Season 2 Episode 13: Dogs vs Cats and AI with Dr. Janelle Shane

We hope everyone is doing ok and handling the physical distancing!  Our program this week has science news about the periodic table and a fun discussion about why dogs breeds look so different when compared with cat breeds.  In the Ask an Expert section we are so fortunate that Dr. Janelle Shane chatted with us about AI and her amazing book "You Look Like A Thing And I Love You!"  It's such a fun and educational talk!
You can find the links below to her book!

Dr. Janelle Shane on Twitter:
https://twitter.com/JanelleCShane

Dr. Janelle Shane's Website:
https://www.janelleshane.com/

You Look Like A Thing And I Love You!
https://www.janelleshane.com/book-you-look-like-a-thing

Dr. Shane's TED Talk
https://www.ted.com/talks/janelle_shane_the_danger_of_ai_is_weirder_than_you_think

The ASAP Science Periodic Table Song:
https://www.youtube.com/watch?v=rz4Dd1I_fX0

Bunsen on Twitter:
https://twitter.com/bunsenbernerbmd
Bunsen on Facebook
https://www.facebook.com/bunsenberner.bmd/
InstaBunsen
https://www.instagram.com/bunsenberner.bmd/?hl=en
Bunsen Merch!
https://teespring.com/en-GB/stores/bunsen-berner

Genius Lab Gear for 10% link!-
10% off science dog bandanas, science stickers and science Pocket tools
https://t.co/UIxKJ1uX8J?amp=1

Support the show (https://www.patreon.com/bunsenberner)

