 What we're going to do is talk about AI technologies that you can actually use. Yes? Good? Not DNA. We're switching. We're not talking about healthcare anymore. This is actually closer to what I love to do, is to make stuff. Okay, so here we go. We're going to make stuff. First thing we're going to do, we're going to go to QuickDraw. QuickDraw.withgoogle.com. QuickDraw.withgoogle.com. I'm going to show it to you. And then we'll play together, okay? Yes. Talk. This is, we can talk together. QuickDraw.withgoogle.com. It should look like this. Any questions? Can you guys get to it? Give me a thumbs up or something so I know you got it. Alright, great. Okay, so what we're going to do, we're going to click the yellow rectangle. Have you guys done this before? Have you seen it? Some of you. Great. Okay. If you haven't played before, you will not be sad to play again. It's fun. If you like Pictionary. Okay, we're going to draw six pictures and then we're going to wait on the screen together. I'm going to draw a peanut apparently. You draw your own pictures, okay? Use your site. Oh, I have no clue what you're drawing. Okay, you should be on this well drawn screen. You guys get there. Thumbs up for me when you have drawn. Thank you. I have two, three. Thumbs up. Thank you. Okay, so here we are. We have drawn six pictures. We gave Google some information. We did give Google some information here. What did we give Google? Yeah, our own Googles. How do we draw? I drew this one in particular. Who do you know what that is? You can tell it's a bird? Yeah, okay. So Google, when it gathers this data, you can click on it and see if you got a word that was wrong. And you can see, I thought I was drawing a mouse, maybe a seed boat or the Great Wall of China. Whoa. I thought my drawings looked more like those things than my bird. There's not a human mind happening here, right? This is a different thing. But I'll show you why it didn't recognize that. If you're on this well drawn screen, there's a word that says data that you can click on right in the middle. So if you can find that data word, it shows you the 50 million drawings that people have given it for free. And you can see all of them. So if you click bat or bathtub or bear, I personally am going to choose butterfly. It's one of my favorites. What is it at now? 114,000 butterfly drawings that people have given to Google for free so that they can learn what a butterfly looks like when someone draws it in 20 seconds or less. The important thing here though is, do you see that when I mouse over one and you can pick something else if you like something else? It's actually showing the way that people draw it. So even though my bird ended up looking like a bird that you guys didn't have at most people draw it, that's why Google couldn't recognize it in the end. Because this data set is not about the final picture. It's about the initial shapes that people use in order to make a drawing. So this butterfly, people are either going to start with a long skinny oval. People usually point it up when they say almost all point up. And some oddly shaped B letter and reflection sizes. So you can see that they either start with the long oval or they start with that backward curved thing. And mostly from people who grow up writing English, we draw from the left because that's where we start our letters in the top left corner. So that pattern is what makes a butterfly to Google, not the final image. You understand the data is actually the way that you draw it. So we have given Google all this data and now it knows what human sketches look like. And then, let's go back to the slide. Not that one. We can close that one. Here we go. This one. We're going to go to autodraw.autodraw.com. You get to go there too, autodraw.com. This is a product that Google developed from your free data that you gave away to it. That is what happens with AI. We give it data and it learns. So I'm going to type in autodraw. Here I have a blank piece of paper and the tool that I want to use is this black magic pen here. And I'm going to draw a butterfly. I think I'm drawing an ear. And then by the time I get to the two big shapes, I have a list of items at the top. And now I click and I have a clip art maker. So Google turned this data set into a product. And I can click on this shape and I can resize it. I can make it bigger. I can change its color. And all of a sudden I am an artist. I can totally draw. There's a little bit of a problem with this data set though. Let me give my magic pen again. As far as I know, this is my sad attempt at a horse. Sorry, horse people. It looks like I've got some dinosaurs up here. There's a camel. There's a horse. So even though I did a really bad drawing, it knew that other people draw equally bad horses. But there's no unicorn. Which is really interesting to me that there is no unicorn. However, let me get my magic pen back. There is a dragon. Three dragons in fact. So I think Autodraw is a really great example of a data set that is perhaps designed or developed more for the male typical presenting gender. There are more options for things that boys like in this database, typically, than girls. And there's no reason for this to be true except that the people who created this program didn't think about unicorns, but they did think about dragons. But Autodraw is learning. So there's a hamburger menu on the left and you can see, there's some options here. You can read about the program, which is nice how it works. But you can also see the artists. And the nice thing is that you have power and agency to choose, let us