AI for Kids

How Kids can Shape AI's Future (Middle+)

• Jaiden Li • Season 1 • Episode 26

Share episode

Send us a text

Ever wondered how a teenager could shape the future of artificial intelligence? Meet Jaiden Li, a high school student and AI researcher with a unique journey from China to Singapore and then the US. Tune in to hear about her innovative vision for an AI translation tool that could break down language barriers in understanding policies and laws. Jaiden's passion for languages and math, along with her personal experiences, fuels a conversation that seamlessly blends cultural insights with AI's potential.

From addressing the psychological effects of consuming negative news to the pivotal role teens can play in AI regulation, this episode covers it all. We dive into the challenges facing young minds in today’s digital landscape, exploring how AI can be leveraged to promote mental resilience by balancing news consumption. Jaiden provides her perspective on how students can contribute to shaping fair and effective AI policies, especially in educational contexts, highlighting the ethical dilemmas posed by technologies like deepfakes.

Whether you're a budding AI enthusiast or just curious about the intersection of technology and education, this episode promises to inspire and enlighten.

Resources:

Support the show

Help us become the #1 podcast for AI for Kids.

Buy our new book "Let Kids Be Kids, Not Robots!: Embracing Childhood in an Age of AI"

Social Media & Contact:

Listen, rate, and subscribe! Stay updated with our latest episodes by subscribing to AI for Kids on your favorite podcast platform.

Like our content, subscribe or feel free to donate to our Patreon here: patreon.com/AiDigiTales...

Speaker 1:

Welcome to the AI for Kids podcast, where playtime, learning and creating collide bit by bit. Ever wonder how your phone recognizes your face. How does a game learn to get harder as you get better? This is AI. This podcast is designed for kids like you and your human parents, making the complex world of AI easy to understand and, most importantly, fun. So are you ready to unlock the mysteries of artificial intelligence? Subscribe and join us on AI for Kids. Hi everyone, welcome back to AI for Kids Today. We have an exciting guest on today. Please welcome Jaden Lee, a high school student and AI researcher. Yes, I said both of those things in the same sentence. Jaden, can you tell us a little bit about yourself and how you got interested in this wild and crazy topic called AI?

Speaker 2:

Of course. Hi everyone. My name is Jaiden, so the way I got into AI was initially because not because I liked STEM in particular. I actually loved learning languages as a kid. I was born in China and I moved to Singapore when I was five and then moved to the US when I was eight, so I had a lot of exposure to languages around the world, and something I realized was that translation tools, including Google Translate, are often not very accurate, and whenever I tried to communicate with my parents or relatives who were a lot more experienced in Chinese than I was, I found that there was often a language barrier that was very difficult to address without the help of a native speaker, which is why I became interested in AI, specifically AI for translation.

Speaker 1:

That is so cool. Okay, so I have so many questions and I know we're going to like do an icebreaker in a second. One is a good point that you brought up that a lot of the translation tools aren't great when you actually use them. Like, of course, it'll get you by so you can be able to talk about those things. But I love that you actually saw a problem and you're like oh, I'm going to learn about this thing to do more. So thank you for talking about that. I think that's so cool. I'm really excited to have you on today and learn about how you're actually utilizing that. But before we get into that, let's talk about your favorite subject. When you were a little kid, did you always love learning about languages and technology?

Speaker 2:

Um, here's the thing when I was first learning English it was really, really difficult for me. I remember in my first ever English class in kindergarten I thought that you could just like write by, like stringing random letters together. So I wrote letters in my book that was just like J-I-O-P, and then I showed it to my teacher and she was like what on earth is this? So I don't think learning English was definitely my favorite subject growing up. I think as a little kid my favorite subject was probably math, because it's very universal. It's hard to argue that oh, there's this one right way to do this and yeah, it's just like a universal language. So I guess I'd have to say math.

Speaker 1:

Which is another type of language when you actually get down to it. But that is such a good point, and kudos to you for learning English as a second language. My brand second language is also English, so sometimes when I use words in like different contexts, I have to remember that it has rules. But then it gets really weird when you think about like cultural pieces and just community. So shout out to you for spending time learning that, but also shout out to you for loving math, which is also another universal language and it's the same across all groups. I love that. If you could build an AI tool to solve a problem, what would it do and why?

Speaker 2:

Okay, if I could build an AI tool to solve a problem, I think I would make a tool that can translate policies like laws and regulations, into other languages. Into other languages because, as someone who didn't grow up speaking English, obviously moving to the United States meant that every day we had these like small inconveniences of just like not being able to understand certain rules and like being punished as a result, not because we didn't want to follow them, but just because we didn't understand them right.

Speaker 2:

So AI translation is getting like obviously it's getting better and better, but there are very few tools that are fine-tuned for like policy translation and that are actually trained on policies in the target language Not just the actual contents of the policy, but also like the formatting, because, like, different countries have different ways of putting things into law and whatnot, and the structure of these laws are very different. So, yeah, that's probably what I'd build.