Transcript

spk_0:   0:08
Hello, science enthusiasts. My name is Jason Ziolkowski and your host. I'm a high school chemistry teacher, but you probably know our dog Bunsen burner. He's the Twitter science dog. This show takes what's best about Munson's account, the science of empathy found there and spends it into podcast. For every week you'll learn some new science in her science news section. We'll also talk about some really interesting dog or pet science every week. There's an amazing expert that has interviewed, and we get to learn so much from them, and we end the podcast with stories and trivia. This is the science podcast. Hey, everybody, thanks for tuning into another week of the science podcast. We've got a great show lined up for you. We're all still happy and healthy. This is like 38 days of quarantine. I'm back teaching, so I do go into my school. Only about 40% of the staff are reporting into the school for teaching from their most most people are teaching from home. I'm definitely physical distancing at my school. If you want to know how our province Alberta is doing um and where are area in the world is doing? I want to say our curve has been flattened, but that just because the curve has flattened doesn't mean that everything goes back to normal every day. We're still getting new cases in Alberta, but the growth has slowed and the deaths have slowed as well. So that's really good news. Um, I'm pretty sure the schools in Alberta are done for this year. We won't be going back until September. Very sadly, I've mentioned this before. Our family loves going to these comic con things. The Calgary comic con was postponed and that it was just canceled. And I heard rumblings that the enormous fair that happens in July and Calgary, the Calgary Stampede it was cancelled as well this week. So, cove, it is affecting everybody's life in Alberta. And the other thing that remains to be seen to is that Alberta is a huge oil and gas producer in Canada, and, uh, oil dropped below $0 a barrel this week. So for people who aren't involved in oil economy, that definitely gives us cause for concern. So who knows what the future will hold? On a very positive note. There's some exciting news that will tell you guys about in a couple weeks. I want to keep it kind of a secret, and probably in the next week or two will be launching the Bunsen website. Uh, and that's just where I've been moving all of the Bunsen merch. We have some adorable merchandise. I don't normally plug it on the podcast, and it will be in the show notes. That's probably good, but you can head there now with the websites. Probably not gonna be open when you hear this, but it's Bunsen burner bmd dot com onto the show for today in science news. We're gonna be looking at the periodic table, Yes, chemistry. I get to talk about chemistry and in dog science, we're gonna take a look at why dogs look so different. Yet cat breeds really don't you know there are some definitely some cats that look different, but the vast majority of cats don't. And our guest is Dr Jenelle Shane, who studies and writes about artificial intelligence, and she's gonna talk to us about that and her hilarious an educational book. You look like a thing and I love you. He Bunsen. Did you hear that? Self driving cars were adding Microsoft to its AI program. Yeah, they were wanting to write their own autobiographies OK on with the show, because there is no time like Science Time this week in science news. It's about the elements on the periodic table. One of the things that I missed about a week ago was I do these science shows basically like there's a guy in the States named Steve Spangler and I steal all his stuff, basically so as a as a science teacher and chemistry teacher. It's a long story, but a short story. Somehow I got into the gig of putting on these really fun explosive signs shows for kids, and it's I get hired. Then my school sends me out all of these different, like middle schools and elementary schools all over central Alberta, and I have a blast doing the shows like the end of the show. I do this massive liquid nitrogen explosion. I have a big budget to do really cool experiments, and all the kids go wild like it's crazy. It's it is so much fun. It's a lot of prep work, but it's absolutely so much fun. And the school I hired to go to because of covert 19. Everything got cancelled, which was really unfortunate, but this would have been like the third time I've been to this school. So the kids, I think it's for their Grade five program. The teacher has me come out and do this chauffeur on Lee the great Five kids. So the great four kids here have heard about the show since they've been in great three. So the excitement has been building. Anyways. Last year I went to this elementary school and the entire class had memorized the periodic table song by ES AP Science. And before I did my show, they saying it to me the I don't know if you've heard of the I'll put a link in the show notes, but it's this adorable song where these guys ASAP science sing all of the elements to the periodic table like there's hydrogen and helium and lithium beryllium like they go through everything on the periodic born carbon everywhere on nitrogen all through the air. I love the song, and I've got most of it memorized cause I'm a giant nerd. But all of these kids memorized the song, and it's funny because they got to a point of the land tonight's and activities where there's just too many and and I think they ran a time to memorize it. So what the teacher had the kids do was go, uh, and they look to each other as the song list in the background. Music was point Okay, so it's a long story to get to the periodic table. But one of the things is really interesting is how new elements discovered. So for the longest time on the periodic table, there were like blank spots missing on the bottom row. Or they had crazy names like a nun IAM. And when I had kids in Grade nine, they'd be some of the kids who were really interested in chemistry. But, like, Hey there spots missing and I have this huge periodic table in my room and, uh, kids who were really artistic once every two years, they asked me if they could fill in a a little square with an element because they're like, Well, what should go here? And I'm like, I don't know. It hasn't been discovered yet. People are working on it. So one girl did this little sketch of, uh, h It's the element of surprise. Ah, and she drew a little ninja is super cute. And then another girl, I think two or three years later, filled in the next block with it was, um um and it was SpongeBob holding the orb of confusion. I don't know if you've seen that episode. You know, I'm talking about in the mix em and Patrick go juror like there in the state of confusion. So it's the element of confusion. You can see how great this joke is. The people who love chemistry in 2016 that whole rows filled in, which was really, really cool. What were some of the elements? Well, one of the elements, I think it was 1 14 N. H is Nyoni, um, and nine Hone is the short form like that part of the word and it comes from the translation for Japan. It was found in Japan, and the element means the land of the rising sun, which is really, really cool. Element 1 15 was also released in 2016. And that's got the element symbol M c must cov in discovered in Moscow or the region honors the Russian land that's home to the nuclear research center that discovered it. Then you have Tennessean on. And that's the recognition of the Tennessee region where the element was found. And last one was a Gunnison and a goddess in his Ogi. And that honors the science Professor Yuri again. Ihsan. So why didn't they name it are gonna sand instead of began Asan. Well, all of the so O g. This element that was discovered is the last one in the noble gas row, and they all end in O n uh, with the exception of helium. So look, you got neon raid on are going or gone. It's on, right? So that's why they just I get it. Maybe they should have just named it after the guy because helium kind of breaks the mold anyways. Now this story that went into discovering these elements is really interesting, and I have heard tell that there are some scientists that are working on the next row down element 1 19 an element 1 20 But to get to that spot, it takes an enormous amount of work. How do they make these elements? They're not found in nature you can't just go dig a hole and find yourself some Tennessean. Oh, you wouldn't want to either, cause it's probably super radioactive, But these elements are made by protons. The more protons you have with the nucleus of an atom that that gives the that gives it its atomic number. And that's what gives it the element, its characteristics and protons. Air held together cause protons air positive. They're held together by neutrons and neutrons, kind of the glue that holds him together. It's way more complicated than that, but that's the best and fastest way to explain it. So Tennessean was made by colliding to other elements. Another one called Berkeley Um, and calcium. And this is not an exact science. It's not like you just put some calcium like Not like your bones. It's not like they're putting a bone in a beaker. Um, the Bunsen would like that anyways, but they're shooting these particles at each other like smack smash, and it took them so many trials to get enough of it to collide. To show that it's stuck together and Tennessean decays super rapidly. It just decays into like the loses protons. It's not super stable on, and there was a lot of controversy about some of these elements that even back in 2012 they were saying, We can do this but they couldn't prove that the elements stuck together for a while. The element that was made in discovered in Japan. Nyoni um, that took the experimenters, ah, quintillion type collisions to get it right. So launching protons of it was it was calcium and I forget the other element. But they they were launching these protons at each other until, you know, on on random chance one hit and it's stuck. This story that goes into putting something on the periodic table is really interesting, and it's also dangerous and tedious. It's painstaking. You'd have to record everything perfectly, and then you have to wait sometimes many, many years and for your element to go before I you pack and then have the AIPAC rule to see if you're on the periodic table. One of the exciting things that has theorized is something called the Island of Stability, and it's just a predicted. It's a predicted Siris of super heavy elements that just don't decay like they don't break apart like as fast as the ones that we've made. And what could you dream up of some super heavy element? Well, we don't know. It's like trying to think of trying to imagine a color that you can't even imagine. And that's what these elements would be like. It's just the arise when they map it out. There's predictions that eventually if you hook enough protons together, you'll reach a point where they won't break down as quickly. You might get some stabilizing effect of having all of those neutrons jam together. As of yet, it's it's just a theory, but that's very cool, because we can't use any of those new elements like Nyoni Nyoni, um, Tennessean A Gunnison like they just they're gone in fractions of a second like Blink and you miss it kind of thing. So it's not like you. We have a new element like chromium that changes our lives because weaken played it and stop things rusting. So if you're ever wondering a little bit about the periodic table and I think about the periodic table a lot, those were the elements that are on it, and you might be seeing in the next couple of years, some rumbling of some new elements. But for those scientists to get their stuff on the periodic table, they may be waiting a while. I just thought I'd throw that out there because it was in the news again about how exciting and how cool it is. It's almost like an epic adventure to create and get your own element on the periodic table that science news for this week this weekend. Dog science. We're gonna take a look at cats vs Dogs, the epic showdown of the feline versus the canine. And to be honest with you, cats are great. Dogs are great. We're not going to debate, which is better? It's dogs. But it was this a really cool kind of discussion about white dogs and cats have such differing variants, meaning that there are so many different types of dog breeds that look vastly different from each other, and so few cat breeds, comparatively, that do look different. I don't know. Somebody was showing me like this is a cat and they're like Bubble. It's above a black cat. I'm like it's a cat and they give You showed me 10 different cats. I'd be like That's a cut. That's a cat. That's a cat. They're all cats. There's some other cats that are quite a bit different, like the doctor. Evil cats. I think one of them is like some kind of Egyptian Sphinx cat. And the one before that? Well, they're both still Mr Bigglesworth. Uh, handsome kind of Persian, fluffy cat like obviously, they look different. But when comparing two dogs, dogs air completely on a different plane. Right, you tickle. Got Bunsen bringing Bernice Mountain Dog looks completely different from nearly every other dog breed that there is out there. There's some that are similar. There is like he sometimes people are like, Oh, see an Australian shepherd? Because Australian Shepherd, some of them do have the markings I'm like. No Australian shepherds are about 1/3 the size and super nimble. Occasionally, people think he's a Saint Bernard, and I get that, Um, I'm like, No, he's not constantly drooling all over the place. He's a big dog, but he's not as big as a Saint Bernard. Some people are like, Oh, you've got one of those Burmese mountain dogs and I'm like, Yes, he's from the Burmese region. Ah, he grew up with pythons, but again, that's just me being sarcastic. So if you're wondering, there are 42 recognized cat breathes in the United States, according to the C F A and the C F A stands for the Cat Fanciers Association. Yes, that's what it that's what it's called because they're so fancy. History. Cat lovers And in the States, there's the A K See the American Kennel Club, and they recognize around 190 breeds. But worldwide, the World Canine Organization recognizes 340. So there's just way more breeds of dogs. And amongst those breeds of dogs, you have way more variants. Why? Well, the story is really interesting, and it cuts to the short form of it is genetics. Dogs look different because they have had mutations over thousands of years selected for our human ancestors selected those traits and called traits away that we didn't want. Cats are just really new to this whole artificial selection business when when people who study cats look at the different cat breeds quote unquote cat breeds, humans have really only been shaping cat breeds for about 100 years. Yes, there are some cat breeds that were shaped way before that. But the vast majority of cat breeds have only been shaped for the last 75 years, whereas dogs have been shaped for thousands of years. That's a big time difference for genetics. Toe work, its magic If you go back far enough and you can and you talk to anthropologists and people study old cultures and they look at different pets that old cultures had we've been our ancestors have been living with pets for a long time. I mean, you can look to the Egyptian Times and the the pharaohs of old just loved their kitty cats. And I mean, there is that I did a story in Season one about the little frozen puppy they found in the permafrost that maybe he was a pat. Dogs have been living with humans for around 20,000 years, plus or minus of 1000 years or so. Cats. It's a little bit more nebulous, but around 5 to 10,000 years, so they haven't been living with him. Humans is long, okay, but 10,000 let's even give Katz 10,000 years. 10 cats have been pets to humans. Domesticated cats are smaller cats are like sabertooth, smaller cats. I don't know. While ago, let's give Katz 10,000 years. Why wire dogs So much more different. There's so many more different looking dog breeds than cats. And not to be rude to people of cats, but dogs air just more helpful to humans. They just they just are. Our early human ancestors found out right away that they could train a dog to do Ah, whole bunch of different tasks and they could couldn't train a cat to do anything. Of course you can. Cats are great for vermin control, right? But they're not big enough to pull a cart right there. There's no herding cats. They're not going to go there and heard a sheep like Go cat go and the cattle display round and not do what you're telling it to do. You can breed a cat to protect you, I guess. But, I mean, I don't want to fight a cat. Cats are crazy in there, All claws and teeth. Um, but they're not on the same page as a guardian breed dog, right? They don't have cats protecting the penguins on Middle Island, so the humans had incentives to shape these dogs that were way more pliable than cats. That's what it came down to. It came down to the ability of humans to shape genetics, to do what we wanted the dogs to do. Make them bigger, make them stronger, make them more nimble, make them listen. And it's not so bad for cats because cats get to live longer than dogs. And as we shape dogs, there were piggyback mutations that followed some of the traits that we love in dogs pick a random, pure bred dog breed. There are traits that follow other traits. Throat, the breed, dark coats and some types of dog breeds. You want a really dark coat. There is an increased likelihood of that dog getting bone cancer. And that's something that we all we know too well. Like when Chris and I were searching on researching Bernese Mountain dogs, there is a really, in some lines, really high incidence of cancer, and they have, ah, sadly, a very short lifespan. If you don't watch for that And, man, did we do our research, knock on wood with Bunsen? There's that, like as you select for traits, you may have to take a bad trade to get the trait that you want in cats that are that looked really different from other cats. There are piggyback mutations as well, like Persian cats with the squishy faces have way higher incidents of kidney disease, right? So nobody's trying to breed terrible traits into their dogs. But there's been some unintended consequences of that, and dogs and a lot of veterinarian, um, practitioners like vets are standing up and saying, We need to stop. We need to basically stop breathing this into dogs because it's giving them a terrible quality of life. And it's really controversial, and I don't want to name any dog breeds out there. But if you there some of the younger vets that are being trained there, they are telling people, Please do not get this type of dog breed or they're giving suggestions to breeders to stop this trait because there is a piggyback mutation that goes with it. And that's a new thing that this was not happening 20 years ago like nobody cared about this. Or maybe if they did, they wasn't really talked about. In a nutshell. Why do cats and dogs look so different? The simple answer. As I said before his genetics, the more complicated answer is genetics and function. And the more durable answer is it doesn't really matter because there are companions now. Very few dog breeds are needed for the function that they were bred to do. Like we don't have Bunsen Hall carts around for us. So we are working to train him to do that. But it's not like my It's not like my life would be over bunds and couldn't pull a cart. He's literally sleeping on the floor right now. I'm talking. He is doing nothing but being a couch potato, and we couldn't love this fuzzy monster couch potato dog anymore. That's the story that look different because of genetics and function. But in the end, maybe it doesn't matter so much anymore. That's dog science for this week. Hey, guys, before we get to the interview section, I thought, I tell you a little bit about how the podcast is made possible. It's made possible with listeners just like you through our patri on page. The podcast is always gonna be free for you to download and listen to you and have fun with us But the way that we make the show go and pay for the fees and and everything else that goes with running one is through the donations per month on her patri on page. If you're interested, head over there to pay tree on dot com. Backslash Bunsen burner. There's a link in the show notes, and there's four tiers of support. Why else might you want to join the Patri on page? That's for the cool perks that we've got for you there. Each tier has some really cool perks with the top. Tier's getting a shout out in the podcast and play time with Bunsen on our for Bow. On top of that will send you some swag from time to time and postcards a couple times a year. If you want to support us, we so appreciate it. Okay, thanks back to the show on the science podcast. I am delighted to have Dr Jenelle Shane, who has a whole bunch of publications that have been used by The New York Times. Her writing has been wired. She's given Ted talks, and she has a book called You Look Like a Thing and I Love You, which is Ah, hilarious and whimsical. Take on artificial intelligence. But also so educational. How are you doing today? You know