know, and to tell them what you want. So if you are a girl that would like to see more girl products in this program, tell them what you want. If you are an artist and you want to submit your own drawings, you too can participate in the development of Autodraw. So I like that because it's not a closed system. It's not like you're using some art program that you don't have options. Autodraw is ready for you to help with it. Okay, anybody draw anything cool with Autodraw? Autodraw.com? Did you guys make anything? Did you draw? Okay, well at least you could see it, I guess. Okay, let's go back up here. Autodraw. Yeah. Okay, Pinscreen.com, if you have an Apple phone, this is more fun for you. You can go to the Apple store and you take a selfie, a single selfie of your face. And it makes a 3D avatar. Sorry, my friend. Oh, look for the Google logo if there. Pinscreen.com, we're going to go over there. Pinscreen.com. This is an avatar made from a single selfie. And so you can use these style them however you want and you can watch a video of it. Which I think is pretty cool. So this technology, one thing that's really cool about it is that UNC Chapel Hill did some of this first work in the computer science department. They were creating avatars and they do full body stands from something that's either a tablet or a laptop. They're using very low power and they capture your full image. So like I would want to be able to find so that when you move into virtual spaces, you have an avatar that's a lot like you. Sorry. Okay, you get the idea? That's Justin Bieber there. You guys use Snapchat, just where we put a filter on it. It was like a huge breakthrough in AI. And so this is just sort of an extension of that. Just being able to transform your face. Alright, got the idea. Let's do something else. Crowdsource.google.com. Crowdsource.google.com. We're going to spend a few minutes here. One thing that AI does really, really well is label an image search for... Hey, sorry Zoom people. Sorry Zoom people. So here we are at crowdsource.google.com. And we are going to think about how all those labels about corn. It's corn. How do we get a six-year-old into the mix and not just some corn, right? So let's go to crowdsource.google.com. Oh, I feel bad about my Zoom friends. Okay, crowdsource.google.com. You'll see this is a website that Google uses in order to learn how to label images. So we're going to choose image label verification. And I used to use swimming all the time. But guess what? Google got all the information it needed from swimming. So we don't need to do that category anymore. Yeah. Hello. Oh dear. Sorry Zoom people. We're 14 times. Okay. I didn't turn it off. It was not me. We're okay. Okay. Image label verification. We are going to choose... Oh, let's see. Let's do cats. You guys want to do cats? Everybody loves cats. Oh, cats that are done. That's because everybody loves cats. They should take those off. Let's try another one. Okay. Image verification. Okay. Let's do chairs. Certainly no one did chair. Is this a chair? My friends is a hot dog. So I'm going to choose no. You can play on your own. You're going to get different images. You don't have to use mine. Okay. Does this image contain a chair? Is that a chair or a couch? Oh. But I have to say yes or no. Are we already struggling to define whether or not this image has a chair in it? Okay. We can't even decide if that thing in the background is a chair, friends. So this is why computers... Sometimes you get really mad at the results, but we actually have the same conversation. I'm going to say no. I agree with you. This is not a chair. Is this a chair? That is... I think that's a church. Church not a chair. You guys say no? Okay. This is a hobbit hole. But does it contain a chair? A bench. It's a bench and not a chair. Okay. Good. Does this contain a chair? That is a... That's a table. Thank you. Yes, that is a table. Does this image contain a chair? Yes. Why is it a chair? She's a sitting on it, but what makes it a chair? She's sitting. It's for one person that contains a chair. Okay. This appears to be a television and an entertainment center. I see no chair. Do we agree? So we are teaching Google about the differences in some of these images. Okay. So now these people are sitting. Is there a chair? No. So sitting is into chair. So we are teaching the Google very specifically what a chair is just by giving images. So if people are sitting, but there's no other thing, there's no chair. Right? If the thing is for sitting, but it's too long, it's not a chair. The thing can be made of wood. The thing can be made of fabric. Oh, what's this? Another kind of chair. Made of plastic. Is it a chair? It's a very iconic chair, right? Everybody has this $9 version in their backyard. Their grandma's backyard. Yes. It is a chair. Yeah. Yeah. So this is very specific. Does it contain a chair? So in this case, it's trying to figure out if there's any sort of chair in the image. With swimming, swimming used to be one of the ones that I did. And the first picture that would come up would be a fish. It's the fish swimming. Some people would say yes and some people would say no. And then we get to water polo, which is a sport with a ball and you throw it. It's like soccer sort of, but in the water. And polo involves swimming, but were they swimming? Not necessarily, right? Throwing, defending. And so some of those nuances is what the data is learning. What happens if we all go in and just click yes for every photo? Yeah, Google gets stupider, right? Yeah, for sure. So this type of information where we give labels is