes. Right. That's what they were talking about in the forties and fifties.

Andi 3:40

Mm-hmm.

Tanner 3:40

So the first time the actual phrase was used. Was in 1951 when a computer program that could play checkers was created and it learned by playing against humans. Mm-hmm. And so this is why it was artificially intelligent. this idea was really taking off because in 1956, Dartmouth in the US, they had a summer research project and they called it the Dartmouth Summer Research Project on artificial intelligence. 1956 Mm-hmm. So by that time, it was already well known enough an expert could get an invitation to this, and they would know what it's about. Yeah.

Andi 4:15

Yeah. I mean, it's not a very inspired term. but Okay. Like, to specifically mean like artificial in the sense of like a man made manufactured device or tool. Then Sure. But I still feel very skeptical that in all of English history, even if we just ignore middle and old English and just think of modern English, we were talking about 500 years. No one ever put together the words artificial and intelligence, and then meant it like that.

Sups 4:44

in that sense, I think this one is right, because the period that he talks about is about the time Just after the Second World War when mm-hmm. there were like a lot of research programs and it was being taken more from like a lab to wider access. You know,

Andi 5:02

so like it does make sense Because early computing, like, video games Yeah. Aren't a thing. The internet of course is not a thing. so like for them to be jumping to what to us is now feels like a very modern technology. Maybe that's where he's trying to trick us. If you think like, no, AI must be newer. Oh, actually it's from the fifties, but I think it's even older. Mm-hmm. But I

Sups 5:23

I have

Andi 5:23

to hear more facts.

Tanner 5:25

All right. Fact number two, AI chat bots hallucinate false information because they were trained to repeat patterns.

Andi 5:35

no, So AI hallucinations. Mm-hmm. This is interesting cuz when I was researching my episode, it did this.

Tanner 5:41

yeah. We're definitely gonna bring up hallucinations and lying in the next couple

Andi 5:45

Yeah. So yeah, the next few episodes this month are going to be predominantly AI generated, where the AI is actually creating the facts and the content. So yeah, it's pretty weird.

Sups 5:56

Yeah. I think it just gets stuck in its own loop. Mm-hmm. it is not hallucinating in that way that we understand

Andi 6:02

hallucinations. Right. maybe you should explain what an AI hallucination is. Okay.

Tanner 6:06

So hallucination means an AI chat bot will confidently tell you a fact And it's false and it's been trained to predict what the next word would be in a phrase or where the next sentence would be. And so it can just. Kind of make up a fact it's just designed to repeat a pattern and figure out what you want to hear, And It has no value for truth. hallucination is basically a lie But you can't really say it's a lie because it's not. Nefarious. It's not

Andi 6:38

It's not trying to trick you I mean that now we're talking about the philosophy of lying. Yeah. Right. Does malicious intent do you have to have that for it to be a lie? Mm-hmm.

Sups 6:45

Yeah.

Andi 6:45

But yeah, that makes sense. So like chatbots we're talking about like those are language based, right? Like I actually did computer generated research when I did my masters. It wasn't an artificial intelligence, but it was computer driven. And it does this exact thing where it, Can search a lot of text and then when you give a word. It'll list like the most common words that come after. And so that's how like the AI is trained. Mm-hmm. Yeah.

Sups 7:08

and it goes back to the idea that AI at the very essence of it is math, right? It's starting off with calculations, so It's trying to look for those patterns and when it can't find any, they're so advanced that they can create

Tanner 7:22

Right.

Andi 7:22

Yeah. But going back to the fact this was a big fact and we're like totally, we have bitten that it comes from patterns. So we actually have to be more skeptical of that. Mm. It reminds me a lot of like, autocorrect, Because if you do like that thing on autocorrect where you just hit like the next word. And you make a sentence based on like what the next word would be. That's what you're describing, right? Yeah. But that doesn't seem advanced enough if my, if my phone's auto correct, can do that. I would think there'd be a deeper, more complicated reason why the AI would be mm-hmm.

Sups 7:58

Yeah.

Andi 7:58

I guess my thing would be, well, what does he think we should know if it's not patterns, Right. If he's lying. So I don't know. That's

Sups 8:06

tricky.

Andi 8:07

it'd have to be something like, oh, of course it's because of this, I should have

Tanner 8:11

Right. right. So should we move on to fact

Andi 8:12

three? Yeah.

Tanner 8:13

Fact number three. AI tools have been used to discover new planets.

Andi 8:20

Wow, that's a very simple yes no.

Sups 8:22

Hey, ai, find me a planet. Here you go. Is that how it

Andi 8:26

feels like? Is that a hallucination? Did it hallucinate planets?

Tanner 8:29

right.

Sups 8:29

Oh, dang.

Andi 8:30

So I have a feeling this has to do with like finding exoplanets. Yes. And how. It's much easier for us to collect the data than it is for us to analyze it. So it makes a lot of sense if you would, could use an AI to do it because it would just learn, oh, this method's not working here, let me use this other one. Mm-hmm. It's like having an intern.

Tanner 8:52

Yeah.

Andi 8:53

The question though is have they actually done it

Sups 8:56

already? Yeah. Doubt it. Okay.

