I Don't Know About That - Artificial Intelligence

Episode Date: November 23, 2021

In this episode, the team discusses artificial intelligence with Data Scientist & Data Science Ambassador in the e-commerce domain and Education Lead for Women in AI Upper Austria, Katherine Munro. Follow Katherine on Twitter @KatherineAMunro and Instagram @KatherineAmabel. For a write up from Katherine of her questions and answers - including some questions we didn't get to on the show - check out this Medium post: https://tinyurl.com/5ubayv3e And if you're a German speaker and interested in learning about machine learning and natural language processing, you can view Katherine's LinkedIn Learning course here: https://tinyurl.com/yd4fr9cs Go to JimJefferies.com to buy tickets to Jim's upcoming tour, The Moist Tour.See omnystudio.com/listener for privacy information.

metal cars wooden cars which one is driven by a complete moron you might find out and i don't know about that with jim jeffries the answer is metal cars and not driven by morons oh you said which one is yeah i know i'm a moron it's only people who know that if they're listening to the podcast regularly yeah yeah I'm not dreaming about morons. Oh, you said which one is. Yeah, I know. I'm a moron. It's only people who know that if they're listening to the podcast regularly. Yeah, yeah. How is your wooden car, Luis? How's it going? The passenger side window doesn't go up anymore.

That's all right. Don't have any friends. Is it wooden windows? No, no. Also, your license plate is like completely off. Fuck again? I keep straining. Every time I drive, I straighten it out.

But each pothole, each pothole. It's good. It's because when he gets to 88, it goes into the future. And it's just that they're spinning around. I actually did put a fuzzy dice on it today. That's Mexican for testicles. Fuzzy dice. What is the fuzzy dice?

There was the fuzzy dice revolution for a while, and then people went, Revolution. No, it was a- It was a renaissance. I'm older than a lot of you. I'm not that far. But there was a time where fuzzy dice was the ultra bit of fashion.

And then there was the little tiny boxing gloves. Remember the little tiny boxing gloves? Oh, yeah. And nothing else has really had it oh no there was the couple of dream catchers the crown air freshener remember this number i'm saying hanging from the revision mirror how did fluffy dice get the monopoly and nothing else is really cool you need two of things otherwise it bounces around too much i i reckon a couple of testicles would be a bit of fun. Well, they have them on the trucks. Truck nuts. They're there at the back, though. Somewhere where you can look at them and enjoy them.

Well, if you get a camera and you put it underneath your car, then you can see the truck nuts. That's true. That's the thing. It's always these simple, like whoever invented the Rock Pet or whoever invented the Fluffy Dice, there's someone, there's some family now sitting on Fluffy Dice money. Oh, yeah. They're killing it. Oh, Grandad invented the Fluffy Dice. dice there's someone there's some family now sitting on fluffy dice money oh yeah they're

killing it our granddad invented the fluffy dice we've been really good life's been amazing yeah they're sitting up in the hollywood hills looking down on the rest of us yeah who haven't invented fluffy dice all their furniture's made of that same fluffy material it's all we know i believe fluffy dice fueled the chinese army and now they're a threat because they were all made in China. You never get a good quality fluffy dice. Has anyone ever had an Italian made pair of fluffy dice?

I've never had a pair of fluffy dice. Probably because they count their numbers differently. They don't use the dot system like we do. Right, right, right. Who doesn't the Italians everyone uses the same dice

no no no Italian dice is different it's just a couple of hands going like that's two like four yeah and then six just says you didn't see nothing they're Italian American dice dice, those ones.

They're different. Jersey dice. Jersey dice, yeah. Jersey dice. They're musical. So I've had a fun week. I went to Disneyland and lost my child and another child, one of his friends, for maybe five minutes, maybe five minutes.

And when you lose a kid, you're not that worried because they're nine years old. They're going to find somebody. They're going to sort it out. It's like a two-year-old family. But I was there. I won't say the name of his mate, right, but I had the mask on so no one was bothering me in Disneyland. I was just walking around.

No one was bothering me. But then I lost my son and I was like, Hank! Hank! Hank! Now, it turns out my voice is fairly distinctive. So all of a sudden people started noticing me. I'm standing there yelling in the middle of a parade. Hank.

Two fucking people stopped me for photos. Stopped me for fucking photos when I'd lost a child. And you took the photos with them, didn't you? I did. You've got to be good to your friends. But pick your fucking moment. Why don't you just follow me around until I find the children? I tell you what, find my children.

Yeah, help you look. Yeah, find my children. Then I'll do a video. I'll call your friends a cunt. That's what people come up to me all the time. Can you do a video and say that Dan's a cunt? And I'm like, sure. And then I go, Dan, you're a cunt.

And I don't question who Dan is. Maybe Dan's having a hard time. Maybe Dan's sitting there in a chair going through chemo and this is the last thing he fucking needs. But I always help you out. So you stop me in the street, I'll call anyone a cunt. Just get your phone ready, know how to use it, and I'll call your friends a cunt.

That's what I do. I give. I give back all the time. I like how you specify know how to use it. Oh. Oh, use it. Oh. Oh, my God. Oh, my God. When we used to do the meet and greets and we used to make people

take photos themselves, like the person in front of them would take the photo of the person behind them. Now I get Jack to do it for the most part or someone who works at the theatre. I know every phone. I know how they all work. But we used to get people, I'm taking a photo of myself now. Okay, how do I turn it around? It's a fucking iPhone.

We all have one. It's also your phone. No, no, no. They're in the photo. They've asked another person to do it. Often it's them just doing the picture. Yeah. But I feel like people are a little bit more savvy.

Back in the cold days of 2016, fuck me, we had some problems. What have you got for us, Jack? I have Jack of the Trades. Jack of the Trades. Jack of... Fuck. I feel like we're about to get sued. I feel like we didn't... Okay. That's royalty free.

So we're good. That's from Apple. If anyone's suing, it's Apple. I forget what Jack of the Trades is. That's when I give you some headlines of the week, and you guys tell me what you think. All right. First one. This could be a whole podcast.

It could be, but it's just a tiny segment that people will skip. Wow. News story one. Britney's free. Woo! Yeah, Britney's free. I'm a big fan of Britney being free. I think she should have been free. But I also follow Britney's Instagram where she's very clearly mentally ill.

You're the biggest star in the world. Stop sending naked pictures. You must want attention, Britney. She's got naked pictures on there? Yeah, she's not naked. Check them out. She's not naked. What's her handle?

Did she post something recently that I haven't seen? Britney Spears. What's her handle? Might she post something recently that I haven't seen? Britney Spears. What's her handle? Might be Britney Spears. Britney Spears won. It was a joke. Okay.

Britney Spears. All right. I'll find you some nudie rudies. I'll follow her. Okay. So there we go. You're Britney Spears. You're just going off to bed. Oh, yeah.

This is her baby's feet and her. No, no, no, no. I'll find you. She's thinking about having another baby. Oh, no, no. It's good that she's having another baby. Now, often it's feet and her. No, no, no. I'll find you. She's thinking about having another baby. Oh no, no. It's good that she's having another baby. Now often it's just fun things. Oh yeah, she's got crazy eyes. It's a lot of Britney just spinning. It's a lot of dance videos. Britney's spinning around like a

nut job, right? We've seen you dance. That doesn't mean she should be. She should still be getting all of her money. Oh no, no, no. I'm free Britney. My favorite were the videos of her like set up with an easel and oh wow, yeah. What's Britney up to? That's just her cup and her boobs. But you were the videos of her like set up with an easel and oh wow. What's Britney up to? That's just her cup and her boobs. But you remember those videos where she

was like painting and there was like classical music in the background and you expect it to be some like really good painting and then the reveal is like a three year old. It's just butterflies and you're like okay. She does a lot of these where she just kind of rocks back and forth. Is that the biggest star in the world?

I don't follow her, so I didn't realize that she was. That's a lot of, there we go. There she is. She's not naked. She's got tiny flowers on her. Yeah, because she'd get deleted from Instagram. This is why if she was fully naked, I'd say don't give her her money back. She's not ready.

The little flowers are the only thing that's keeping me going. Yeah. I think Brittany should give her her money back. She's not ready. The little flowers are the only thing that's keeping me going. Yeah. I think Brittany should have her own money. It's her money to piss up the wall, and I'm looking forward to her Brewster Million-ing this whole fucking thing. We're going to find Brittany by the side of the road with a bit of cardboard going. No.

She's got tons of money. She doesn't have a husband? Who's the baby? She's engaged. Whose baby is that? To who? I don't know. Sam something.

She knows babies. Did she have a baby with Kevin Federline? Yeah, she had two kids. Oh, with Kevin Federline? Yeah, with Federline. Federline, yeah, yeah, yeah. But they're like teenagers now. I just follow her.

If she got a father, she'll be happy. Good luck to you, Brittany. I follow you. 36. My wife's a big fan of your music. We listen to you in the car. Good luck to you. Good luck to you. She's going to be a Of your music We listen to you In the car Good luck to you Good luck to you

She's gonna be a chef She's gonna be It says artist Mama Pray every day Chef in the works She's taking a cooking course Yep

That's what she did That's the first thing She spent her money on Frozen dinners They should have that There's that Amy's brand Not that good But if you had Brittany's

I think Frozen burritos I think She should have a restaurant that just serves asparagus and it should be called Spears. No, it could be Spears, but it could be all different kinds of Spears. There's like... Like kebabs. Yeah, things that are on Spears.

Yeah, or the Indian meal. When you fight about the check, you're given a Spear. Good luck to you all. Free Britney Spears. What do you mean? What do you mean a spear? That's not... Like throwing spears

at each other. Yeah, I think that's more of a Game of Thrones thing. It's not a real restaurant. Okay, whatever. So I don't know if you guys remember last year, but there was a couple...

I do, I do. I forgot we were doing your segment. We were all... Good news headline. We were all home for a while. I don't know if you remember last year,

but there was a virus. Okay, next headline. But there are a couple sightings around our airport LAX about a man with a jetpack. And pilots were seen as they were flying in to go, what's going on with this jetpack guy? And the FBI looked into it

and they finally got a good picture of the jetpack man. And it's a full body balloon of Jack Skellington. Wow. There's no jetpack. Are you sure it's not Chris Angel? Well, he's not in Las Vegas.