spk_1:   22:10
I'm doing great. How are you?

spk_0:   22:12
I'm so good. Hum. One of the things I was I'm always curious about because the podcast is all over the world that people listen to whereabouts. Are you in the world? Where you calling us

spk_1:   22:22
from? I am calling in fun. Boulder, Colorado.

spk_0:   22:27
Oh, so you are. You were close to the mountains then?

spk_1:   22:30
Yes, right on the edge of the Rocky Mountains. And, uh, now we've got it Interesting spot here where the Continental Divide is really, really close to the plane. So we have a whole bunch of different ecological areas just kind of mushed up against each other, and it's really dramatic to be living here.

spk_0:   22:50
Oh, that's you know, it's very similar to Alberta, like a little bit west of where we live, Um, southwest, because then you were nestled up against the Rockies as well. So that's very cool. Your geography is pretty similar to Albert. I think you're a little warmer, though. Uh, we're kind of cold up here, so

spk_1:   23:07
yeah, in fan, how you do is in the fifties today.

spk_0:   23:11
Oh, nice. Nice. Okay. All right. All right. Uh, just just curious. And for our guests listening. Dr. Shane, could you tell everybody what your education is like? Where where are you? Kind of in your, I guess, your journey through science and technology.

spk_1:   23:29
Well, I went all the way to the journey destination of a PhD. So did an undergrad in, um, electrical engineering at Michigan State. And then I did a master's degree in physics at University of ST Andrews in Scotland. And then I did a PhD in electrical engineering again in Ah, at the University of California, San Diego.

spk_0:   24:00
Well, that's so cool. So you did some of your education in Scotland?

spk_1:   24:03
I did? Yeah. Left there for a year.

spk_0:   24:06
Wow. Um, it's that's one on my bucket list of a place to go visit. I hear it's just breathtaking in places. Did you get up to see some of the sights? Are you so busy with your education that never happened?

spk_1:   24:19
No. The work life balance was okay there. So I have met a bunch of photographers and we went out and saw thing, so yeah, with within a couple of months of me landing in Scotland. I was on the overnight 14 hour ferry up to the Shetland Island. Uh, wow. Blown about by the wind up there. Yeah, there's there. So maney, gorgeous places one of my favorite. The place is kind of, uh, not very heavily trafficked by tourists. Is this double bridge in a place called Rumbling Gorge, where they have an old one lane bridge across this very steep gorge, and then they've got another bridge directly above it that replaced it. So when you take a look at it, it is this weird bridge shape that you'd never see who was a bridge over a grid.

spk_0:   25:13
It's crazy. So you have a PhD in electrical engineering? One of the things I'm always so curious about is when you were growing up were like once you got to post secondary, did something, did something bite you and grab you for that area of of your education. Like was there something in that area that really fascinated you?

spk_1:   25:35
Well, my particular concentration was on photonics, on optics, lasers, things you can do with light. And I thought all that was really cool. So for some of my master's degree. I was working on this thing called optical trapping, where laser tweezers is also called, where you wouldn't think that you can actually use light to push things around. But on a microscopic scale, it works under some kids conditions. You can actually pick up small particles like about the size of, ah bacterium, and you can move them around in a focus of a laser beam. And then, if you can do fancy things where you can use a hologram to turn your one laser beam into hundreds of programmable laser beans, then you can have these particles trapped in three D. And so people find that interesting for studying particles suspension, for example, so you don't have to rely on the particles doing a similar thing every time. If they're bumbling around randomly, you didn't start them from some known configuration and do that repeatedly and do repeatable experiments. If you can put the particles where you want them. Of course, you can also use them to make a holographic you know, Tetris game out of these microscopic particles, or and and the statue back. I did not, but there is some entertain Andrews that research group you know, was known for golf. So there's some that made a swinging golf club that hit a little microscopic, uh, golf ball. And they also said that the, uh, the dance ah, strip the willow. So the world's smallest, the willow, played out micron sized particles under a microscope.

spk_0:   27:30
That's hilarious. So I guess it's true that that's one of the things we always tell because I teach high school chemistry. And that's one of the things the physics teachers always tell the kids. You know, you stick with you, stick with physics long enough, you may get the play with some crazy lasers, and this is the proof of it right here. So all of the students toiling through physics, guess what? You could stick with that. You could pick up things with lasers, but is way more interesting than that. I'm sorry I distilled it to something silly like

spk_1:   27:59
that, you know, that is an attraction.