really helpful, but it can be used nefariously. Do you guys think it's fun? You are actually participating in helping the Google image recognition software get better when you use CrowdSource. How are we doing on time? Pretty good. Okay, let's make some more stuff. You can always go back and play with CrowdSource. Okay, W3 schools. You guys have learned a lot about technology today and you haven't necessarily learned how to make that technology. This is one of the best websites as a resource for online learning about computer science, very broadly. But so I want to show you a few things about it so that if you want to be more of a maker, you can be. This is a place where there's a ton, a ton, a ton of references of how to code, how to learn any type of code, but they specifically have a section right here under tutorials, learn AI, learn machine learning, learn data science. It assumes at the start that you know nothing. You don't need to know a single thing to start here. As a former developer, 10 years, I worked in computer science in building databases and web development, primarily business-to-business systems, the back-end side of things. And although I started out with some college knowledge, what was way more important was actually that I built my own stuff. And so if you are figuring out whether or not you can afford college, if you have a strong portfolio and you teach yourself stuff and you are good, you can find that entry-level job for real. So you can teach yourself, and I want to make sure you know that this is a place to do it. The important thing is that you build things that work so that you have a portfolio to display. And so in their learn AI section, I just want you to know they have the first example, image classification. So this is not the Google version that we just did. This is the code. This is how it works. This is the math that you've learned, and I wonder what it was for. This is what it's for. And so you can go through and learn all about how to create these various things right here on W3Schools. It's completely free. Questions? Anybody actually going to try this? Check it out? Yes, good. Okay. One of my favorite things here too is just like if you just want to mess around with a website, you can. There's really good stuff in here. There's guides and forms, all kinds of things. And one of the things I use all the time is that there's this HTML colors, and you can actually see what all the colors are named and how they're picked, and all the different types of names. So if you have your very favorite color, I have a friend who's really into pink, and all of her stuff is like branded pink. And so she has her favorite hex code color. And so this is just a great place as a reference if you're working on other things and you need some specific information. They have all these great tutorials. You don't even have to sign in. You can just use it. Okay. Write it down. W3Schools. Remember, if you forget, it's www, right? How many Ws? There's three of them. W3. And then you'll learn. If you want to be a coder, if you are interested in computer science in particular, you will have to learn forever. It never stops. You'll always have to be learning. The stuff that I coded doesn't even exist anymore. The languages that I use, they aren't around. So you'll always be learning. Okay, Vocaloid. Have you seen a Vocaloid? Do you watch them? You know, yes, a little bit. Some of you know what I'm talking about. These are like anime characters that sing and dance. And you can go see them for real on a stage. Please tell me you know what I'm talking about. Yes? No. Okay. I do feel like we should at least maybe visit the YouTube and see a Vocaloid performance so that you know that I'm not crazy. The O-C-A-L, there it is, Vocaloid Los Angeles. Hatsune Miku, anybody? Anime fans. This is a real concert. There's 70,000 people here. You can see them. Oops. The real musicians and the performer is an anime character. Anybody in band? Band kids here? Few. Okay. So musically, there's a revolution happening. We're never going to get rid of real live in-person performance. But singers and musicians are moving to the digital space. And some of our biggest, this is a very popular character around the world, particularly in Japan. Some of our biggest characters are not people. Okay. So if you want to be a person making art and music from people that aren't real, Vocaloid.com is where we're going. Vocaloid. This is a singing synthesizer. If you really are into music, digital music, if you want to create music without singing it yourself, you should download this software to a real computer. It's really how it works best. But I at least want to show you there is a free trial version right here in the corner. You can get it right there. And we can preview some of these songs right here on the main page. Now these are AI singers. They are not real people. They make music. I don't know if my people far away can hear them. Okay. And what's interesting about them is that they are not just singing. But you can adjust every bit of what they sing. And by that I mean, we'll watch this a second. The way the software works. There's a person at home on your home studio. So that line is the graph of the singer. But it's not a real person. So just like in any other software program, you can grab the line and move it and change what they sing. So you can see he's changing the words. He's typing in the words and then he's moving the pitch. So you get the idea. When you use Vocaloid, you can start with a voice that isn't your own or a completely digital voice. You can use