Why did they think he had a jetpack on his back? Because it just looks like a dude floating in the air. But Jack Skellington's so thin, he doesn't look like he's got a jetpack on his back. I thought it was Jack Skellington with a jet pack. Oh, did he have a jet pack on him? He might have. That's how he was flying.

Remember that, was it Chris Angel? I don't know, who was the other guy? David Blaine? Yeah, when they made Shaq float over the- Shaq floated? He supposedly made Shaq float over a house, and then there was just a balloon of Shaq, I think. But that's the magic in there.

No? I don't know about that one. I think it was Chris Angel. He's a good guy. that's the magic in there. No? I don't know about that one. I think it was Criss Angel. He's a good guy. It's an illusion. Yeah. Oh. We're going to have the longest lunar eclipse in centuries next week,

or this week. It's going to last three hours, 28 minutes, and 23 seconds. And the moon name, it's called a beaver moon. Wait a minute. A solar eclipse is when the moon crosses in front of the sun. A lunar eclipse is when the sun crosses in front and the moon name it's called a beaver moon wait a minute which is a solar eclipse is when the moon crosses in front of the sun a lunar eclipse is when the sun crosses in front of the moon that can't happen what about what about a total eclipse of the heart when's that gonna happen that's in two weeks oh shit so what's what's what's a lunar planet in

the way what's a lunar eclipse a lunar eclipse yeah i don't know anything about eclipses yeah the solar eclipse is the... Jack's the one with the article. Is the moon crossing in front of the sun. What's a lunar eclipse? I think a lunar happens at night, right? Lunar eclipse occurs when the moon moves into the Earth's shadow. This can occur only when the sun, Earth, and moon are exactly or very closely aligned with the Earth.

So the moon looks kind of like really orange. We won't... So we won't see the moon? We will see. It's going the moon looks kind of like really orange. We won't, oh, oh, oh, oh. So we won't see the moon? We will see it. It's going to be like reddish orange. A beaver moon.

You'll see a beaver moon for three hours in 20 minutes. Why is it called a beaver moon? I don't know. I don't know how moons get their names. They always have some crazy names.

There's like a blood wolf moon. Next week on the podcast, we're all going to be talking about the red moon. That would have been a laugh. Blood moon. Wait, what did you say? Blood wolf moon? Blood wolf moon. That sounds way cooler than beaver moon though. Yeah, guys we're all going to be talking about the red moon that would have been blood moon wait what did you say blood wolf blood wolf that sounds way cooler than beaver moon though yeah beaver moon's hilarious beaver moon is like beaver moon's hairier yeah it's like when a girl moons you out of a car

window but she bends over a little too far me me and my brother at one stage when i was young we got very into mooning people and my brother just just got What were you got into it? Well, my brother had just gotten his license and we did a lot of trips to like basically Australian equivalent of 7-Eleven which they have loads of now, but they didn't have them back then. And we were getting a lot of slurpees. That's all we needed to drive to, you know. My brother had his license. Do you want to go get slurpees? So we'd get in the car and off we'd go. We'd go to McDonald's. But you knew you were going to be mooning people. Well, we grew tired of all get Slurpees, so we'd get in the car and off we'd go. We'd go to McDonald's. We'd go get a Slurpee. But you knew you were going to be mooning people. Well, we grew tired of all the Slurpees and the Big Macs.

And then we found out that on Sunday we'd read in the newspaper where all the weddings were going to be. And we would drive. This is very complicated, actually. Yeah. So Scott was 16. I was four years younger. I was 12 or something like that. So in a $400 car, not safe, we'd find where the weddings were

and then I would drive by with my ass hanging out the window. So fun times were had by all. Just by you and Scott. Yeah, until I got a bit of rice caught up there. Expanded. People would live in there. Expanded and blew it up. That's how you get your hemorrhoids.

But I like to think the people who I mooned as we drove by had a good time. They probably went home. Most likely. That broke up the ceremony a bit. You know what might be happening now is people are hearing this for the first time and now they're like, you know what? That was Jim Jefferies mooned at our wedding. If you had a wedding on the northern beaches of Sydney in 1990,

you got yourself a mooning, ladies and gentlemen. No, we didn't do that much. Maybe four times. That is a lot, though. We got really into it. That is a lot, though. It was all in one day. Yeah. A couple of them were.

We really found those weddings. The segment has gone gangbusters. It's going good. We're getting a lot of chats in here. It's good. But also, you have a moon. You moon out like a moving vehicle. It feels going good. We're getting a lot of chats in here. It's good. But also you have a moon, you moon out like a moving vehicle. It feels a bit,

you're not stable. Yeah. It's a drive by moon. You have to brake fast. Dangerous. I wouldn't do it in America. You'll get shot. In Australia,

it's a much more innocent mooning country. You get stabbed. I haven't been mooned very well for a few years. I think mooning's out of fashion. I think you'd be done for sexual assault now if you mooned someone. Back in the day, it was just a laugh. Watching a teenage boy's ass hanging. I was going to say, that's pedophilia.

Yeah, I did it. You're the victim. I did it and everyone else got arrested. Don't look at my ass. Russia's in hot water. Which is weird because it's so cold over there. They must have done something really wrong. They tested out a new- Climate change.

They tested out a new anti-satellite missile and blew up one of their own satellites, sending debris everywhere in space. And now it's orbiting. Everyone in the International Space Station had to go into their cubby holes or whatever for protection because it's like a scene from the movie Gravity when there's

just space debris just flying around. So that's going to circulate the it's going to orbit the Earth for centuries. What if it blocks the Beaver Moon? We're going to see space debris in front of the Beaver Moon now thanks to Russia. So they blew up their own satellite. They shot a missile I believe from Earth. And what were they trying? Were they trying they blew up their own satellite they shot a missile i

believe from earth and who what were they trying to were they trying to blow up their own satellite they were testing to see if their anti-satellite missiles would work so they blew up their own satellite to not piss anyone off but it's pissed everybody off so russia seems to do that yeah i feel like everyone no one's happy with russia right i always think because i went up like call of duty or something like that, what must it be like to be like German or Russian? And in every computer game or in a Bond film, what are we, the bad guys again? Do they have movies in Russia where it's like,

here we are hunting down some dumb Americans? They must have some. For sure. I think China has some movies where Americans are the bad guys. Yeah, where it's just like some bloke just eating a hot dog. Rawr, rawr, rawr, rawr. When I was in Disneyland, I saw one of the great, I've already told Forrest and what's your name?

Jack. I've already told Forrest and Jack this story, but I saw it. Whenever you're in like Disneyland, you see all these fat fucks, right? Fat fucks who are on their mobility fucking scooters. I'm all right with the elderly, right? You want to go out with the grandkids, but it's just like I don't want to walk. There's no Disney for you.

If you can't fucking walk it, you can't fucking do it. You've got to be 60 or over to have the mobility scooter. Anyway, so there was some Midwest fat fuck on his fucking scooter. How did he know he didn't have a disability? He was about 350 pounds. Yeah. And also this T-shirt says he didn't have a disability, that he was just a fuckwit.

Not disabled, just fat. His T-shirt said this. He had like the straight out of Compton font. You know how a lot of people are like straight out of whatever, straight out of prison, straight out of Sydney, straight out of Compton font. You know how a lot of people are like, straight out of whatever, straight out of prison, straight out of Sydney, straight out of whatever. He said, his T-shirt went straight out of patience. Yeah. With liberals, right?

You can't add a bit on the bottom of the with liberals bit. You can only have the block letters. If your parody doesn't fit within the parameter of the words, you can't just go with liberals and cunts and the fucking woman down the road. That's every conservative comedy shirt. It's just way too much info. You know what? And I'm not.

Look, I consider myself. I'm a bit out of patience with fucking everybody. Everybody fucking annoys me. But I am out of patience with some bloke just sitting there going, these liberals. What do you contribute, you dumb fuck, with your fuck? Anyway, he was in me way for a bit. Kelly, what was the t-shirt we saw in Atlantic

City? Oh yeah, the first guy I saw in Atlantic City, his shirt said if you're not into oral sex, then shut your mouth. And I was like, we're here. That's a joke. That's just like, it's kind of violent. It's like, if your mouth is open, I'm shoving my dick in there.

Yeah, but I tell you what, you didn't say anything to him, did you? I kept my mouth shut. You kept your mouth shut. In our distance. I told you what we were talking about on the little private episode we did, which went very well. I enjoyed that. But there's a lady who's got the Jim Jefferies logo,

the two J's back to each other above her arsehole but it written underneath it april 18th oh good on you good on you you're bloody you you got the anal sex day above your arsehole and a tattoo bit of fun russia's are not water yeah all right i'll give you one more headline you guys get to choose do you want a cool science headline or a fun entertainment headline i want a fun entertainment headline about cool science okay i'll do my best um give us both okay i'll give you both i'll give you the short one first no we can do one okay so i'll do both i don't let forrest fool you i think the entertainment one's really short uh journey announced a new world tour with Billy Idol and Toto. Oh, cool.

Who was the band? Toto? Oh, Journey. We had Steve Lukather. Oh, Journey. We had Steve Lukather on our podcast. Oh, I thought you said Germany was announced a new world tour. And I go, what happened the last time they did it?

You know what I mean? They call them tours now instead of wars. Well, what's interesting about that is because Steve Lukather's son, Trev, is engaged to the keyboardist from Journey's daughter. I think his name's Jonathan Cain. But they're engaged, so I wonder if that's how the tour came about. It's like, hey, we're going to be family soon. Might as well.