spk_0:   28:02
Did you get a kick out of it? And I don't know if you watched Austin Powers when Dr Evil went laser like he with his ah quotation marks on. Imagine somebody who studied that would find that hilarious.

spk_1:   28:13
Yeah, We're often people would often proposed to put our experiments on sharks for one reason or another. I

spk_0:   28:22
love it. I love it. Oh, I'm so glad. I'm so glad I asked about their conversation. Went this way. We have to get back to the main part eventually here, But that is that's just great. You'd have to have Ah, pretty good power source and have to be waterproof. They're on a shark. I mean, you have to have a waterproof fleas or to put it on

spk_1:   28:42
shark. Yeah, there are practical considerations,

spk_0:   28:45
practical and ethical, there to some forms to fill out some forms. Yeah, a lot of paperwork. One of the things we have. One of the things that jumped out about you and when I was looking for people do invite on the podcast was your work with artificial intelligence. That's like a huge area of discourse all over the world now, both in science fiction. And from a practical standpoint, that's not something that the general population probably has a good handle on because we don't think about Ai. Aside from Sky Net in the Terminator movies, could you tell everybody a little bit about what ai is. I know that's a really broad question, but I was hoping you could help me out with that.

spk_1:   29:29
Yeah, it is a tough one to answer just because, as you said, there are so many different definitions of it. So you've got the AI front that in many ways is the most familiar. The sky net. You know, R two D two. All of this science fiction level. Ai, that's at least a smart is a human being. And that kind of a I is nothing like what we actually have in real life. Uh, so the stuff we have in real life is way, way simpler, way more limited. Ah, but you know what, actually, is it? Ah, no. Is it depends who you ask eso. That was one of the things I had to kind of figure out when I was sitting down. To write this book is okay. Is gonna be a book about Ai. What even is that because, you know, some, You know, if you ask computer scientists, computer programmers, they'll say one thing. But then sometimes in marketing, you'll get people saying that their thing is an AI, and it's actually let this really simple spreadsheet or sometimes it is literally a human being hired to pretend to be an AI. So there's always that happens, happens that happened

spk_0:   30:42
literally. Like, What's his? The guy that was actually in the C three po suit?

spk_1:   30:47
Yeah, yeah, literal person in a robot suit. And they call it pseudo ai sometimes. And they sell. Yeah. You know, sometimes there isn't a I but the human takes over if things get too tough for the AI which happens all the time because our the eyes we have today are not nearly as smart as C three p o, for example. Ah, so yeah, So there is me, like, looking at okay, a eyes all over the place like this, all these different definitions And I ended up just kind of going back to let's use what software developers computer scientists use when they're talking about a I and that is a specific kind of computer program that instead of being told step by step instructions on how to solve the problem, you give it the goal, and then you let it figure out how to solve that problem via trial and error. And that's also sometimes known broadly as a machine learning program Machine learning algorithm.

spk_0:   31:48
Okay, there we go. All right?

spk_1:   31:50
Yeah. And then under that umbrella of machine learning algorithms, you have all these other things that people may have heard of, Like neural networks and evolutionary algorithms and generative adversarial networks. All these sorts of things fall under that machine learning, Uh, algorithm, Umbrella,

spk_0:   32:08
your book is titled You look like a thing and I love you and the title of the book within the first couple pages. Hopefully, I'm not spoiling it, but what hooked me was you were trying to teach an AI how to do pick up lines. Is that kind of the gist of the first bit of the book?

spk_1:   32:26
Yeah. And it's not a spoiler so much as an explanation of where the heck that title came from. Because the title is one of the used pick up lines at Ah Eh eh, I came up with

spk_0:   32:39
some of the pickup lines. You're a I came up with the AI in quotation marks. We're We're hilarious, like one is you must be a triangle, cause you're the only thing here. So how how did the like and this happens over and over in the book Took I think you try. You're explaining to to everybody just how bad ai is doing simple tasks right off the

spk_1:   33:03
bat. Yeah, the problem is one this simple tax are a little bit beyond what the's a eyes can handle. And usually when things tend to get broad and tend to get complicated or there isn't enough data, uh, there isn't enough brain power. I kind of go through, like, where are all these sources of failure, like, Why are a is of today bad at things sometimes. And, yeah, this is one of these ones when I'm generating text like pick up lines. What I'm doing is giving ah whole chunk of existing pickup lines too, a algorithm whose task it is to learn to imitate the text and given maybe one letter, it has to learn to predict what are the letters that come next in the typical pickup line? But, you know, especially this particular A like I gave it these pick up lines and that was all it had and had no clue that it was doing English or that these were even like words, sentences like for all of new. This could have been a computer program, Britain and Finish, and it had to try to figure out from scratch like what kinds of letters go together. Like, When does it use vowels? Maybe it doesn't have as many asterisks in these pick up lines as you might, I think if you were guessing symbols at random and it's checking its own guess is against actual existing pickup lines to see how well it does at predicting them. And so without any like direct input from me, other than here are the examples, it gradually gets better and better at predicting letters that come next to each other so it could. It learns gradually to spend it. So it learns gradually to spell the word you, for example, and learn to spell the word love and I things like things that come very frequently and it'll learn letter combination. So that's how you get words like Triangle, which were definitely not in the training data. But I thought it might be pronounceable and this is true. Ah, but this is also how you get this algorithm spontaneously spelling words like fart, which are not in the trading data, which are pronounceable, and it doesn't have information about what kinds of letter combinations and words ought to be avoided. Like all of those is exactly what's there in its data set in its examples. A pick up line?

spk_0:   35:43
No. Do you have to tell the AI it's on the right track? That's the part I'm I need your help with. Kind of like wrapping my head around is, once it starts to do this, how how does it know it's on the right track? And then it's instead of just making gibberish.

spk_1:   35:58
So it is reserving some of the pickup lines that it can test itself against. So, uh, every once a while it will check and see how it's doing. It'll start with a letter. Why, oh, and if it adds cue to the next to the end of that, then it will look at his guess and say, Oh, that wasn't right. It actually was the letter you at the end of why Oh, so okay on. I will adjust my own internal structures the next time I guess a you or more likely to get you in that spot. And so it's just this gradual refinement that it does on its own. Just by looking at this training data, of course there is a There is a danger here, too, in that sense is doing all its guessing and is trying to imitate trying to predict what the examples that given it a technically a perfect solution to his task would be to memorize the entire list of pickup lines that it gave it and spend it back to me word for word, because that would be predicting each one of those perfectly. So it's become Sometimes I gave battle between me and this algorithm wants to memorize everything. And me, who wants to give it so much stuff that it can't possibly memorize It will have to come up with some general strategies that are gonna work a little more generally.

spk_0:   37:22
So you're, like, bad. I know, like if it just like repeats what it you're like your you get frustrated, you like. Ah, stupid. I

spk_1:   37:31
like OK, yes, technically, that was correct. But that is not useful to me. And it's not like I can. It will adjust itself. It is not like I have a way to tell it. While it's training. No, no, be no, don't do that. Don't copy needs. Like I just have to start over the start, the training over from scratch and try something different. Next time

spk_0:   37:51
was that the idea for the book was to try and and show the public the process that the current AI stuff has to go through.