your own voice and modify it. You can change what is said, how it's said, the pitch, the timber, all the good things there. And if you want to play with this, download free right there. Cool, right? You can manipulate accents, vibrato, rhythmic feel, and change your own vocal production. There's another program like this. It's called Descript. I find it fascinating. We can't do as much with it. But if you're a podcaster or video editor, this to me is the coolest, coolest thing. So what happens is I go in and make a voice print of myself. That means I read a script so that this AI knows what my voice sounds like. And then I can... There it is. You can see. She talked. It recorded into like a Google Doc type thing. And now someone else that isn't me can go in and edit what I said in the text version and it updates my video. So I could record something okay where I say a bunch of pauses and things. And then I can hand that off to my friend right here and they can edit out all of that. So that I can have a completely clean podcast or video. And if I've done a complete voice print, you could even take my voice and write a script for me and I could say it without ever having said anything out loud. Say that. Yeah, yeah, right? You have to trust somebody with your voice print. Totally, totally. You have to be careful with that. Okay, you can generate singing voices. How about AIartists.org? AIartists.org. This has over 40 tools. If you are into AI or if you're into art and you're not sure about AI, this is a fun place to play. AIartists, I-S-T-S has the S on the end.org. I spelled it wrong, didn't I? A-R-T-I-S-T there. AIartists.org, largest community of artists using AI. There's a little bit of a debate right now about who owns AIart. If you're not the creator of the original art and you modify it using an algorithm, is it the original artist who contributed? Is it the person who created the algorithm or is it you who came up with the unique genius of merging various things? I think what will end up happening is we'll end up with some sort of policy. You know how a lot of cool hip-hop music is resampled and remixed? We're probably going to end up with some sort of visual legislation that gives credit to all of the contributors and financial credit too. Those systems aren't yet in place though. If you're a person taking paintings from the 16th century and having an algorithm merge them with future scapes, there's not really any way to tell who should be compensated and who gets credit for that. This community, AIartists.org, if you scroll down, you can see cool, really nice art and what people are doing and some history of it. And then also there are resources which are creative AI tools and generative art guide, but this creative AI tools is the cool place to be. So we go to create, look, 41 tools. 41 tools. So lots and lots of AI that you can make. And they're bisections. So music, art would be like visual art. There's also movement and dance, sketches with Bill Jones, really cool stuff. PoseNet is much like a Kinect if you've ever seen a Microsoft Kinect where your body turns into like points and you can interact in an important way. But we're going to try Deep Dream Generator right here under AI-generated pictures. There's also some neat GANs over here, but we're going to try Deep Dream to start. We use a lot of the Google ones because they're really accessible. A lot of the other ones require sign-in. So Deep Dream Generator is a human AI collaboration and there is a text-to-dream tool and you can just try it, get started. Don't put up a wall for me. Well, you used to be able to do it without signing in. Can I please text-to-dream? You can see these are artworks that it has made, it's collaborated. You didn't used to have to sign in. Yeah, free, you need to sign up. Okay, well, let's try A-S-D-F. Do we have to make a password? Let's see. If you feel like signing in, you can. I have a Google account. If you don't, you can watch here. Okay, so the way this works is that we're going to generate an image at the top and what you do is write what you want it to create. So I can use some words, give me some words, something cool, something you'd like. Favorite sport? Any of you play a sport? Basketball. All right, now tell me your favorite animal. Horse. Okay, we're going to do horse basketball. Oh, we could have unicorn basketball. Can we do unicorn basketball? That's a cool idea. Unicorn basketball. Okay, we want normal quality. We don't necessarily need to enhance anything. Let's generate it. Unicorn basketball. Now, this has never been created before. We could have a really long sentence here. We could also specify some of these generators. You can be really specific. You can say like photorealistic or Salvador Dali. And if you, you can imagine something very different from 4K photorealistic versus like a melting image of Salvador Dali. It is actually creating a image here that, that if I did this again, I might get a different result. Okay, what do we get? We got unicorn basketball. It did it, right? Yeah, yeah, stable diffusion. Yeah. Yeah, all right. Did you do it right now? Very good. Any Discord users here? Unsurprising, yes. I realize my fellow adults have not always followed you onto Discord, but I have. So I want to make sure that you know also like the best. Yeah, you've done it. Yeah. I know. And so it was on beta for a while and it was totally free. So mid-journey is on Discord and it's an AI generator. There is a cheat sheet that I think is worth downloading. This is, I have to give credit. Mid-journey is an AI generator like Deep Dream, but