That's a fun tour. Tour together. Fun tour. My friend Nick Dalyaly who played rodney on legit he came to the show the other day and uh he had his birthday just recently and the thing that he wanted for his birthday more than anything else is the luke of the biography man oh really yeah yeah he got it that was his best gift there's a photo of him on instagram like yeah i got it oh i bet i could probably get something signed for him yeah Yeah, he'd love that. There's an episode in legit when we wanted to do this thing

with Dan Ratzenberg where Rodney was on the whiteboard writing something, just writing something down, and then John was going to come over and do a diagram that was length of time of your relationship versus blow jobs, like how they go down with this graph. And so I'd written out this graph for John to write. And I said, Nick, you just write something on the other side. And I said, Nick, write whatever you want. And so John's like this and you go, Nick's standing in the way of what's written there. And this actually happened when you watch the footage, it actually happened because we didn't know what Nick was going to write.

And it was like this. He goes, this is the amount of blowjobs you get when you're in a relationship. And then eventually you're dead. And this is a picture of her dancing over your grave. And we're all like, oh, like this. And then Steve goes, what are you trying to tell us, Dad? And then Nick moves out of the way.

And then John just reads it. Docking rocks. Like, Nick likes old school rock and roll. Docking. Docking, man. He likes docking. And I was like, who's docking? And Dan Bacchanal looked at me. He was just so much angry.

He goes, you don't know docking. It rhymes with rocking. Yeah, it was rocking with docking. I think that's the way they said it, yeah. Yeah, yeah, he likes docking. Never heard of it. So if you can get me an autograph of docking as rocking. Yeah, it was rocking with docking. I was like the way they said it. Yeah, he likes docking. Never heard of it. So if you can get me an autograph of docking as well. Yeah.

First I have to find out what that is and then I will get rid of it. D-O-K-K-E-N. Oh, docking. Okay. Yep. I have seen that. Docking rocks. Give me the next one.

Last one, Jack, and then we'll start. Last one. A new state of matter was discovered called super ionic ice. And basically at extreme pressure... Why did you put the inverted commas? Because I have it quoted on here. I just read exactly what I'm saying. You read one of them too.

It's just the end. Because this one's sweaty. Basically at extreme pressures, ice can remain solid, but have a temperature hotter than the surface of the sun. And the ice isn't clear. It becomes black because light interacts

with it differently. So it explains how the core of Neptune and Uranus and ice planets exist. I think it's cool. Great. Told you I should have done one. That was a legitimate yawn. As it was coming to me, I thought,

I want him to hear it. Told you. Hey. Okay. Please welcome our guest today, Catherine Munro. And now it's time to play. Yes, no. Yes, no.

Yes, no. Yes, no. Judging a book by its cover. G'day, Catherine. Thanks for being on the show. I've been told that you stayed up very late for us and you're in Austria right now and that you're an Australian in Austria.

That must be fun. That's right. That must be fun. Yeah. Two countries are named so closely together and they hold so much things in common. What's in common? Well, the spelling's almost right there.

Yeah, okay. Exactly. I always think that's the reason I almost right there. Yeah, okay. Exactly. I always say that's the reason I moved here. The paperwork was easy. Yeah. Okay. You've got pictures of Austria behind you. You're fully into Austria.

You're right there. There's nothing I can tell from behind the thing. Okay. So you're a very poised person. I'm going to say that you're a professor, that you lecture people. You have to ask questions. Are you a professor who lectures people? I have and I do teach people, but I'm not a professor.

Oh, you're not a professor. Are you involved in the world of science? Yes. Okay, the world of science, which is everything. What we're going to talk about today, you use every day, I would imagine. I guess so. I don't know. Ah, the hallowed tushy 5.0.

Exactly. I was one of the principal research and developers for this. All right. The fact that she used the sentence principal researcher developer means that she's involved in something more interesting than stand-up comedy. I'll tell you that much. All right. Principal, something I use every day.

Is it a product? It's in very many products. Oh, it's in very many products. Oh, okay. So it's a thing. Okay. So is it, are you an expert? We've done water.

Water? Yeah, water's in everything. Everything's sun. That's true. You use water every day. Look, she's drinking it right now. Maybe it's water. Maybe it's water.

We haven't done water. Oh, you mean we've done water quality. Okay. How many different episodes of water can you do? How about this? You have a car where it's very heavily used. Oh, gasoline. No.

I don't know what car you must have. I have a muscle car. Oh, it's electricity. No, no. Yeah, it's another. You have two cars. I have two cars. The electric car. Yeah, the electric car. Made by our old buddy Musk. Musk, Tesla. She's an expert in Tesla. This, what she's going to talk about

is very involved in that car. Oh, oh, wheels. Okay. Ooh. How have we not done the wheel yet? The wheel seems like the first invention we should chat about that. It's going to take too long.

Today we're going to talk about artificial intelligence. Ah, yes, yes. AI, they call it. They do. He gets one point already. I know a girl who was in the movie Artificial Intelligence. Frances O'Connor. I haven't met her in many years.

Sounds like you know her. No, no. A girl I dated back in Perth 25 years ago was friends with her, and so I met her. But anyway, that's neither here nor there. No points for that. All right. Hayley Joel Osman was in the film.

Let me introduce. Catherine. Steven Spielberg directed. It was a movie. Catherine Monroe is a data scientist and data science ambassador working in the e-commerce domain, conducting research as well as corporate training in AI, machine learning, natural language processing, and data science. She has a background in computational linguistics and deep machine learning. Catherine worked in research and development for Mercedes-Benz

and the user voice interface team and completed her master's at Germany's prestigious, I don't know if I'm going to say this right, Fraunhofer Institute. In Stuttgart. Yeah. Specializing in building natural language, understanding systems. Catherine is a speaker,

moderator, writer, and editor. She has education lead for women and AI upper Austria, part of a global network aiming to increase the diversity of voices in AI and a volunteer mentor at female coders who run free tutorials, teaching women how to code on social media. You can find her at katherine a

monroe and monroe is m-u-n-r-o no e at the end and then instagram katherine a mabel different last name oh i'm a bow okay sorry he's not the middle initial hey mabel no it is katherine a is your middle initial right then monroe because that's her Twitter. No, no, it's pronounced Michael Jafox. All right. Well, I'm sorry. I'm the worst. I'm here for you, baby. All right.

Well, we'll have those up on the screen, too, and they'll tag, and everyone will be able to find it. All right. What we're going to do. Oh, Catherine, just tell us a little bit about how you got into AI, and then you ended up in austria and all that stuff yeah so it's uh quite complicated i mean i tried and failed like

five different degrees wasn't sure what to do found linguistics in the end uh thought that was awesome studied that at newcastle gym maybe you know it i do and uh yeah studied that and then decided i wanted to specialize did some research and found this thing called computational linguistics in Germany. So I went there to study that. And, yeah, I had some connections across the border in Austria, so I came here to work. Forrest used to shag a bird in Newcastle.

Okay, you say that every time it's brought up. No, we're not going to. How many times has it been brought up? Many times. Any time Newcastle is brought up. Newcastle's brought up every episode somehow. I can't remember it being on Newcastle. No more stories. Shut up.

Okay, so I'm going to ask Jim a bunch of questions about artificial intelligence and at the end of that you can grade them on accuracy 0-10, 10's the best. Kelly's going to grade them on confidence I'm going to grade them on etc. If you score 21-30 AI 11-20 AI 0-10, 10's the best. Kelly's going to gain my confidence, my grade amount, etc. If you score 21-30, AI. 11-20, AI.

0-10, AI. The thing is, they stand for different things. Yeah, I imagine. 21-30, what do you think it stands for, AI? 21, that's the best score? Yeah. Actual intelligence? That would have been a good one. We can make it that.

I was going to say artificially intelligent. What about AI 11-20? Ah. Actual intelligence. Ooh. That would have been a good one. We can make it that. I was going to say artificially intelligent. Yeah, actual intelligence. What about AI 11 through 20? Ah. Ah. Intelligence. I think that's better. I was going to say Allen Iverson.

But, all right, zero through 10, AI. Asshole intelligence. No, that's pretty good. All right, we'll switch into that. Ass. And then actual. Okay. All right, cool. Thanks for writing those, Jim. No worries. All right, we'll switch into that. Ass. And then actual. Okay. All right, cool.

Thanks for writing those, Jim. No worries. All right. What is artificial intelligence? Artificial intelligence is many times referred to as AI. Another point. Artificial intelligence is when you make a computer start to, like, have things not for itself as such. But what it does is you put intelligence into making jack laugh for some reason why are you laughing so much jack

why is no one else laughing jim's giving it his best all right do you want to move down to this mic um so so it's it's it's like it's like your navigation system in your car that's like the peak of it and then that's the peak? yeah that's the peak of artificial intelligence

when they brought out the fucking TomTom man I thought we're in the future now that was the thing I used to work at a Stratfield Car Radio

place in Australia where I sold mobile phones and they brought out the navigation system we couldn't sell any because no one believed that it would work, right? It was like a $7,000 thing you had to put in your car. And then I used to make the joke, I go, you use a male voice

or a female voice, the female voice will give you the wrong directions. And we all laughed and then no one bought it. In which domains is AI used? Right? In which domains is AI used? In computers. And that's where it's used. Right? In which domains is AI used? In computers. Ooh. And that's where it's used.

Name some types of AI you interact with every day besides your navigation system. Speak and spell. Every day you do that? Still use that. Okay. How do you think my spelling's improved? No, your laptop would have a bit of it.

Siri, she's artificial intelligence all day, right? She's there. I like when you get excited, you thought of something. Fucking Alexa. Alexa, they're all women, the artificial intelligence. No one ever goes, Brian, turn on the lights. What is an algorithm? Ah, well, I can give you an algorithm to tell you what the algorithm is.

The algorithm is the difference. You know when people are coding in computers, right, and they're typing all those different things? It might look like gobbledygook to you, Forrest, but when I see it, I go, ooh, they're making the game pitfall for atari okay how do machines learn um uh by us inputting things into them by going here's a bit more information here's a bit more information and soon what will happen is the computers will start teaching each other and then we're fucked like we're speaking to you right now you seem like a very nice lady but you do know

you're about to kill us all right i've seen the terminator movies you're the bit where the terminator comes back and he has to kill you because you're about to kill us all okay um we'll skip this we'll get back to this other question what is deep learning ah it's um that's when you're high and someone says something that spins you out and you go that's bloody good, man. That is good. Okay. What is the Turing test?