spk_1:   38:02
Yeah, that was definitely part of it. And in fact ah, you know, as I was outlining the book at first, that's what I kind of had focused on was how do you generate text and what can you learn from failures and generating text? But as I got farther into making this book, I realized it would be more useful if we covered other kinds of things, like image generation and the kinds of algorithms that do self driving cars and image recognition. And, ah, build language translation. So, really, I am talking about a whole broader set of kinds of algorithms right now, but what kind of underlies this motivates this is I would get people reading my blawg, reading these things like the pickup lines and saying how wait is this a I Because isn't a I supposed to be smart. That's what

spk_0:   38:58
I thought. That's what I thought Before I read your book. I was shocked. I was shocked. So terrible. It is. Sometimes

spk_1:   39:04
it's a lot of smoke and mirrors. That's a lot of ah, cleverly setting up the problem to make the A. I appear as if it's smart and choosing a problem that's narrow enough that the AI is not biting off more than it in two. And eso you get the sorts of things where ah out things like playing chess is actually a kind of narrow problem that I could do really well on. So if you're playing chess against an AI is going to seem really, really smart. But then there's things like folding laundry, for example, which is actually a really, really tough problem. And really broad, because if you think about it, there's all these different kinds of clothes. There's all these different ways that clothes can fold. There's all these legs and shapes and textures and things, and we do not have AI today that can handle laundry or handle being a personal assistant and making phone calls. Bacon barely transcribe voicemails, and they still make a lot of mistakes I

spk_0:   40:07
loved in the book. The you know, it starts off, of course, with the pickup lines. But you do talk about lots of different areas in a I a couple of my favorite parts of of your book, and I'm gonna ask you maybe what your favorite part was. I love the section where was talking about how a I would make a sandwich, just how complicated it would be for, like, an artist like a a I to make a sandwich from scratch because it may find the right ingredients alone. And then it's like, Yep, those air delicious. And then it's gonna throw them in all the time in all of these other sandwiches where they're not delicious together. And that would be very confusing for it, because it be like, Wait, this was good before. Now it's not, Um, yeah, and I just I just thought that was fascinating. And then I think obviously you gave the A. I like it could use whatever once it was picking like crazy stuff that put in the sandwiches.

spk_1:   41:00
Yeah, I had. I had a lot of fun with that thought. Experiment was just like, How can I illustrate the inner workings of an AI because I really wanted to kind of get under the hood and show how this can even work. How can something to in itself and how could really simple parts be making more complex decisions? So yeah, it actually I picked sandwiches in part because it came down to what can I draw? Because, of course I had. I illustrated a lot of this book and I had planned to illustrate these sandwiches and so that kind of all right, I think I think draw sandwiches.

spk_0:   41:40
I love it. I was reading my aunt. My younger son is in marching band. So I was chaperoning when one I was reading your book like a couple weeks ago and I got to the sound. Which part? And I burst out laughing and the guy sitting beside me I could not explain to him why it was like it was like I was trying to explain him. This book I was reading and it was just I was like, speaking of the language, this guy So but But it's got some really funny parts. What was there an area of the book that you just enjoyed? the most

spk_1:   42:11
writing how I had a lot of fun in particular with the section about a eyes that hack their own simulations. Because this is a you know, it kind of reminds me of all the weird things that biology will do. So biology will, if there's an energy source like some bacterium will evolve to use it like Oh, is the deep sea, then farting hydrogen sulfide. Let's see, we can eat that or oh yeah, let's see. Let's see, there's radiation coming from the sun Let's see if we can eat that. And there's, you know, just from these this really simple evolution that comes up with these ways to exploit the environment. And what I love is that since the early days of doing machine learning programming in computers and this goes all the way back to the 19 sixties, Ah, we have had even really simple programs. Ah, hack their simulations. And it's as if you have bacteria that are evolving to take ah advantage of energy sources. But in this case, the energy sources will be things like math errors. So you get these virtual organisms that learn that in their simulation, if they move really fast. Then there will be rounding errors from the math that governs movement. In each of these rounding errors is going to give it a little bit of free energy. Or you'll get other organisms that learn that if they wiggled just so they can glitch into the floor and then the math that's supposed to stop them from occupying this vein spaces, the floor realizes, Oh, there's a problem. Pops them out of the floor. Hey, that's free energy. That's free momentum that they just got. And so you'll get these organisms that just glitch all the time into the floor.

spk_0:   44:10
There was one story of ah, a simulated. I forget what it was. I'm sorry. Maybe you can probably clarify it where it figured out that if it did like a whole bunch of different steps, it would actually cause the program to crash, therefore, winning and beating the player. And they just kept doing that over and over again so it would win. And that was like an unintended consequence of trying to beat the player.

spk_1:   44:33
Yeah, there is a program that ah was accidentally given access to the answers when it was trying Teoh I think it was sort numbers or something like this. And so it learned that if it deleted the answers, that would get a perfect score. There is another program that was supposed to put a list of numbers in order is supposed to sort them. But what is programmers told it specifically to do? Waas, ah, eliminate the sorting errors? And so it found that if it eliminated the entire list, there would be zero sorting errors and it would get a perfect score.

spk_0:   45:15
The more we talk, it just sounds like a eyes like it's almost sounds like you're trying to work with, like a little little toddlers like toddlers, that you trying to corral this toddler to do a thing, and it's just takes the path of least resistance.

spk_1:   45:30
Yeah, it is a, you know, and it understands less about what's going on than your typical top their swells. So

spk_0:   45:37
that's right. That's right. You talked about how it's it's brainpower is. So there was a animal you referenced that would that was about this intelligence oven ai

spk_1:   45:47
about. So this is a very rough guess. Just based on, you know, ballpark. How complex are the virtual imitation brains that power some of these ai algorithms. And then how complex How is the brain of various creatures? And my best guess, Ah, is that we're somewhere around the level of unearth word. So our imitation who reigns like these neural networks, as they're called, they have more virtual neurons than an earthworm does. Right now, we may be somewhere along upwards of a honeybee getting toward frog right now. Ah, but each one of these virtual neurons is a lot simpler than the real thing. And so this is me saying Okay, well, accounting for that accounting for the fact that one rial neuron is actually like a complex neural network all by itself. I'm gonna wildly guess we're still somewhere around earthworm level. And sure enough, you'll get some examples where people will train pigeons to do the things that you can get a I to do now. So image recognition pensions are actually pretty good at it. And if you can figure out how to motivate them properly trained them properly. Ah, there is a study where the people were getting pigeons to identify the pictures off cells with breast cancer and pictures and showing that they actually did pretty well. Like I think it was about approaching as well as people do.

spk_0:   47:22
That's awesome. Bird's Air hack a smart, though way where we learn about them. They're just they're on their own plane, and the more we learn about them. It's so

spk_1:   47:30
cool, I said. The pigeons are way smarter than these algorithms. But on the other hand, the algorithms get too deep, devote their entire worm power to one problem. They don't have to also worry about cooing and shuffling around and thinking about dinner.

spk_0:   47:44
They're not as cute. I'm gonna say that pigeons pigeons can be pretty adorable.

spk_1:   47:49
You stuff. You stick a pair of googly eyes on any old algorithm and you'll sympathize with it.

spk_0:   47:55
Oh, that's true. Yeah, your that's right. The art in the book is pretty whimsical in, and it does. It does humanize these pretty the's programs in their terrible issues. Well, speaking of pigeons and pigeon zehren animal, this is a good kind of Segway into the question. We always ask our guests, and that's for a pet story. Do you have one you could share with us?

spk_1:   48:15
Yeah, So the pet in question has finally wandered off. She's but the first part of this, uh, this podcast interview kind of perched right on the desk in front of me. Uh, and she is a £12 tortoiseshell cat. And this particular cat really likes train rides. So not actual trains. Because she's an indoor cat. She doesn't go outside. But if I start pushing a chair around and making little to to noises that health, she will run up and jump on that chair and right around for as long as I care to push it.