Andi 8:58

me more.

Tanner 8:59

Okay, so you're exactly right. This is about exoplanets there's been over 5,000 exoplanets found outside the solar system, and last year they found a new one with ai. And so what they do like at NASA is they take photos of the sky like space, right? Mm-hmm. And they're looking for the beginnings of exoplanets. So they call them proto-planetary discs. these are usually found where like stars are born. And so they are taking a photo of a star hoping to find a disc, which would tell you that a planet is forming, So that's a lot of steps already for human to do. And so they take all of this data and the researchers gave an AI this, and they trained it and they asked it to find proto-planetary discs and thus exoplanets. It did that and they confirmed that it was on the right track. And so they said, where else should we look? And it said, look over here. yay. It flagged the area around Star HD 14 2 6 66. Where they hadn't identified any planets. The humans checked and there was indeed an exoplanet forming there.

Andi 10:03

Planet 6 66.

Sups 10:06

isn't this like a script of some, I think it's like

Tanner 10:09

strip

Sups 10:09

some marvel.

Andi 10:10

I mean, it's one of those things where, honestly, the trick is because I genuinely believe humans it will absolutely be a fact that AI will help find planets. That's just a fact. So the question here is Have they already?

Sups 10:23

Yeah.

Andi 10:24

It sounds like it's probably real because I think AI now is on that level in NASA can definitely use way better AI than like, you know, we can get, so like, I think they, it must be real. Right?

Sups 10:38

of all the three that I've heard so far, this is the one that I'm most skeptical about for the simple reason, if AI were to do it, don't you think this would have been a bigger news? Like, I'm drawing a bit blank. I

Andi 10:49

I mean, that's not that I'm

Sups 10:50

saying like, I know everything, but

Andi 10:52

That's always a bit tricky cuz you're relying on your, memory recall. Right. You might have just missed this. I don't know. But like, it's possible. I guess maybe we would've thought it of been a Bigger news,

Sups 11:02

yeah. But it's a very 50 50 fact but anyway, I dunno, I'm skeptical about this one. Yeah. Okay.

Andi 11:08

Okay. Interesting. We have one more.

Tanner 11:10

All right. Fact number four, Sunspring written by an AI won the Academy Award for best live action short film in 2017.

Andi 11:19

wait, what?

Tanner 11:20

Have you ever heard of the film Sunspring?

Andi 11:23

No, because you invented it. Obviously,

Tanner 11:24

it's not invented.

Sups 11:25

Exactly

Tanner 11:26

It's a nine minute film released in 2016 written by Benjamin, which is what the AI said to call it. Oh,

Andi 11:34

Oh God, that's always really creepy when the AI is like, call me this. It's like, no, I'm not gonna do that. Yeah,

Tanner 11:39

So this was in 2016. So before chat, G P T and all of these advanced ones came out. Those came out in 2022. So with this one, they fed it sci-fi screenplays, and they said, Write a movie that's similar to these sci-fi movies from the eighties and nineties. Mm-hmm. And so it churned out this script it, called it Sunspring. And so they filmed it with Thomas Middleditch, who is in Silicon Valley. And that one episode of the Office, he's Dwight's brother. yeah.

Andi 12:10

Okay.

Sups 12:10

Mm-hmm.

Tanner 12:10

Yeah. He's the main character.

Sups 12:12

Okay,

Andi 12:12

Okay. I know. that guy.

Sups 12:13

know.

Tanner 12:13

that guy.

Andi 12:14

that's good context, Thank you.

Tanner 12:17

And so it's like on this futuristic ship and there's these three people living together and they're in a love triangle

Andi 12:24

Love it. And

Tanner 12:24

first line in the film is, in a future with mass unemployment, young people are forced to sell blood. That's the first thing I can do.

Andi 12:33

Wait, that's not the future. You know that it basically exists today, right? I don't remember exactly when, but I do remember when it was kind of really trendy to have AI generated like scripts. Mm-hmm. yeah, like write an episode of Star Trek. Mm-hmm. it would be pretty awful. So I guess this, Seems possible. I don't know if you guys have realized, but I am not the biggest movie fan, me missing, I know, right? Oh, wow. Me missing some Oscar and not noticing it from many years ago. that checks out that I, you know, I don't know if I would've remembered this was real, if I had heard about it. So,

Sups 13:14

Yeah. I think the time period is right. I think around that time it was really popular. Nine minute Also checks out out of. Totally not believed. If it said that it put together like a

Andi 13:24

feature. a feature Oh

Sups 13:25

film, It would've yeah. I have a feeling this might be true.

Andi 13:29

it might be true, it I think the movie exists. But I would say maybe the least believable part is that it won an Academy Award. Award because I'm sure it was awful. And the only reason it would win is because of the novelty. But I don't know if the academy is the kind of people who like novelty.

Sups 13:50

Yes. which category did it win

Tanner 13:53

best live action, short

Sups 13:55

Okay. Right.

Andi 13:56

So

Sups 13:56

did put in the word short.