Turing. Alan Turing. Oh, yes. Alan Turing. I know. You've seen the movie. He made the first computer. He was also a homosexual.

And. You get points for that. Yeah, but that might be the test. Yeah. This might be the test. Yeah, okay. This might be if a good-looking bloke walks by, I go, ooh, he's all right. Did I pass?

Did I fail? It's all your perception. That's the question. Well, what is a Turing test, and do you think you can pass it? I don't know what it is, but I do think I could pass it. So give me some confidence, Kelly. Yeah. What is the Chinese room experiment? Ah, be careful.

Thank God this is not the live podcast. First of all, the room's made it up of folded paper. And it's whether you can get out of it with two sticks. I have a lot of questions here. I think I should just skip ahead. We'll get to all these questions. Don't worry, Kevin. I think you're already going really, really fast.

No, what it is, is it's the same as the cat in the box test, right? So is there a cat in the box? You don't know either way. Is there a Chinese person in the box? What do you do? What is an AI winter and are we in danger of another one? An AI winter is. There's no way you know any of these.

No, no. An AI winter is when you go, oh, God, it shouldn't be cold this time of year. But it is. But it is. Okay. And you start the sense like this, I reckon it's too cold. Okay. Oh, wow.

What is the difference between weak and strong AI? I don't know why I'm asking any of this.'s too cold. Okay. Oh, wow. What is the difference between weak and strong AI? I don't know why I'm asking any of this. No, no. Okay. So weak AI is the stuff that we have to input the information a la the speak and spell. Strong AI is where it's like your PlayStation. You go to bed, you come down, it's downloaded a game for you.

You haven't even fucking done anything. It's updated all at Call of Duty. That's some strong-ass AI right there. What is, quote, the singularity? You've heard that term, right? Yeah, yeah. No, no, no. It's a dating page that Jack goes onto.

That's not a bad name for an app, honestly. What was that, Catherine? I knew he would say something like that. She's a fan of the show. I knew it was going to be about Jack. Poor Jack. How does... That was the saddest I've ever heard you, Jack.

That's fine. Yeah, singularity. It's very hard because he's the only person on the app. I made it too. He checks in every day. All right, there's me. Swipe left. We'll get to all these, but I'm going to skip a few.

How does playing Floor is Lava help advance machine learning research? All right. Okay. This is easy, right? Because it gives the machine some parameters that it can't, you know, if you go the floor is lava, the claw that picks up things, a lot of the AI is claw involved, right? If you watch a car going, mate, they go through the thing

and there's all these things with claws coming out. That's what, if you look at Terminator 2, they had to eliminate that hand. It's all claw based. And so the claw, you can pick up things, but not on the floor because the floor's lava. So those claw machines where you pick up like the stuffed animals? Oh, that's very basic AI, Kelly. That's very basic AI. That's weak AI. Just wanted to make sure we had that in here.

That's weak AI. The picky up machines. The picky up machines. Let me ask you a couple more here. How does Amazon's Alexa and other smart devices understand what you want? Well, for me, I find it very difficult because i haven't figured out my accent in this country yet so they figure it out if i do an american voice is how they really figure it out um no because keywords it's all about keywords and that's how the machine learns for us so if you go alexa turn on the lights they hear turn lights and then they do a bit of guesswork, right? They go on.

They turn them off. Oh, whoops, that's not what I wanted. No, no, no, no, no. They go, turn, and you go, oh, you just had the lights off. Probably turn them on. Learning. Wow. Yeah.

Here, I'm going to ask you two more of these. I don't use any of those, you know. I don't use any of them. My wife just looks at the phone and goes, Siri, set my alarm for 4.45, please. And it just fucking does it. I go, oh, and Siri, and the phone just stares at me. It's probably in the settings. What is?

I don't know. I move with the future. What is the trolley problem? You want a hint? Yeah, it's that fourth wheel that always goes a bit wobbly. That's a different podcast, yeah, sorry. Okay. No, give me a hint on the trolley problem, I'll tell you. It's related to autonomous vehicles. Oh, okay. So when you've got like your Tesla and your Tesla knows how it's driving and stuff

like that, it can pick up cars, but it can't pick up trolleys, man. Can't recognize trolleys. It's in my car. Like I like to look when you go through. I was in the car with a friend of my son's and he goes, I know a great hack for this car, a little Easter egg for this car. And on the Tesla, if you hit the auto driver, which you may go twice, one, two, if you go four times, one, two, three,

four, like that, it'll start playing the more cowbell sketch and the road will become rainbow. What? Yeah. I'll do it for you on the way home, Jack. It's quite a trip. No way. Yeah, yeah, yeah. It'll do that.

And I was like, oh, yeah, what's your Easter egg? Just tap it four times. I'm like, I'm trying to drive here. What is algorithmic bias and why is it dangerous? Okay. It's the same as, you know, if an algorithm. Okay. So let's say when you get your credit score and it starts saying that you

haven't got enough credit because you haven't got credit to begin with. So the computer is biased against you, but it doesn't know you in person. So if you met me when I first tried to buy a house, I couldn't buy one because I didn't have any credit cards because I didn't need them because I had money and I was just taking money in my ATM. But then I had to get a credit card to build up my credit to be able to do it. You see what I mean? So the machine was biased. The computer says no, right? So computer says no in comparison to if the computer

just got to know me, it would have said yes all day. Would have said yes all day. Yeah. Would have said yes all day. I'm, you know. It's up for debate. All right. Will, last question. Will AI steal your job? Not necessarily your job, but jobs.

Have you seen Siri do a stand-up set? Fuck, it's some boring ass shit. I mean, in general, the world's like jobs. You know when people tell you to turn the lights on all day? What's all that about? All right, Siri, you fucking hack. No, it will steal many people's jobs. Yeah, eventually. You know when people tell you to turn the lights on all day? What's all that about? All right, Siri, you fucking hack. No, it will steal many people's jobs, yeah, eventually,

and some jobs have already been stolen. Eventually, so we're going to have AI is going to start driving the trucks for us. As soon as we have the trucks driving for us, then we stop having the truck stops, and then these little towns are going to die because the truckers aren't going to stop there, but our deliveries are going to get to us more and more. We're all going to become Wally. Wally is the future. We've already got the fat thing down.

We're already being fat. Oh, you mean the people in Wally, not Wally itself. Wally was the robot, right? Yeah, yeah. Wally's AI. Wally's fucking with these emotions. Should have said Wally earlier. Picking up garbage.

Also, he's just a ripoff of Short Circuit all day. He was a small Short Circuit. Okay. Catherine. Short Circuit was AI. Short Circuit. Put that in there, yeah. He started to get feelings.

Short Circuit, towards the end, he was dying, and we were all worried he was just a fucking robot. Fucking Short Circuit, man. Hi, Catherine. How did Jim do, zero through ten, ten being the best on artificial intelligence knowledge look i know people always a bit too nice to jim on this show and i would really like to i'm sure you did your best but it wasn't great

you made up some points on the last one so i will give you a five wait wait wait which last one short circuit the track he had a couple buzzwords in there that were kind of getting on the right track. Five. Well, see, this is, this is,

I know I'm still being nice. This is the thing with AI. I don't have to know much about it because eventually it'll do it for me. Eventually AI will give me the knowledge that I need. You should have said at the beginning, we shouldn't have asked anything. 10. Yeah.

How do you do on confidence, Kelly? 12. Wow. Yeah. Extremely you do on confidence, Kelly? 12. He was extremely confident. 17 total. Give me the Turing test right now. I don't know what it is. You don't even know what it is, yeah. I can do it though.

Okay, so that's 17 points. If I know it, then I'll get a perfect score. I'm going to give you a 1. All right. And et cetera. That means you have 18, which you're... What we've got to start doing is we've got to start giving me some type of monetary Yeah, that's probably true. I'm going to give you a one. All right. And et cetera. That means you have 18, which you're, ah. What we've got to start doing is we've got to start giving me some type of monetary punishment for doing badly on these scores. We'll not do that.

I don't know. There's got to be something for it. Make it 200. We'll do Bitcoin every week. Yeah, here's some Ethereum. All right, Catherine, let's get to these questions here. What is artificial intelligence? Jim said it's when you make a computer think.

Navigation in your car, peak AI. What did you say, TomTom? TomTom. Yeah. What is artificial intelligence? Well, I mean, it's a little bit tricky to define because different people have different answers. But very generally, it's about trying to simulate or mimic natural intelligence, which would be like what we get in humans or animals, but instead in machines and in particular in computer programs. But one of the reasons it's tricky is that the definition kind of keeps shifting, like what we used to consider AI back in the 1950s, these sort of simple rule-based expert systems.

We wouldn't really consider them AI anymore because we've come so much further. So it's a bit of a shift in goalposts. But, yeah, that's general. Is a calculator AI? No, I would not think so. I've always been amazed by those things. Yeah, I mean, you've really got me there. I don't know how they work.

Yeah, they fucking did. They've been around before AI was able to do that kind of calculation, so they must work some other way. Do you remember in an Australian high school when you got to about year seven, you went through year one to six at carrying the one and all that type of stuff, and then like in year seven they gave you a Casio calculator that we all have the same calculator, and then you were like, why the fuck did we learn all that shit before?

This thing's marvelous. I'm going to pass every exam. But then the problem is everyone's got a calculator. Oh, God, I didn't pass that. Well, because now you know how it works, so you need to know that. What if you're on a boat and then you didn't have a calculator and you need to calculate something? I reckon there'd be a lot of pi involved in AI.

I can't remember what pi was used for, but 3.14. Okay. At a point. In which domains is AI used? Jim said computers. Well, he's not wrong. Basically, I mean, the bulk of the answer for this one, I would answer in the next question about the types that you like interact with every day. So, I mean, the bulk of the answer for this one, I would answer in the next question about the types that you interact with every day. So, I mean, I just brainstormed a bit of a list of AIs that I guess you would use every day. And you've got some of them, like definitely Siri and Alexa and stuff talking to your phone.

When you search the internet, like when it's auto-completing your queries and actually serving the results. When you're consuming stuff on Netflix or YouTube or on Twitter, like it's protecting you from hate speech and things like that. Online shopping products being recommended to you, your self-driving car, your spell check, your automatic email categorization. Maybe if you unlock your phone with your face.