better, super good. And this is a cheat sheet that was developed. It's at firebasestorage.googleapis.com. And it has all, like you could take a screenshot of this or something. These are the main commands on Discord to generate art within Discord. So Imagine creates the image, blah, blah, blah. Help, help, info, et cetera. But I think mid-journey is one of my favorite AI generators right now. You can also play online with Dolly and Dolly, two really powerful image generators that are sort of blending what's real and what's not real. How are we doing? Pretty good. Okay. Okay, let's talk about GPT-3 and GPT-4. Do you know about them? Anybody heard of them? Yes? Okay. Are your teachers afraid yet? Do you think that teachers should be afraid of this technology? Yeah. Okay, so GPT-3 and GPT-4 have huge amounts of data on how people write. And for small local newspapers who want to publish sports scores, those are probably already being written by AIs because of baseball game. There's been so many years of baseball games and they're all the same. It's like what happened in the top of the ninth and who got a hit and what's his RBI's. And the AIs can just write the whole game now. You don't even have to send a journalist. And so at Stanford, there's a major called computational journalism. And it's like they've almost given up on traditional journalism, although there's a huge need and importance in journalism. But computational journalism is like, you know, there's actually a story in the data. We can mine like Twitter feeds to see how people are feeling with semantic analysis. And we can write some stories using GPT-3 and GPT-4. And so this is an example of how text generators work. But GPT-4 is so powerful, it can write like where if I was an English teacher, I might be worried that my students won't write their papers anymore. No. Right. Right. It's not plagiarism. Well, it is plagiarism. But it doesn't always give the same results because it's generating it each time. So let's write a paper. I need a subject. Tell me something funny you did in the last week. Something funny. Do you have a small sibling? They did something crazy? No. Yes. What? Okay. So we're going to start with Harry Potter. Harry Potter. And we're going to take that little fragment, spilled his cereal. I guess he spilled his cereal. Harry Potter spilled his cereal. And then what did Voldemort do? Maybe. Let's be more interesting. You were a basketball player, right? Basketball? Basketball? Basketball? Okay. Tell me something that happened on the basketball court that Voldemort might have done. Okay. Voldemort scored a three-pointer. Let's try it. Let's ask the AI to write us a story about these things. Harry Potter spilled his cereal. Voldemort scored a three-pointer. Ron stared into Harry's eye. He reached for his wand and started to put it on his head. He started to wail and yell at the door in his ears. Voldemort pushed and Harry pulled up. Harry screamed. He started to run, but Harry caught him. He started running and Harry got to his feet and grabbed the floor, pushing him back. Anyway, you see. It knows who we're talking about. It's written something somewhat nonsensical. But the better algorithms, if we had picked something that is more academic, it would get closer and closer to what a human would write. So these are fun to play with. Especially if you take a prompt like, astrophysicists have discovered. What have they discovered? That dark matter like the sun is extremely close to our solar system in mass and that its red light is just as large as our sun. If I don't know anything about this subject, I might actually believe it. It's kind of fun. Yes, yes. It doesn't only have to write like in standard English. It can write in a dialect or in another language. It depends where you are, though, whether or not there's enough data. I'll give you a really good example. We did crowdsource, right? Like crowdsource.google.com. And if I put in wedding, I am probably going to get white dresses and tuxedos and matching people at a Christian altar. I'm probably going to get something that is close to an American wedding. What I am not going to get is a wedding that would be appropriate in Zambia. Right? And that would look very different. Or India, where there might be a sorry involved and really bright colors because the cultural context of the search is in America. Now, if I went to Google in India, I would probably get the right results for India. But that's a great point. It's like, if there's not enough data, then we might not get as good results. So it probably writes better in English or Mandarin Chinese than it would in a smaller dialect where there's less data available. How are we doing? Oh, we're pretty good. Okay. The last thing is that it's really important to make stuff. There's a guy, his name is Douglas Rushkoff. I really recommend his books. He wrote this title up here. He wrote it 25 years ago and he said, if you're not the makers, then you're the product. All this stuff you learned about today was about data. You're giving away your data all the time. It's almost impossible not impossible. Really, really challenging. So the way to avoid some of the worst consequences is to be a creator of some of the solutions. It doesn't mean you have to code, but it means that whatever you are, whether it's a journalist or a school teacher, that you're aware of the ethics of these technologies and that you do something about it. Okay, questions for me. We've got 10 minutes. Do we have any way to have the Zoom people ask me questions too? Because I'd love to hear from them too. Do