spk_0:   48:56
That is a little what's what's your cat's name?

spk_1:   48:59
Her name is Char

spk_0:   49:00
Char. Aw, that's so cute. How did this were you just like moving chairs for company and the cat hopped up. And And that was a thing from that day forward.

spk_1:   49:10
Yes, there isn't very much that gets her to suddenly perk up and come galloping over. But it turns out pushing a chair is one of those things.

spk_0:   49:20
How long have you had your cut

spk_1:   49:21
off? Had her for three years. So she's 11 now, I think.

spk_0:   49:26
Okay. You got her when she was eat? Yep. I'm not sure Bunsen would fit on a chair. He's pretty big. Maybe like lazy boy. Yeah, but but But our dog Bunsen he loves he loves car rides. He lives for car rides. If you really were for going anywhere, he is ready to go on in a car ride. So, um, may I know what it is? Some animals just love that they like the humans or machines to do their work for the moving them place to place. They're smarter than us, I think.

spk_1:   49:53
Yeah. I mean, he people have, like in the eventual, you know, hypothetical ai takeover to what cafs have basically already accomplished. And that you know what? We will be spending our lives catering to the winds of an algorithm. Specifically, it doesn't have to even be a complicated one. Like we will go and rescue our Roomba from under the couch over and over again. We will. Yeah, way so the verbose have basically taken over already. In that sense,

spk_0:   50:24
you put some googly eyes on a Roomba. It's adorable.

spk_1:   50:28
Yeah. Yeah, that's I think that's one of the things that makes you know robots and algorithms. They seem smart, smarter than they actually are. Is because we have this human willingness to play along and to see some kind of thought agency and these things, especially if they have googly eye side on them. Oh, my goodness. But often there is not very much there, or it is actually a person remote controlling that meal delivery robot, for example.

spk_0:   50:59
Thanks for sharing. That's that story about your pet. Um, one of the things I was gonna ask you about is your Is it a blawg ai weirdness dot com? Um, that is it that you're a catch that you're attached to. Could you tell everybody a little bit of both? That is, I did a little dive into it. Ah, a couple days ago. It's pretty

spk_1:   51:16
cool. So yeah, weirdness is, ah, a come for block Where I post experiments that I do with a I so usually involves giving a i human things to imitate and because it's not very good at it or make some mistakes, they come out weird, but sometimes in an entertaining way. So I gave it the name so entertaining flats, for example, to imitate. And this is actually in partnership with Ai animal shelter that needed to name some kittens and the the names that the A I was coming up with were like Mr Tinkles and Recchi on There's Cat called Sofa. There is another cat called Stocks Poops, Jack Salute pickles. Yet that was another one.

spk_0:   52:12
That's awesome. Is this where I it was in your Ted talk? You were talking about ice cream flavors to, and it just like, butchered some of the names of the ice cream flavors like there's one ice cream flavor that was strawberry cream disease like it was just trying its best. But coming up with a horrific it's pretty flavor.

spk_1:   52:31
Yeah, and that was a case of it. Not knowing that perhaps the word disease is not to be used in ice cream. It just knew it was likely to be pronounceable.

spk_0:   52:41
And that's the idea. Very awareness is of people who people can check that out and they'll see all these whimsical failures of AI or just the process there.

spk_1:   52:49
Yeah, and I poke around with image recognition algorithms to, uh, or image generating algorithms. There is, ah, one while paper I really liked recently where they were demonstrating an image generating algorithm and they demonstrated with human faces and then they also demonstrated it with Internet cat pictures, and it's really interesting to see the contrast between the two is the human faces is a fairly well controlled data. Settlers. Just the human face is seen from the front and the cat pictures as well. Imagine your typical handful of in that cat pictures like it could be a bunch of kittens or could be a person holding a cat. Cat could have 1/2 the can't be curled up, and that was a lot more for the AI to try and keep track of and figure out. And so it's cats would be weird. They would end up as he's amorphous blobs with far too many legs or maybe a floating tail. It would try to do the kitten clouds, and for some reason it would manage to do one kitten correctly, usually more or less. And then the rest would just be kind of like a smear of eyes and for and very, very alarming. That would be nightmarish. Oh, yeah, it's so cool.

spk_0:   54:08
The heart. That's awesome. I'll have to check that. I miss that. When I was checking the site, that's that's cool. I'll make sure that the Blawg is in the show notes to the podcast, so other people have easy access to it as well. I have a question for you about a I in the future. So I guess where we are right now is nowhere near the science fiction you know, that I love or were like near future ai like like the the movie Her with Joaquin Phoenix, Phoenix and Scarlett Johansson. I don't know if you see if you've seen the movie her where the operating system is voiced by Scarlett Johansson, who's an AI and that's like sold is near future. Do you? Do you see that near future coming for us, where our future way off, where AI gets to be That point where it could be used for good or evil? Um, I think we have to worry about that. Or is that like many, many, many years off?

spk_1:   55:03
Well, the kind of AI that's as smart as the one in her or ex mocking Ah, like these are definitely science fiction level a eyes on. I'm kind of mind with the, uh, researchers who think we won't see that, uh, anywhere in our lifetimes. Ah, and possibly maybe not ever. Um, so that's really in the realm of science fiction. And I know there are some people who think that we are, you know, 10 20 years away from this. But if you look back over the history of a I research, you will see people saying that we are 10 20 years away from a human like thinking machine almost all the way back. And I think we as people have this really persistent tendency to underestimate how smart we are compared to these algorithms and how far these algorithms have to go. And, you know, the history is littered with all of the's failed machine learning algorithm applications where people tried to build a voice assistant Facebook tried to build a Facebook AM assistant that was supposed to be able to like, do anything you ask it to do, look up movie tickets or make plane reservations. And they ended up giving up after a few years when they realise the problem was much harder than they had realized, and they were going to always have to have human step in and rescue the eyes. So I think we tend to make these kinds of mistakes and not realize how much work we're doing when we do something simple like, uh, you send a message to someone, make a phone call. Uh, there's actually a lot of complexity going on there, so I think we I think so. I think that kind of science fiction future is really far off in the realm of science fiction. I love reading science fiction, so it's interesting to think about what that might look like someday. But I think the AI that we have to worry about using, for good or for evil is today's actual AI. So we're all we've already got algorithms that people are using to control a lot of different aspects of our daily life. So, um, they come in contact with algorithms that recommend videos on YouTube or that are recommending new songs on Spotify that are tagging our photos or sorting are ah, sorting. Our junk mail like these are all applications that are built in lot of smartphones now have a eyes that are making the pictures turn out a lot better. So if your smartphone has something like a portrait mode, that is Ah, I have a lot of heavy lifting They're doing a lot of heavy lifting there to figure out what the background is and how to artfully blur it. So it looks as if it was taken with a really expensive camera. So we've really we have a I algorithms doing all sorts of things all around us. About some of them are making really important decisions about us, like, you know, hiring algorithms, for example that are deciding who gets an interview and who doesn't. Ah, based on maybe scanning a resume, maybe even looking at a video of somebody. And we don't have a lot of evidence right now that these algorithms are actually fair and we have a lot of ways in which we know that they might not be fair. Specifically, if you remember how these algorithms work is, a lot of them are copying examples of human decisions. And if the humans decisions have bias in them, the A. I will copy the bias because it doesn't know any better. It doesn't have enough context to know that this isn't what we want. And so we end up with a lot of algorithms that are biased and ah, faulty in several different ways. like, maybe they're biased and they don't work. Or maybe they do work. If what you wanted was an exact copy of the decisions humans would have made keeping in mind, humans are biased.