So yeah, I mean, that's heaps of domains. And then maybe some others that are not every day for you, but it's like defense, medicine, logistics, retail. Yeah, it's all over the place. Netflix, you know when it used to be more like you could comment under things and you had to go on your laptop and then it was like these things and you had to press shows you like. Now it's not so much as it shows you what you had to go.

Thumbs up for that one, thumbs up for that one, thumbs up for that one. When that first came out and they were starting to suggest TV shows to you, I had my first Netflix special and Netflix announced that it wasn't for me. My son was very young at the time so I think I'd been watching a lot of kids' programs but they went, you won't enjoy this. Maybe they were right, though.

You don't want to watch yourself. I didn't watch it, so it was a good point. Is the speak and spell AI? Jim's really been high on that. I've never heard of speak and spell. It sounds like a kid's toy, but... It is. Okay.

Could a kid's toy help E.T. get back to his planet? I don't think so. That's how he rang. He phoned home through a fucking speaking spell. It's a tool. Well, he uses an interface. But spelling correction, definitely. So this is powered by a language model. So, yeah, you model the probability of different words in a language and so it can use that to sort of you know suggest corrections

and stuff i'll tell you how speak and spell right they'd go spell cat and then you'd go k-a-t and they'd go you're wrong it knew shit man all right um yeah uh what is an algorithm? Jim said, you know, when people are coding, it might look like gobbledygook to you for us, but they're making pitfall. Remember the game pitfall? Yeah, I do. It's on an Atari. It'd be a bloke swinging a vine over, over, over a hole or jumping over some alligator heads. That didn't happen by itself. What is an algorithm? Let's just explain that. Cause that can't be.

It's a bit like a recipe. Basically it's a set of instructions that are written in code um for a computer to execute something okay good for you know what i getting back to the last question we asked i do enjoy it when some people get upset when things are suggested for you to buy but i like that like i like it too. They know me. I went through a run of hot sauce. I bought one bottle of hot sauce and then they kept on suggesting different hot sauces. I've got about seven brands.

Sneakers for me. They're always like, you like sneakers? And they show me that they always know what kind I like and I enjoy it. At first I didn't think I was going to. The thing is they often show you the ones you've already bought, and I don't know why that is. I don't know why they're not intelligent enough to figure that out, but I get that all the time.

They haven't brought it into the prostitution world. They've never gone, here's a blonde, you like them? I don't think that's legal. I don't know. That's what I've heard. I mean, on certain websites that I'm sure that you are partial to, they definitely would use recommended systems. Oh, no, no.

The porn sites definitely, they know what's up with me. They know what's up with me. They suggest the right things. I go on the Pornhub and I'm like, oh, it's just nice that you care. How do machines learn? Jim said, by us inputting information into them. Then the computers start teaching each other and then we're fucked. Correct.

Yeah. I mean, the start of the answer, I was so hopeful for you. It sounded so good. And then it went a bit pear-shaped. But like the simplest methods are often based on statistics. So like spam detection and email, you count how many emails have the word Viagra in them and how many of those are spam. And then you do that for every single word in every single email in your training data set. And then when you get a new email that has the word Viagra and some other words as well, you can use the probabilities to say whether it's... We've lost you. We've lost you.

One second. Can you hear us? Just nod your head. We can't hear you right now. We can't hear you. We have a thing, a computer program in our podcast

if Viagra is mentioned four times. Wait, talk. It has to shut down. Is it plugged back in? We can hear you now. Yeah. Yeah.

I'm not... You cut out right after spam and Viagra. Is it possible for you to remember the whole answer to that question again and start from there again? Okay, great. You can just start from answering it. Yeah. So basically, the simplest methods are based on statistics.

So for example, spam detection in emails, what you do is you start by counting how many emails have the word Viagra in them, for example, and how many of those are spam. And then you repeat that for every word in every email in your training data set. And then you have probabilities. And then when you get a new email in that has the word Viagra, for example, you know the probability that this is spam or not. And that's a simple approach. this is spam or not. And that's a simple approach. And then more complex approaches is where I thought you were going in the right direction. It's kind of about learning from seeing the answers. So we give a model a bunch of data points and ask it to produce an output. And then it's wrong, but you show it how wrong it is. So if it's trying to

predict a number, you say this was totally too high or totally too low. And then it tries again and again and again until eventually it has learned to basically model this input data. And then you can give it new examples and it will do a pretty good job if it's been trained enough of predicting the output. we didn't ask you this jim but what is a neural network and when were they invented do you know what that is that's your nerve system man and it's a network yeah yeah yeah it's it's we're talking about ai so what's a neural yeah it senses when you're down or when you're happy i don't know catherine what's a neural network when was was it invented?

Well, the nervous system is kind of on the right track. Like, it's loosely inspired by the biological. Fuck. One second. I blame Louise, man. What's going on? I think you can hear us, right, Catherine? Yeah, it just cut out.

It's something on our end with a wire. They're fixing it. Hold on. Yeah, it just cut out. It's something on our end with a wire. They're fixing it. Just hold on. One second. Sorry about that. It's like we've got two of the three stooges butting around with each other. We need some AI in here.

This is why I'm not close to you, man. We just changed the wire out. Shouldn't happen again. Sorry about that. With all the AI and all that type of stuff, how can't you guys get a wire that works? Like, I'm serious. Like, every iPhone charger works for a bit and then it stops working.

We should have more people working on that. I don't know if that's AI, though. It's all computers. That's a wire. Yeah, that's where it started. Okay. Getting back to neural network, so exactly what is a neural network and when was it invented?

Yeah, right. So, well, firstly, I exactly what is a neural network and when was it invented? Yeah, right. So, well, firstly, I'll give Jim a chance to get a point. Do you think a neural network is a new thing or an old thing? This is a trick question, so I'm going to say it's an old thing. Yes, yes, you're right. Boom shakalaka. Boom shakalaka. It sounds super new and modern, but, I mean,

the idea has been around since, like, the early 40s, and the Perceptron is an early sort of version from 1958. So it's pretty old, but yeah, it's basically inspired by the brain. So, you know, when we get a stimulus, light, sound, food, whatever, the neurons in our brain, like they get excited to a greater or lesser degree and transmit that signal onto the other neurons. And a neural network is trying to replicate this. So if you want to bear with me, I can visualize it.

Yeah. So just imagine for me like three rows of three circles. Okay. Right. The Channel 9 logo. Right. Yes. Okay.

I mean, this is one of the reasons why I love the show, because I don't get to talk to any Australians. I haven't met any for five years. It's been five years since I've been over here, actually, so I never meet them. So I like the references, the joint references. There's no Australians in Austria. You're the only one.

Maybe. No, no, like, you know, I haven't really seen much of the family, didn't get a chance, and I don't meet other Australians. They don't come here. They go to London or Berlin. They're in a lab. We perform there. Me and Forrest went and did shows there. We're in Meerhofen. Meerhofen.

Meerhofen, some small village. Like a small ski village. I think it's also a brand of bread. Not sure. Anyway, back to the Channel 9 logo. You've got these three rows of three circles, right? And these are the neurons. And each one is connected to the next row of circles, right?

So this would be a three-layer neural network. So what you do is you give some input into the first layer. So it's just numbers. It might have originally been a picture or music, but, you know, it's encoded in numbers. You give this to this first layer of neurons and they manipulate that signal a bit and put it onto the next layer. And they manipulate it as well and push it onto the output layer. And then again, it has a guess, has a go. It gets it wrong.

it has a guess, has a go. It gets it wrong. And based on the strength of this error, you just sort of rearrange the strength of all those connections between the neurons. So how much of that information flows through, you just change it. And through trial and error and doing this thousands and thousands of times, you end up with a network that manipulates an input signal in such a way that the output signal is actually what you wanted. Like a Furby? Like a Furby. Because a Furby, you're taking him out to talk, and then by the time the battery's run out, it's really come to life.

I forgot about Furbies. Yeah, Furbies. I wanted a Furby so bad. Furbies is AI all day. I always used to have a tic-tac-toe machine. I didn't always beat it. It figured out my algorithm. I always went for the top corner and then diagonal that way

so I'd get the two-way, but anyway. So what is deep learning and how is that? Jim said it's when you're high and someone says something that spins you out. Yeah. Yeah. It's much simpler than that. So I just explained like a neural network,

deep learning is just stacking a lot of them on top of each other. So it's deep. That's it. Yeah. And it enables the network to learn more complex functions between the input and the output. So something like language modeling, where a model is learning to take our audio input signal

and then write words, like understand words. It's not very simple. So some kind of deep neural network can learn a complex function like that or machine translation, for example. Yeah, you can't get a computer high, Jim. Yeah, yeah. Even if you try to shove an edible in it. What if they're the people and we're the computers?

Whoa. No. Fuck. Okay, here we go. What is the Turing test? Andim just said alan turing oh yes he made the first computer um he was arrested for homosexuality it cracks me up every time you say homosexual because you say homosexual and he's a homosexual i say it like all people wouldn't like anything yeah

and jim's i said do you think you could pass it? He goes, I don't know what it is, but I think I could pass it. I think you could pass the Turing test. So what's going on there? Explain to us, please. Yeah, I think Jim probably could. So basically, yeah, Alan Turing, you know, he's a famous mathematician, codebreaker, computer scientist. And the Turing test was basically a thought experiment.

So to try and devise a way to determine if the machine is intelligent and the idea is like you have a human interrogator who is interrogating another person. Ah, Luis. What are you going to do for your next job? Sorry, it just cut out again. Sorry, Catherine. What's going on?

Why does it have to be so complicated? It's just a recording. I don't understand. I don't understand. Sorry, Catherine, I'll get it fixed. All these cables to make this fucking talk. I'd be happy if we just put an iPhone in the middle and we all yelled at it.