you guys have questions for me? Yeah. So you're talking about upscaling images? Yeah, so do you want to know about the math of that a little bit? You know, okay. Yeah, so that's really cool technology that didn't used to exist. Another interesting place where AI, we don't even think about it, but it's totally AI is autofocus and the way your phone knows what to focus on and some of the cool technologies like night sight and stuff, those are all AI technologies. The photo space is like really, really good right now for AI. Totally. Questions, what can we make? There and then there, okay, go ahead. Do you mean like when I visited the doctor? Yeah, when I visited the doctor, they had recorded my son's visit and they were testing to see if that technology would work for transcription. So when you visit the doctor's office, they write down what happened in that visit. They say something like, this patient has presented with a fracture of the tibia and we recommend six weeks of cast and then two weeks of physical therapy, something like that, where they're saying what happened in diagnosis. They might also share different information like the patient presented with their mother, they appeared to be well taken care of. There might be some more, like some information that's not, it's not just about your physicality but like your circumstances. Are you in a safe place where you're likely to get better, that kind of thing. Or what additional support you might need, like they need help with Medicare in order to pay for this treatment. That might be in there too. But the idea is that that becomes a part of your medical record. So one of my questions from that recording was that it was recording while I wasn't in the room. So previously the doctor would leave the room and then type up their notes or dictate into a microphone and that would be typed up by essentially like a secretary. A medical transcriptionist is the word. And then that would become part of your medical record. But if it's a recording that happens from an AI that translates speech to text, that's the AI piece. Could the conversation that I have privately about my treatment go into that record? And it's probably not a big deal if it's about a broken arm. But it might be a very big deal if we were deciding on what cancer treatment to get or something like that and whether or not we wanted to continue treatment or whether or not we wanted to go see a different doctor for a second opinion. Is that being recorded? Is it part of my medical record now even if the doctor wasn't in the room? I don't know. I think the part that bothers me is that I don't know. Questions? Yeah. Oh, wait. You and then, yeah. Yeah, the comment is about podcast.ai, which is a website. I did not play it here because there's some swear words in it. But it's completely AI generated as a podcast and there's enough vocal print of Joe Rogan that it's pretty accurate. The only way that I could tell that it was really fake was that it mispronounced common, famous people's names. I think, was it like Bruce Willis or one of the names was wrong that they had said. And they're like, oh, nobody who's a personality would get that one wrong. Steve Jobs was on the air even though he's passed away. So that's the interview. So podcast.ai, language warning. Here and then there, yeah. So my job is chief operating officer. So that means most of the time I run a company. And sometimes I get to come and hang out with you guys, which is like the best, the best. High schoolers are the best. Totally you are. Why do you say no? No, like you guys are actually like super interesting and smart and I wish I could talk more to each of you because I always learn from you. Like anybody who played Detroit, you become human. That's an AI. You played it? Yeah. That's an AI game where you get to play as the AI. And I learned that from you guys and then I went and played it. So if there's any media that I'm missing, I want to know about it from you guys too. You could email me, text me. So my day-to-day is to help kids learn about AI. That's the job of my company. It's not always my job. But I was a teacher previously. Yeah. My generation says who it's written by. Very good. It's totally plagiarism. Yes, if you use an AI to write your paper, you didn't write your paper. Just to be clear. Yeah, if you wrote the topic sentence, you did not write the paper. I think there's a college professor. I work with the National Humanities Center. So I work with 15 different colleges who are developing ethics classes with AI. And that's one of the cool things I get to do in my job. And one of the professors in that program is doing a class where you start your topic with an AI and then you write your paper as compared to what the AI wrote. I think that's a really great way for teachers to incorporate AI into their language classes is to start with the algorithm. Because what's really exciting in the end is what you think. Yeah. Yeah, thank you. It's called I Am Woman. I Am Woman is the recommendation. That reminds me there's a book called Clara and the Sun. Clara with a K. And Clara and the Sun is a Pulitzer... It won some sort of major book award, but it's about when it's a futuristic novel, when children have robot friends as companions as they're growing up. That's really good. And then Ray Bradbury has a short story of an electric mother from the 60s, which is equally interesting about what if you could have the perfect mother. Looks like people are coming back. Thank you for spending time with me in this breakout session and having a chat. Thank you virtual people as well.