spk_0:   59:42
So I believe in your Ted talk you talked about. Um this is a quick aside. This is a really interesting area. Like where a. I can, you know, potentially have unintended evil consequences where they trained it on resumes. But this place that was using it on Lee generally hired men so it would train the AI. We use the data set to exclude women, basically, because there was like, Oh, well, we they've never hired women or I've never seen the world women before. So if I see a resume with that, it goes in the do not hire pile is that is that there's idea

spk_1:   1:0:18
it is. And this particular algorithm is really sneaky about figuring out who was a woman as well. Because, of course, you know, people don't generally put that explicitly along with a resume. But it was able to look for things like word choice or able to look for ah, historically women's colleges or to look for things like women's softball team or even sokrati and use all of these clues. Yeah, and so you'll get algorithms that will learn how to discriminate based on race is well, even if they're not explicit explicitly given data of the race of people who you know humans were making decisions about, it can look up. You can often use other clues like what's their zip code? Because, uh, especially in the United States, a lot of places are very segregated and by zip code, And so based on someone zip code, you form a fairly good guess as to what race they are. And so algorithms would use these to implement these kinds of Ah, you know, the kinds of biased decisions it would see. Well, you say, Well, I don't know how to predict other than to say, Hey, everybody from this neighborhood seemed to not be getting loans. So let's go ahead and implement that.

spk_0:   1:1:46
So we don't necessarily have to worry about Skynet killing us. We just have to worry about the AI being a jerk. Yeah, like how humans are

spk_1:   1:1:54
jerks. Yeah, and there are some jerks who are using a eyes to do jerky stuff too. Oh, that Teoh. Also be some cause for concern.

spk_0:   1:2:03
That's terrible. Well, if you know that I'm I'm so glad we had this talk about the future of AI. Um, because you're right. It's really cool. And it fills you with wonder to read science fiction. And I thought that, you know, you're always you're always optimistic that things are gonna be so cool within your lifetime. Um but I guess that also should alleviate people's fears that this AI is going to go rogue and and, like, take everybody out. Um, But I guess the only thing we have to worry about is them taking you out because they are They're biased because of horrible hiring pat practices that were used Point that people already So Ah, well, that's a negative. I'm sure there's so many cool things ai is doing that we don't even think about. I'll tell you that portrait mode on my phone. Um, I didn't know that was an AI, but man does that. Does it take such cool pictures like, That's amazing. It does all of the light shading everything. It just is perfect.

spk_1:   1:3:05
Yeah, and they're they're also, you know, some of this is based on probability. It says. Okay, well, the lighting. I've been told that the lighting in a good photo looks like this. And so I will, after the lighting of the picture you took Baker look like these examples of good pick pictures, and you can get some really studying photography. That way, it's It's pretty impressive.

spk_0:   1:3:28
Well, we've come to kind of the end of the podcast, and one of the things we always ask her guests at the end is for super fact. Um, a super fact is anything that you can share with us that is kind of mind blowing. Do you have a super fact force? I know our conversation has been so interesting already

spk_1:   1:3:46
I have a short story on this is one of an algorithm hacking its environment, but it's one of my favorite ones. And so this goes back to actually the 19 nineties, and there was still student who has had the assignment to build an algorithm that could play tick tack toe and to make the problem a little bit more challenging. You could play. You're playing tic tack toe on an infinite board, and not just the standard three by three. I guess it's called Knots Across in some place that game with the X's and zeros and things. And, uh, yeah, so, uh, the and they would play remotely against Ah ah, there other computers that would connect in and each of these algorithms have it's different strategy to try out. And so one of the students said, Okay for my algorithm, I'm just gonna have it use machine learning to evolve its own solution to this problem. I won't tell it exactly how to play tick tack toe, but let's see how it does. And they discovered, came back and discovered that the algorithm was winning almost all of its games. He was just trouncing everybody else. And when they went back to find out what it's winning strategy Waas. This was, well, it had managed to take advantage of the fact this was an infinite board and that it was playing remotely against its opponents, and so his opponents would make a roof whatever. It would completely ignore that move, and it would just pick the move that was farthest away in numbers that it could possibly think of So would have this huge number of coordinates like, All right, My next move will be five million miles away, and then the opponent would then get the move back and say, All right, now, let's take a look at the board. So here's my move. And where's my opponents? Move and okay, Drawing, drawing, drawing, drawing, drawing. And the board would get so big that it would crash when trying to render this huge five million mile board. And then the machine learning algorithm would win, and

spk_0:   1:6:03
it just couldn't. I just couldn't handle it. Is couldn't handle it. I will. That's so cool,

spk_1:   1:6:07
unintentionally sinister. I mean, it remotely learned to remotely crashes opponents computers so that I could win that. It's but yeah, as it's delightful. Yeah,

spk_0:   1:6:20
I like that unintentionally sinister. My son is playing a video game called While he was replaying it. That's called her Horizon. Zero Dawn. I don't know. It's like a post apocalyptic type video game, and that's the cool thing that happens in it is. The humans developed this thing to save the Earth, and the algorithm decided the best way to save the Earth was just to get rid of all the humans. So it was on and it had no. And it wasn't like evil. It just that it solved the problem. The humans asked the salt. So your solution was the beat. This person, that tick tack toe. So you crashed the system. The win every time. That's funny. Yep. Well, thank you so much for sharing. That's a perfect. And also, thank you so much for being a guest on the podcast. This has been such a cool chat. We've kind of gone over time. But, man, I could listen to you talk about artificial intelligence for hours, but we don't have that kind of time. Um, your book is you look like a thing, and I love you, and I believe you can get it just about anywhere. Is that

spk_1:   1:7:19
correct? That's correct. And if you like signed copies, you can get him from the Boulder bookstore. They'll they'll mail in 10 to 1 of them.

spk_0:   1:7:29
That's awesome. And where can people find you on social media?

spk_1:   1:7:33
I'm on Twitter at Jenelle. See, Shane, I'm also on Master Don thin. If you do that, I'm Jenelle. See Shane at wandering dot shop. Say science fiction. Uh, community there and I'm sporadically on instagram that as Ah, Jenelle Shane Onda. Of course, my tumbler Blawg. You can get to me ai weirdness dot com

spk_0:   1:8:04
Well, thank you so much for giving up your time to talk to us about artificial intelligence and maybe bringing us back down to Earth from the lofty ideals of C three PO and Star Wars, making it more real for us.

spk_1:   1:8:16
No, thank you. And I'm so glad you enjoyed the book.

spk_0:   1:8:20
I loved it. I loved it. One more Plug for me. Get the book. It is so cool. It is so whimsical. It is so funny. And you'll learn something too. All right, Take care, Dr Shane.

spk_1:   1:8:29
All right, take care. Nice talking with you. Bye.