This is meant to make it better. Leave all this in. I want this all left in the podcast. I want you to listen to your failure while you're editing. I want you to have to sit through every minute of this and go, fuck me, it wasn't the wire, it was something else. Do you want this to be like the promo clip too? I like how you two are walking around

like mechanics kicking a couple of tires. It's probably the fucking Johnson rods or whatever they said in Seinfeld. You know what I mean? No, you're not touching anything. Neither of you are touching anything.

Let them work it out, Tim. They're just pointing at each other. Oh, you moved the chair. That'll do it. It's a shorter cable. It's because the chair is too far away. Oh, this will show us the way. Have you tried

turning it on and off? Do control alt delete. This is going to be fun to edit. No, no, no. I don't want you to edit it. Jim's going to check it. I want the comments underneath you to be pure abuse, Luis.

It's so much fun when it's not you being berated, you know. How many times did I berate you? Never. Never. That's why it's always so much fun. Exactly, because I haven't put you in charge of a simple task. Well, it seems complicated, to be fair.

Are we ready? We're plugged in. We hope this doesn't happen again. Can you talk real quick? Yes. Yay! Okay, we were talking about Alan. You're doing a good job. I was just teasing.

We were talking about Alan Turing. I'm glad I get no blame. Do you want to leave that in there, Jim? Yeah. Sure. That was fun. We were talking about the Turing test and Alan Turing and then you got cut off because the wire went bad. So if you could just tell us a little bit about

Alan Turing and the Turing test again, please. Yeah. All right. I'll start again. So Alan Turing, as you know, he's a famous computer scientist, code breaker, mathematician. You should have got him to do the fucking same. Yeah, maybe. He's dead. And the Turing test is basically a thought experiment as a way to try to figure out if a machine is intelligent. So the way that it works is you have a human interrogator and they're interrogating another human and a machine, but can't see those. And it just basically asks questions. And it's the machine's job to try to convince the interrogator that the other one, that the human is the machine. And basically, if it succeeds in doing that, then we might say that that machine can think.

So that was the idea of the test. And Alan Turing thought that by 2000, we would have machines who can pass it. We're not really getting close to that. So yeah, Jim, I certainly hope you would be able to beat the Turing test. How far are we from actual hoverboards? Like I feel like 2015, back to the future, they said that was all going to be happening. They got a few things right.

They got a few things missing. How close are we to floating cars or is this not in your wheelhouse? It's not my domain at all, but I mean. Just have a guess though. Maybe 20 years or something. Yeah, all right. But I mean, I've seen a lot of prototypes and stuff at the moment.

Could end up being closer. The last few years having like, you know, the sort of progress has increased pretty rapidly. And everyone's being shocked by it, like especially with the sort of increased use of deep neural nets and stuff. So I could be wrong, but, yeah, I think it's still a copy. Serious question. How far are we from robots teaching other robots without us having it?

How far are we from a robot inventing another robot, building another robot, and then so forth and so on? The domino effect. How far are we from Terminator? Two, not one. That's actually the singularity that we were talking about before. Yeah, we can jump ahead to that, yeah. Yeah, yeah.

So basically the idea of the singularity is this sort of intelligence explosion where the AI starts to be smart enough to create other AIs that are even smarter until the point that we can't keep up anymore. No one is really sure how long this will take or if it is even possible, but some people are scared of it, especially as we try to search for this thing called artificial general intelligence. This is like people are trying to make an AI that's capable of many different tasks and some people are scared that as we try to do that, we might hit this point, but nobody really knows.

Okay, so how smart are the computers now? Are they as smart as me, smarter than me, or are they just still pretty dumb? Depends on the task. And again, we kind of get to one of the other questions about the weak versus strong AI. A weak AI is really narrow, so exceptionally good at one thing. There are plenty of tasks where there is no human alive who could be as good as that, like machine learning or calculations or

different things like this. But it's super, super narrow and they can't generalize at all. or different things like this, but it's super, super narrow and they can't generalize at all. And then on the other hand, the strong AI is more broad and more general and can do lots of different things. And we're not very close to that. We're getting better with things like few shot learning and so on and multitask learning, but where it's still a huge, huge problem. but where it's still a huge, huge problem. I'm not going to be worried until the automated phone thing gets your right. It never gets it right. You're always like, yes. Did you mean no? No, I said yes. And when that starts working correctly, a hundred percent of the time, that's when I'll get afraid. Yeah, no, I can't use those ones.

I just start yelling at them. They edit in typing now when you say an answer, like, let me look that up real quick. Yeah. What is that? Yeah. That's annoying. Um, Tamagotchis.

Are they going to progress or are they, have they reached? Maybe. Yeah. If they reach. Um, I had Tamagotchis. What type of pet did you have? Uh,

I don't, weren't they all the same thing? You have a little cat, you got a little dog, you have a dinosaur? I don't, weren't they all the same thing? You could have a little cat, you could have a little dog, you could have a dinosaur. I don't remember, but I do remember that I would like leave them in my mom's purse and they would wake her up in the middle of the night. She'd say, feed your fucking damn cat.

They'd starve to death. Here's a question that we sort of touched on, Jim, but we didn't ask. A famous moment where an AI beat a human at a cognitive challenge. Okay, so there's been some tic-tac-toe machines that beat people and there's a chicken that can beat you at the fair. Not some AI computers. But then there's a machine on Jeopardy.

There's a Jeopardy machine that, and I can't remember the name of the thing, but the guy that now hosts Jeopardy and he's one of the chasers, he beat the machine um okay but what about ai beating a human but this guy beat everybody oh the the chess people they always win but the humans now ai beats the people at chess for the most part not always but for the most is that right i thought it took a while for ai to beat humans at chess or is it now it does that for the most part that the computers will beat you at chess if it's trying. I think the thing is it's algorithms that can play a

regular player at chess and beat them. I think that's been around for a long time. But the first time that an AI beat somebody at chess like a world champion, Garry Kasparov, was in 1997. It was Dee Blue from IBM and then yeah IBM came back again with Watson in 2011 Watson yeah and so this one

and you know if you want to spend a nerdy night in you can see that on YouTube and it's pretty cool to see the little processing Why call it Watson and not Sherlock Holmes because Sherlock Holmes was all the fucking, he was always going ah, Watson. Like he was always telling Watson, I've got the elementary Watson. I've got an idea. Why name it after the second one?

You never go, we're building a superhero computer, Robin. You never do that. You call the fucking computer Batman. Why call it Watson and not Sherlock Holmes? Because it's your sidekick. I don't even know if it was named after. This could have been something else. No idea. I mean, you know, Google, the algorithm that ranks the search results on Google, it's called Google PageRank,

but it's not because it's ranking the pages. It's because the creator is called Larry Page. I just looked it up. Watson was named after IBM's founder and first CEO, Thomas J. Watson. Yeah, he was the assistant of Sherlock Holmes. Oh, sorry. Not Thomas J. Sherlock.

Before that, he did that for a job. My bad, I'm an idiot. So Watson beat the people. He didn't beat everybody. Somebody beat him. Because isn't Watson basically just like Google? They ask a question, it worked really fast, it went through all the things on the internet until it found the answer.

Is that what happened with Watson? I think there was probably lots of different components working together. So some natural language understanding having to take place, then some information retrieval like this. Yeah, a lot of different stuff. But yeah, Google is definitely based on similar kind of things. Like they built these knowledge graphs, for example, where like connecting nodes, you know,

trying to connect all the information in the world in different nodes and understand the relations. Yeah. It's a super interesting and like ongoing problem that they would have definitely been working on back then as well. What is the Chinese room experiment? Jim said the room is made up of folded paper. You get out with two sticks and it's the same as the cat in the box test. Just a bunch of question marks.

I was like, what's the cat in the box test? You know the cat in the box test. Schrodinger's cat. Schrodinger's cat. I might get this wrong. Showbiz cat. What happened to showbiz cat was

you walk in, you go, there's a cat in the box and no one's quite sure. If you open the box, does the thing die? Or if you feed it? You tell us, Jack. What is the Chinese room experiment, Catherine? Tell us the cat thing quickly. I've got to know.

When the box is closed, the cat is either alive or dead and we don't know. So there's two timelines that could exist and we only know what it is once you open the box. That's anything isn't it it is yeah do you know cats only meow when when people are in the room when they're just by themselves they never meow to each other well that's like when a tree falls in the forest you can videotape me you can leave like a camera in there and they just walk around like that then we walk in there we all talk no get the fuck out of here you needy cats what what is the chinese room

experiment catherine right so this is the idea that um like you have a person in a room and you're um feeding them um like chinese characters written on paper just you know slipping them under the door and the person in the room doesn't understand chinese but they have instructions of how to like to generate an output. If you get this, then you should write this. So, rule-based system. So, the person writes the output and passes it back under the door. And so, for the person outside, they think they're having a conversation with someone who can speak Chinese, but that's not true at all. And basically, this is pointing out the flaw of the Turing test. So, we can make

things that seem to be intelligent, but they actually aren't. So, that's one conclusion you can make from this thought experiment. Wait. So, neither of them know Chinese? Neither of them know Chinese? Maybe the person outside knows Chinese. Yeah, they would know it. And they get the impression they're speaking or writing with a Chinese speaker, but the person inside has no idea. It's just following instructions. Did I mention the other day I was thinking about John McCain in prison?

Did I talk about this the other day? So John McCain, he gets caught in Vietnam and he's got another bloke on the other side of the wall, right? I saw a documentary. I locked up abroad on a documentary. And they tap on the wall, tap means A, tap, tap means B, tap, tap, tap means C, so on and so forth, right? And they kept on doing that to each other. And they were locked up next to each other for like a couple of years,

right? Until they figured it out, right? But I always thought that that must've been like a kind of a thing. They would be sitting next door to other. Oh, God, I have to spell the word zoo. They never brought up pizza. Didn't even bother bringing up pizza. But wouldn't they like have really liked how we use like just those abbreviations now? Like, you know, like you just got beaten up. Oh, LOL.

Yeah, you just got beaten up. You got tortured that day and then you go, FML. Is that it? Yeah, I just think that's artificial intelligence in a nutshell, man. Okay. No, but the thing is it's not like artificial intelligence, but it does show some kind of evolution of learning and that is, you know, one thing that AI might try and do.