Andi 13:57

Yeah, of course. Because of course it didn't win. Best picture. We would've

Sups 14:00

course. the winner? Do you remember? 2017. Let's take a wild

Andi 14:05

no, I don't even know if I can remember what won last year.

Sups 14:09

Okay.

Tanner 14:09

Well, even if you asked the hosts I think they would be confused because the winner was, do you remember soups?

Sups 14:14

No.

Andi 14:15

Was this the one where they messed it up? Yeah. Oh, they, okay. Oh, so it was not Lala Land?

Sups 14:20

No, it was the

Tanner 14:20

No Moonlight.

Sups 14:20

Moonlight. Yeah. It's that here. Yeah. So definitely. Okay. I didn't see either of those No.

Andi 14:25

No. Nope. Okay. I did not

Tanner 14:27

She says she doesn't watch movies. Maybe. Have you seen Zootopia Fast animated film that year? I have seen

Andi 14:32

Zootopia.

Tanner 14:33

See? There we go.

Sups 14:34

Okay.

Andi 14:35

so that would also make sense why we don't remember. Yeah, because it, the big thing was the flubbing of the best Picture. so no one's even talking about the AI short. But I don't know, it's, it's pretty sketchy to me that it won. I could see that it existed. Maybe it was nominated, but it won. Yeah, that's the biggest lie to me.

Sups 14:54

I have a feeling this fact is true though. I'm less skeptical about this. Yeah.

Andi 14:58

Well, before we make our official guesses mm-hmm. Our official artificial guesses Tanner, can you please repeat all of the facts and then we will discuss them and guess.

Tanner 15:09

Certainly. fact number one, the term artificial intelligence was coined in the 1950s. Fact number two, AI chatbots, hallucinate false information because they were trained to repeat patterns. Fact number three. AI tools have been used to discover new planets. And fact number four, Sunspring written by an AI won the Academy Award for best live action short film in 2017. One of those should not be believed. It is a lie.

Andi 15:50

All right. So are you still thinking number three?

Sups 15:52

Yeah, Yeah. Mm-hmm.

Andi 15:54

I feel good about number three. Yeah. I'm between one and four. Okay. I think. Four because he's trying to trick us being like, oh, you don't believe the movie's real movie's real. But it didn't win the award. Okay. Okay. And then number one, I would've thought that we would've put those

Sups 16:10

let's see, two Okay. One is the one that I'm most confident about. Mm-hmm. Because I think the date timing all of this checks out

Andi 16:19

Yeah, I mean, I kinda imagine that in the, later enlightenment. There was a lot of discussion about the future and about the nature of intelligence. And I don't know, it wouldn't surprise me if some writer had theorized or come up with this idea. of, you know, an artificial intelligence. Like an HG Wells kind of thing. So that's kind of where I would've guessed. Okay. And so that's earlier than Tanner said. So that's where I'm skeptical.

Sups 16:46

Mm-hmm.

Andi 16:47

and number two, it's like I don't know what it could be, so I'm just gonna believe it. I think it might be number four. Yeah. I don't think it one.

Sups 16:54

yeah. And I'll go with three, I just have a feeling, it should have been a bigger news

Andi 16:59

Mm-hmm.

Sups 17:00

it was, so, okay, I'm gonna go with number three.

Tanner 17:03

All right, So. The big reveal is fact number four is a lie. Uhhuh. Yes, It did not win. It was not even nominated, The winner for live action short film for 2017 was a movie called Sing. It's Hungarian from Christophe Deak

Andi 17:19

Oh, it's not the animated sing with the pig that sings. No,

Tanner 17:23

very different Yeah. No, this movie sucks.

Andi 17:26

I'm sure. It's terrible, right? It's terrible.

Tanner 17:29

terrible. So funny about it. Yeah. Right. The academy would never allow it to win. That's

Sups 17:33

That's kind No,

Tanner 17:34

Yeah. the like lines are terrible, but then they made a sequel the next year.

Andi 17:39

They should just use a, like a better bot every time and see how much better the movie gets. Then You could kind of track over

Tanner 17:46

Right. And even seeing the improvement from 2016 to 2023 with G, PT four. If you ask that to raise screenplay, it's so much better than what this is.

Andi 17:56

Okay. it does make sense. The academy wouldn't touch that.

Sups 18:00

not yet.

Andi 18:00

be like, yeah. Yet. Oh, I, I fully believe that. We'll get there. Yeah. We'll get there, but yeah, not, yeah, not yet.

Sups 18:08

not yet.

Andi 18:09

Well, I should have known,

Sups 18:10

I should have known.

Andi 18:12

Thanks for listening to this episode of I Should Have Known. we'll be continuing with more AI themed episodes, and actually we're gonna get into AI generated episodes after this one. Next week we're gonna hear from soups who got the AI to write us some poetry? So, stay tuned for that. And if you don't know, we now have a YouTube channel where you can. Watch episodes, but most importantly, you can comment and like the episodes much more directly So please like, subscribe all that YouTubey stuff. And What is your favorite AI themed movie that you've seen? And we'll see if Andy has seen it. The answer is probably not, but

Tanner 18:53

we'll

Andi 18:53

out. And as always, yes, for listening, I know some humans who talk a lot like chatbots.