spk_0:   1:8:31
Hey, everybody. I just want to give a quick shoutout to our sponsor genius lab here. Genius lab here is giving everybody 10% off everything at their store. If you use the code Bunsen 10. That's B U. And S e n 10 genius lab here has a tonic. Cool stuff. We love that. They sent Montana bandanna That says PhD emotional support dog is so cute. Also, they have these little wallet sized stencils for doing organic molecules. And if you're not a scientist or a science teacher, there's gear at the site that you would like Aziz Well, including sciences for everyone, stickers and a whole bunch of other stuff. Check them out. That's at genius lab gear dot com. It's time for Rue Are Wow, and I've got a guest host again. And it's another Lindsay Thurber teacher. I have Allentown on the line. How you doing today, Alan? Enjoying this beautiful, beautiful weather we're having here. Just feeling great that winter is finally over. It has warmed up and melted a ton in the last week. Uh, excited. Finally. Just get back into the garden here. You teach. What? At Lindsay Thurber. Ellen, I am a physics teacher, so I mostly teach. Yeah, I teach some general science classes, but physics 20 Physics 30. So the Elevens and twelves, they're kind of like That's my That's my pet project right there. So the Americans that are listening who don't understand great 11 and 12. That's like the last two late years before you go off to post secondary. What is it? Junior and senior? I have no idea. I don't I don't know. I think freshman's like great nines or tens. Well, it took me like a solid year to wrap my head around the fact that Alberta calls it physics 20 physics 30 as opposed to 11 and 12. Right, Because you taught in B, C, B, C and Ontario. You just had a little bi ble to write. Ah, little we won. Yeah, he's a 3.5 months old. Gabriel. Yeah, this, uh, he's starting to tedx now, so that's, uh, he he doesn't like to sleep very much and cries a whole bunch. So that's great for me. I was going to say this. You've been teaching from home and having a newborn new newborn That's awesome to spend. You know, you don't have to worry about commuting as much as kids get older. They become newborns to sleep all the time, and they're usually pretty pretty good. You're probably now into the not so sleeping time when he does go down. It's not for very long. He'll sleep maybe, like 20 minutes half an hour before he's up again. And the thing is to he doesn't like not being held. So for putting down that's common chair, it just makes life so tough. But really working from home is is awesome. Like when else am I going to get six months off? Essentially, like paternity leave? You've been doing some great stuff with your physics, though, for the kids. The kids have been really enjoying it and before we get to were while you have two dogs yourself? A. I have a £250 American mastiff. Thin again. He's is getting up there in age, so we started to slow down a little bit. He's 8.5, but it's amazing how, like even when he gets excited and he starts jumping around how he's able to move all that mass, it's, ah, he makes Bunsen look tiny. Oh, he's Ah, he does that to most dogs. Even especially, it's hilarious to see him with our £13 miniature schnauzer Young, when they started his name again. His name is Christopher. All right? Yeah. My father in law has a miniature schnauzer named Doc. Their cute little dogs. Yeah, he's a little bit yappy, but we were working on that. I don't know if you can. I don't know if you can stop that with those type of dogs. They were bred to be a little yappy watchdogs. Yeah, and it's good for if whoever we're worried about, like whether the mailman's coming or not. Uh, all right, you ready? Get started with war. Wow, I am alright. Today's category is artificial intelligence. Because the guests this week was Dr John L. Shane, who is a, uh, who studies artificial intelligence. So that's the theme today. Awesome. I have no idea about any of it. All right, here's the first statement Humans air currently better at diagnosing. If a patient has skin cancer, they are better than an artificial intelligence program. All right, that's the first haven. So it's humans are better than an artificial. That then an AI program at diagnosing skin cancer. Okay, that's right. Okay, It's that statement. One statement to in a survey by mindshare. 60% of people stated they'd rather talk to a chat, but then a really person when they had an issue. Okay, So rather talk to somebody who's not really. They'd rather talk to a machine than a human. Okay, there's some humans. I'm sure I'd rather talk to that. I'd rather talk to him Yeah, I know like this isolation thing has been like it's been tough, but it's also been kind of nice. I could just, like, not be part of society for a minute. It's great. All right, I heard statement the turning test. It's famous for being able to determine if an A I can pass is a human. But most people don't know the specifics to it. Turing's tests had you talk to four people for 25 minutes and then attempt to find the AI within The group had to talk to four people for 20 minutes. This is a tough one. You want to recap the statements or do you think you're pretty good? Okay, so it was Ah, Turing test we had to do with, um People prefer talking to bots than to humans. And the 1st 1 was Would again, humans are better currently better at diagnosis King. If a patient has cancer than an AI diagnostic program, skin cancer. I have no idea. This is a shot in the dark. So too, are false. One. Yeah, yeah, straight. So I think I'm gonna have to go just based on what I prefer. Yeah, I would rather talk to a bought, then talk to a real person. Okay, Final answer. Yeah, I'm going. Final answer on that. Okay, let's take a look at the first statement. Humans are currently better at diagnosing of a patient Has skin cancer that call so great a eyes are better than humans at diagnosing. If a patient has skin cancer and they're more accurate, human doctors are only human. Doctors are only 87% accurate. And the EI program that they designed for specific types of skin cancer that's just impressive right there. 95% of anything. Yeah. There's something that's actually 97% at a type of cancer on its blood cancer. You know what it is? Uh, the I dogs, dogs. Dogs have been found once. They've been trained for the sent their 97% for blood cancer, and it's spreading like they're moving it to different types of cancer because it seems to be similar across the board. So they're training I potentially to smell chemicals, which is very cool coming from the world of physics where nothing ever works. 97% success rate. That's amazing. Good. Okay, so we're down the two statements Town. We've got the 2029 year, 60% people rather talk to about and we've got the Turing test. So let's take a look at the Turing test. That statement is else. So you're right, people, people do. More than 60% of people do prefer talking to which, at but rather than a real person, which is crazy now turns test was two people you and the AI for five minutes, and then they had Teoh augment that test because the train test has been passed a few times by one on one conversations. So this another inventor can tweak the rules. So now it's a group of five people and talking for 25 minutes, and I think that it's ah, as Ai gets better and better and better, they're gonna have to keep on adjusting that right. The one that beat the Turing test first was an AI that was a child. So it was like a seven year old kid. And what tricked the people was when they're asking the kid like a tough question, that kid was like, I don't know, and then people are like, Oh, it's just a kid That's okay. And that's what fooled. Oh, right. That's what fooled people was. That was a voice of a child. And it made sense that the kid wouldn't know really crazy answers. It's not in what the The a I answered. It's what it didn't answer. Yeah, Yeah. Have you seen What's it called X A machine? Oh, something that's not gonna X market. Yeah. Good movie. So good. I watched it just last month. Yeah, it's got that I Isaac something rather Isaac, Oscar, Isaac or something like that. Popo from the Star Wars movies. You're going to say Po from the Teletubbies. No, no, no. Well, you could also be from from a kung fu panda. There's lots of So I'm I'm right. I'm excited about this. I was really was really nervous about this. So you can bug Sylvie if you want, But she's so nice. I don't know if you'd bugger, but definitely. But Doug Walton. Oh, for sure. Let him know. Yeah. Yeah, for sure. Let him know. All right. Talk to you. Thanks for being my guest host. Well, thank you for having me on. Awesome. All right, take care. Walking around my house aimlessly. Yeah, OK, I'll talk to you. Talked about. All right. Sounds good. See you later. Thanks for tuning into another podcast episode and special thanks to our guest, Dr Jenelle. Shane, make sure you check the show Notes in the podcast for links to her hilarious and educational book. You look like a thing and I love you. At the end of the podcast, we also give a show to our top tier patrons on patriotic. We couldn't do what we do without their support. They are Andrea Persons, Bianca Hide, Brooklyn. Follow Dan Frye, Elizabeth Bourgeois, Judith Martin, Karen Beth ST George, Katherine Lynch, Kathleen's Worker, Mary Coups, Marianne McNally, Ben Rather, Liz Button and Rebecca Rutherford. Thanks, guys. If you want to hear your name, think about signing up as a patron on our patriotic page to support the podcast until next week. Folks, for science, empathy and cute