So you might try to put an AI in a situation where it has to try and figure it out for itself, and that's exactly what machine learning is about. It's learning for itself as opposed to being programmed how to learn. Because they would have just kept on bashing in the wall, do you understand, do you understand, do you understand, until eventually YS came back, and then they're off to the races. So what's your name?

They were next door to each other for fucking two years and all they got out is, I'm from Cincinnati. It took a long time. But I will say one thing just to allay some of your fears about like AI taking over and stuff. Another conclusion you can take from the Chinese room experiment, so some people say it shows that thinking and intelligence is really biological.

And, you know, this example in the Chinese room is not a replica of that. And because of that, we can never build intelligence. We can only try to copy it. And it's really, really hard to know whether that's true or not. But if that's true, then, you know, we might be okay. I want to kill us. Do you, because you work in this field, and do you ever get pushback from people in society who say what you're doing is wrong or is everyone like, this is awesome, well done? I don't get pushed back because the domain that I work in, it's not having any potential negative consequences on people at all. It's basically we are in e-commerce.

We're predicting how well products will sell. That's the bulk of my work is trying to understand that. And we have various uses for those predictions. But I am myself interested in AI ethics, and I want to make sure that if I was ever in a different job that I would be very, very conscious of these ethics, you know, boundaries and so on, so I would push myself back. But luckily I don't have that situation yet. And when you put the AI into Mercedes, right,

so is that just the navigation system or is that like? Well, I mean, I was a work student in doing some very early research. I know you didn't do it by yourself. You don't have to go ahead. No, but in the voice interface system. So I was basically doing more preliminary research, trying to understand whether you can identify character types or like

characteristics, like maybe gender or age or something like that from the answers people gave to certain questions. So it's nowhere near getting into the cars yet. It's really groundwork. What is an AI winner? And are we in danger of having another one? Jim said,

it's when you go, Oh, shouldn't be cold this time of year, but it is. I don't even know how you got that. So what's an ai winner and i'm in danger of having another one is it dangerous yeah ai winter is coming um basically these were periods in the past where there was just too much hype too much investment um inflated expectations and research was simply not able to deliver because AI is really hard. It's ludicrous now, but a fun fact, there's this really famous quote from

these bunch of scientists in 1955. They wanted to make a summer school on AI or something like this. Basically, they said, we think it's a problem that can be solved in one summer. This is ridiculous. It's so hard. What happened in the past was people's inflated expectations were not met. Funding started to dry up. Research started to slow down. The whole thing went underground for years. A lot of people see similarities to the situation at the moment. It's definitely a hyped topic and there's insane amounts of money being invested by venture capitalists and so on. So some people are worried we will get an AI winter again because it's still hard.

And others think, no, this won't happen. The ecosystem is a bit more robust now. But again, I've heard very convincing arguments both ways. So what I'm hearing is it's very hard just to throw money at the problem because like, do you ever go back to the investors and they've asked you to do a certain task and then you go back and you go, this is really hard. Like, is that basically like we had a go, but it's really hard. I mean, my specialty is natural language processing. It's kind of one of the things that is a love-hate thing about it.

Language processing is really, really difficult, and you end up getting stuck in trying to, you know, increase your precision, but then you wreck your so-called recalls. Okay, I'm going into too much detail, but it is super hard. This does happen a lot, and I'm sure that there are awkward conversations like this about where are the results. This might sound a little bit bigoted, but always a good way to start. You've already done the Chinese real quick.

Is there a language that lends itself to teaching a computer AI quicker? Like is English better to do it? Is German better to do it? Is there a thing or is there some languages where you go, oh, don't do it in Thai. It's a fucking waste of time. It's too difficult. Or is, you know, like do you understand what I'm saying?

Like, yeah, like does it lend itself to a language? It's a really great question. The only thing that springs to my mind as maybe easier is Esperanto because this was a human-designed language. So, I presume that it's a bit simpler. But otherwise, all languages have their own little irregularities and so on. So, they're all difficult. Certainly, some languages pose more challenges than others. So, English is a pretty simple language compared to, say, German or Finnish or things like that. But at least all those languages have the benefit that, for example,

every word is separated by a space. And that means the very first step, which is just literally getting all the text data and cutting it up into words to analyze, at least that's kind of easy for us in something like Mandarin, the characters are put together and there is not necessarily one way to separate those characters. So even just getting the individual words out of a sentence can be really tricky. So, yeah, there are definitely languages that are more difficult. I reckon the hardest languages for AI is anything that involves,

that must be hard to get across. Siri, what way to my house? You don't know what you're talking about, Siri. I also like that you thought we knew what Esperanto was, by the way. I knew. Okay. Okay, I'm sorry. It's Spanish, man.

Esperanto. okay i'm sorry it's spanish man esperado it's it's it's a spanish yeah esperanto yeah what is esperanto oh so yeah i mean it's uh she just called you an asshole i'm sorry no no no it's no. I said, it's German. I've been here too long. So when I realized, oh, I was like, ah, that's what you mean. Ach so. No, it's, yeah, it's a design language. So it was designed to be easy to learn.

And designed by who? I don't know. A person with too much time on their hands. I reckon. How much do you reckon if we start up a new language if it would catch on how long it would take a long time hundreds of years right thousands maybe never yeah i'm good um we talked about we can barely speak english i don't think we're gonna come i believe the whole world should have the same language but i think we can't do that until the

whole world figures out powerpoints until we can figure out plugs we all need to have the same plugs the presentation software i'm traveling i'm traveling around trying to charge my phone up oh it's a different plug here why stop doing that wait you don't think we should all have the same language no i think we should know catherine catherine is saying definitely not because languages are really valuable and they keep delivering valuable in ways that we don't know yet like um an obvious example is medicine so some of these um small languages like in the amazon they know about plants that we don't know that could you know contain the cure for cancer or something and sadly a language dies about every two weeks and every time a language dies about every two weeks. And every time a language dies. Not fast enough.

Come on. Fucking Welsh has been a life support ever since I've been born. They're still harping on with that crap. They've got four L's in a row. Yeah, well, they're not going to discover any medicine, Welsh. Yeah, you're right. Ooh, if you put the leek in the soup. Yeah, I never thought about that.

The medicine. Yeah, okay. All right. Because I'm almost on that camp. All right. So what is the difference between weak and strong AI? We talked about that.

What is singularity? We talked about that. It's not the dating page that Jack goes to. That's funny. That's... Yeah. How does playing floor is lava

help advance machine learning research? Jim said easy. Gives the machine some parameters. Picky uppy machines. Weak AI. No, no, no, no, no.? Jim said easy. Gives the machine some parameters, picky uppy machines, weak AI. No, no, no, no, no. I've got kids. I know the answer. You make the kids play Flora's Lava,

that kills a bit of time, and then you can work on the AI. They're out of your hair. Yeah. Yeah, no, it's a good thought. But let's take another example with the kids. You know, if the kid hurts himself, then they'll probably learn not to do it again, right? Hopefully. So, you know, they trip over. I don't know, me kid keeps answering back and I keep hurting him.

I don't know. It's his birthday. I don't, I don't hurt me. Yes. But imagining that you had a healthy relationship with your child. Burns, dang it. Yeah, sorry. And, you know, your kid, he did something, he got hurt, and then he would learn maybe next time not to do it. And this is basically reinforcement learning.

So it's another kind of machine learning. So what you do is you give the AI some kind of reward function that it has to maximize. And it might be a function like score a lot of points. And you don't tell it how to score points. You just put it in an environment and it will move. It'll do something and it will get the points. Or it'll get the points taken away, right? There's a cat in the box out the door. Arnie, come here.

That's how you know if the cat's in the box. Come here. Get over here. Sorry. Lots of fun stuff today. Come here, Arnie. Hey, Arnie. Arnie. Arnie's a robot

dog. Those robot dogs, they're shit, aren't they? Those ones they tried to sell us for years, the ones that all walked up. Busting dynamics, yeah. Fucking useless things. Wow. Jesus. I haven't fed mine in years.

All right, Catherine, I'm sorry. Sorry, Catherine. No worries. All right. So tell me, what is reinforcement learning? Did you get it? Yeah. You keep giving kids rewards when they fall in lava.

You give them rewards when they do something good and you maybe give them a punishment when they do something bad. And it's the same with the AI. And so it learns to, yeah, to achieve this goal. And it's super useful in things like, yeah, recommender systems, like recommending content that you might like. It could be used there or in share trading.

Some people will try to teach like a, you know, like recommending content that you might like. It could be used there or in share trading. Some people will try to teach like a stock trading bot to reinforce the money. That would be a good one. It's super hard because, I mean, if it was that easy, people would just do it and make a ton of money. So it's hard because there are so many factors that influence the stock market. You know, the whole world is super complex.

But, yeah, this would be an idea. So the reward function would be, hey, make me a ton of money. And so it would learn if I buy this stock or sell the stock and the money goes up or down, it learns to maximize this reward. And Flora's Lava is one kind of simulation that I know Google have used. They also do stuff like hide and seek. That's really cool. You can look that up on YouTube because these little simulated creatures, they really learn to work together. And then they even end up learning to lock the seekers in, like to build a structure around the seekers so so the seekers can't move like after thousands and thousands of iterations so um yeah there's lots of uh interesting uses for for reinforcement learning how far are we from like the black mirror where like if i die before i die i can download

all your questions are how far away are we from yeah that's what i want to know um so so like from if i'm about to die and i can download myself onto a computer and then in a way i can live forever even though i won't be alive but people who like me can come and visit that computer that might be a very lonely ass fucking computer i'll tell you that for nothing well we could make a language model based on all the languages that you normally use um and uh and you know we would be able to make a deep fake based on all the languages that you normally use. And, you know, we would be able to make a deep fake based on your appearance and stuff like that. So we're telling you how to simulate it for a while,

but without any new data, you're never going to, you know, change and develop as a, as a being, you're always just going to end up saying the same old thing and the same old thing over and over again. Right. Well, that's pretty much what I do now. It's about 25 words. It'd be called computer under natural torrents. Cunt the machine. again. Right. Well, that's pretty much what I do now. It's about 25 words. It'd be called computer under natural torrents. Cunt, the machine.

Yeah. That's as good as I could have done. I was being thinking about that for 30 seconds. How does Amazon's Alexa and other smart devices understand what you want? Yeah. Well, the first thing I have to say, Jim, I have totally the same problem with you with the language recognition. They don't get my accent. So that probably indicates that they haven't been trained on enough Australian data, so spoken language data. But basically, first, you have automatic speech recognition that'll turn the input signal into text because that's what the computer can work with. And then what it tries to do is do natural language understandings. It's my domain. So simultaneously, it's trying to figure out what domain are you in? Like maybe this is a travel-based request or an entertainment-based request. And then what is your

intent? So if it's travel, it might be book a flight. And if it's intent, it might be play a song. And when it's figured those things out then it knows what it needs to know to be able to finish your task so like if it's a travel domain and the intent is book a flight then obviously it's going to need to know what's the airport you're leaving from and where are you going and so then another neural network will try to label every word in your your utterance with with as many slots as it can that it needs. And it might ask you for more information, but until it's basically got all this information

from your original unstructured utterance into a structured thing called a semantic frame. And then it's pretty simple. This information is used to query an API. So like basically access a publicly available sort of interface to say a travel website and it will query the website for the flight that you want and the API will return a result and then probably natural language generation will be used to turn it back into a text and maybe even it'll be spoken aloud with speech generation,

which would again be taken care of by probably nothing. I want you to repeat this answer now, Jim. What you do is you talk to Alexa and it finds your flights, man. Yeah. Trolley problem. What is the trolley problem? Jim says Tesla can't recognize trolleys. Bang right in on them.

That's the shopping carts. Oh. We call them trolleys. We call them trolleys. I ding ding like a tram exactly yeah so think about it as a tram and now think about it that we said it's a it's a it's an autonomous vehicle problem yeah so so the vehicles, they don't like trams, right? Because they're not cars. Same answer. Yeah, autonomous. Okay, everyone knows that autonomous cars are racist as fuck when it comes to other vehicles.

They're just like, I don't even want to know you, man. Wooden cars. Yeah, wooden car. They wouldn't recognize Luis's car. I think it's a tree. Yeah, Luis's car runs like a clock. So what's the trolley problem, Catherine? Yeah, so basically, again, it's a thought experiment.

So notice there's a lot of thought experiments here today. So the idea is you've got, yeah, this like old train carriage and it's out of control. It's going down the tracks towards, say, five workers on the track. Oh, I know this. It was in the TV show. of control it's going down the tracks towards say five workers on the track so they would all know this it was it was it was in the tv show the one with the christian bell in it at the end they had the trolley experiment and so do you just think in a head now i feel like do you plow through a whole lot of people or do you turn and die and you say well those people they don't know yeah

like so so you know um yeah, the question before is like if you flick the switch and direct the tram to just one person, yeah, you save a lot of lives, but then you actively kill a person. What would you do? And with autonomous vehicles, it might be something like, okay, if two children run in front of the car and the car could save them by swerving and killing the driver, what should it do? And a lot of people will take the utilitarian perspective,

which is costs. So, so they would say, okay, the fewest lives, that's the driver. You have to kill the driver. But the question becomes different when you say, okay, what if you're the driver? And they say, well, I want my car to look after me above anything. Those kids are dead as far as I'm concerned. It's going to be super relevant. I think there will be so many big court cases in the future and stuff about this. I'd have to know the kids if they're shits or not. You know how much a bumper on a Tesla costs? I do. Here's the big question. Will AI steal your job? and it kind of goes back to what jim was saying too

like how are we in danger of not only that are we in danger of ai taking over and terminator you kind of addressed that but uh jim says it will sell many people's jobs it already has we're all going to become wall right it's going to get worse and worse and worse soon we won't even be interviewing you we'll be interviewing a robot who tells us what to do. I reckon I could make a robot right now who could be a late night TV show host.

Netflix has put out some short videos of AI written stand-up. And how did they go? Not good. But they still did better than Jack. What was it I don't know they just released

a bunch of videos like in the Simpsons where they went what are those fools on Capitol Hill doing like that that's all you gotta do you'd be an AI guy

you just go you go our guest is next this thing and Trump whoa he said something stupid there you go

that's all you gotta do that's good so is it gonna steal our jobs is it gonna steal our jobs not Is it going to steal our jobs? Not our jobs as such. Yeah. I mean, I don't want to flatter your ego, Jim, but I don't think an AI will be able to replace you at the moment.

I think I'm irreplaceable. I don't think a computer could do what I am. I want to see a computer that can lay on a couch for 12 hours without urinating because it looks like it's too much effort. Yeah, I'm not sure about that, or I don't know about that. But, yeah, the creativity, also like empathy, so maybe nurses or things like that, any jobs requiring a lot of human thought, they will be, yeah, very hard to replace

at the moment. The jobs that are maybe very repetitive, very manual, not very nice, these ones might be automated, but it's still going to probably take quite some time because, say, the Boston Dynamics robots, they're amazing, but again, they're limited in what they can do. The cool thing is that probably a lot of the jobs or the tasks that will be automated will simply leave room for the workers to do the more interesting and challenging problems. So if you work in a help center for a bank and a chatbot takes over the really mundane, repetitive questions, you might get passed on the more interesting and tricky questions.

So I'm not scared that it's going to take a lot of jobs very soon, but definitely some will be replaced. You made a great point with the truck drivers. And, yeah, that's something that will happen. Yeah, because when they... It will create some more jobs as well. It's going to definitely create new jobs too. Jay Leno, there's a bit of a name drop.

Jay Leno took me to a place that the robots flip your burgers for you, right? It's got a spatula on the end of its arm and it waited and it got on and he flipped the burger like that. I'll tell you what, though, with all that technology, couldn't put fucking cheese on it. They haven't taught a machine to lightly place the cheese on. It can do flipping. It could probably bring you a bun over to the station.

The assembly of a burger. A human being has to do that. It's called Cali Burger. But, you know, I'll tell you. Yeah, it's a real place. So I'll tell you where automation is good, where it's coming in. For war. For war.

Before that, we're not really even training fighter pilots anymore because we've got drones and all that type of stuff. Like whenever I play Call of Duty, I think to myself, I can't believe this is real. No, but I can't. Like the people were running towards each other with guns. Now we just have robots fighting our wars for it. I think that's super.

That's not how it works though. It's only good if you're the country that has the drones, but there's a lot of countries. And where do I live? Drone America Yeah, I'm not paying all these taxes for fucking nothing folks. I'm a drone man That makes a good point, yeah

This is a bit morally dubious where you're going there, but um That's because you're near Germany, you're like, oh we shouldn't do this anymore Alright Look, this is the part of the show called Dinner Party Facts, where we ask our guests. It took me that long to bring up the Nazis. I'm surprised. I thought I would have done it right straight away. It took me right to the last minute. We ask our guests to give us some sort of fact or

something interesting that our listeners can use at home to impress people on the subject. What do you got for us, Catherine? Yeah, I just wanted to say like AI is brilliant, of course, in certain domains, but it's also really dumb and it can fail in really funny ways because it learns to like over rely on So, for example, there's a famous example where they were making a dog versus husky detector, and it was really, really good. It was really excellent. Sorry, I missed that. A dog versus husky? Husky, yeah. So in images, so it's an image classification, and it's trying to,

I don't know the practical uses for this classifier. Yeah, I don't know either. Yeah. Isn't a husky a dog? Is a husky something? Well, that's the thing. That's why they would be doing it because it's tricky, right? Computer dog versus real dog.

Well, no, no. A husky, some people would say a husky is its own kind of breed, like the wolf, close to a wolf, I think is what they're, right? So you're saying wolf versus dog. Yeah. Where is the robots involved? I'm missing this. It's very useful technology.

So we're at a dog fight. No, no, but this is image recognition. And yeah, this use case might be very narrow. I was trying to tell the difference between a wolf and a dog. Okay, I'm with you. I'm with you. Okay, sorry. Okay, cool.

Sorry, probably I said it a bit wrong. This is an awesome dinner party fact. You're going to have a great time repeating this. So, yeah, it's trying to tell the difference, and it was really good. But then the cool thing is we can use, like, model insights to try to understand what a model is paying attention to when it's making a prediction. And when they did that, they realized it wasn't looking

at the animal's face at all. It was looking at the background to see if there was snow or not. So the dog versus husky detector was actually literally just a snow detector. And this happens all the time. There's a super nice example of they had an AI-powered camera for a football match. It was supposed to follow the ball.

The referee was bald, and it just kept falling. So you can really entertain yourself by looking up AI fails on the internet. But basically, AI is still, in general, really narrow. It's not great at generalizing. Whereas humans, we are a few short learners, which means we can generalize really good, learn new tasks really well. And so this is why I think there's still hope for us. We're not dying yet. I reckon a difficult one would be old person's finger or carrot. What about hot dog or legs?

Hot dog or legs? Hot dog or legs, no, but that one tricks me. Or the beach. If it's at a beach, it'd be legs. You ever seen the webpages, hot dogs versus legs? No. You're not sure. It's always like people's Instagram photos of the beach. I've seen the hot dogs

or legs thing, like people post it. Yeah, but people do it on the beach like that. You go, that's a nice looking woman. That's a nice looking woman. Hot dogs. A lot of people bring hot dogs to the beach. All right, Catherine. Well, thank you for being here. Like I said before, you can follow her on Twitter at Catherine A. women hot dogs yeah a lot of people bring hot dogs to the beach all right Catherine well thank you for being here like uh I said before you can follow her on Twitter at Catherine A Monroe and on Instagram Catherine Amable did I get that wrong again Amabelle I'm an idiot

so it's A-M-A-B-E-L and then A-M-U-N-R-O for Twitter. You can teach your computer to do that in seconds. But thank you for being on the show. Thank you. Thank you for being on the show. Ladies and gentlemen, if you're ever at a party and someone says, I got a computer that can recognize a husky, you throw them a bit of snow and go, I don't know about that, and walk away. Throw them snow.

Yeah. Can I say the last bit? Sure. Good night, Australia. Good night. Good night.