Speaker 1:

I didn't even think about that as a problem, so thank you for calling it out. I work my day job is working with governments, so when you talked about that I was like that is such a good point, because a lot of policies are put out, we're asking people to follow these things, but if you don't have a easy way or a translation that helps you to fully understand that and you mess up. You said not because you're trying to mess up, but because it's not easy to understand and some of the words aren't easily translated into other languages, because some of our words people just use the English word of it, but if you don't understand what that means, like it's very hard to follow. That.

Speaker 1:

I love that idea, particularly because of some of the work that I'm doing. We're going to jump into something a little bit more interesting and I think we're going to all learn a lot here. So you wrote a paper using a big word, sentiment analysis, to recommend positive news articles. Before we even go into what you wrote a paper on, can you talk about what sentiment analysis is and why it's important in a way that a kid can understand?

Speaker 2:

Yeah, sure. So sentiment analysis is a specific natural language processing algorithm that essentially assigns a label to like a body of text and it's particularly helpful to determine, like whether a statement is positive, negative or neutral. Those are the three most common labels you'll find in sentiment analysis. At the more complex level you can assign a specific numerical value to the sentiment of each piece of text. So if something is only slightly negative then it'll have a smaller negative value than something that's very negative. But essentially the job of sentiment analysis is at the basic level is just to determine whether the contents of a text are positive, neutral or negative.

Speaker 1:

Interesting. So it allows me, like you said, if I take a sentence and let's say the beginning of that sentence is very positive, but at the end of that sentence I say something that's super negative.

Speaker 2:

Right. It will tag both. No, say what you're saying is like in the context of a review. If that very negative thing outshadows the positive thing, the AI will be more likely to market as negative. And this goes into some interesting cases like you sort of touched on with, like things like sarcasm and irony, because oftentimes it's very hard for AI to pick up on those things. So, yeah, there's definitely a lot of nuance there on those things.

Speaker 1:

So, yeah, there's definitely a lot of nuance there. That's also a good point around things like sarcasm or some of the cultural context where, if you're in it, you know someone's being sarcastic, but if you're an AI that's reading it, you may not know that someone saying something that does sound negative but it's actually sarcasm and you may tag it the wrong way. So that's interesting to think about. How did you come up with the idea to recommend positive news articles and what was the most interesting thing you discovered when you were doing your research?

Speaker 2:

Sure. So in our history class this year, I learned a lot about political polarization. What does polarization mean? Politics, in the US especially, is getting more polarized, which means that essentially on both sides it's getting more extreme, so like people's beliefs and ideas and whatnot and so in this sort of climate, there's a lot of news articles coming out on both sides that fall victim to something that's in psychology called the sensationalism bias, which is that, like, people are more likely to read news that is very positive or very negative, and they tend to ignore the fact that objective, neutral articles exist, and that's something I found in my research as well.

Speaker 2:

The amount of negative and positive news articles in the US they're very similar.

Speaker 2:

The difference is pretty negligible, but the amount of articles that score neutrally on a sentiment analysis model is almost a third as much, which is definitely not good. But what happens is that reading all these negative news every day just takes a toll on teenagers' mental health, and a lot of research has shown that consuming negative news is really bad for a lot of reasons, and it can cause you to isolate yourself, it can cause you to become trapped in this sort of ideological bubble, and it's just not good for your mental health overall. But, on the bright side, reading positive news can help you like strengthen your mental, like fortitude against negative news, and it can help you actually be motivated to action, because when you're constantly reading all this negative news every day, it's very hard to feel as though you have any sense of agency in it. So that was the main reason why I wanted to make a tool that could recommend positive news articles to people, because that can help get you to see another perspective, other than just what you see in the headlines every day.

Speaker 1:

I'm glad you mentioned that because, particularly with your generation, you all have so much access to so much information and so much news that, like you said, can be in the extremes across whatever that spectrum or those different politics are, and one of the things I've always been concerned about is that I know when I was growing up, I didn't have access to that much information and that much like negativity and also a lot of positive things come out of it as well but didn't have that many inputs going into my eyes and my brain.

Speaker 1:

So that's something that I'm glad you brought up, because that does impact kids' mental health and we're seeing, more than ever, a lot of mental health issues jumping up in populations earlier than we have in the past. So I'm glad you brought that up and that you're doing work to help us think about how we ensure that we're also getting those neutral articles and not just the negative ones, and also balancing that with making sure we get positive information as well. I know you're doing a lot of work in that space, but you're also helping your school district draft AI policy. What is AI policy?

Speaker 2:

So essentially, ai policy broadly speaking, just regulating AI and deciding when it's ethical to use AI. So earlier this year actually very recently the Defiance Act was passed, which was in response to the large amount of deepfakes. Essentially, just, you can make AI create false images of someone and make them appear highly realistic, and this has a lot of problems for people, especially women in the context of cyberbullying and violence, and also for prominent political figures, because it's very easy to create a false image of someone doing something and it'll reflect very poorly on them when it's not true. The Defiance Act, essentially, is helping to combat deepfakes. So that's sort of what people are sort of doing at a federal level, but what I do at the local level is essentially help decide what the proper usage of AI is in the context of education. So like, if you guys are using ChajiBT in schools to write your essays, for example, to what extent should that be allowed? Is it helpful as a tool to provide feedback? Is it ethical for ChachiBT to write your entire essay? You know that sort of stuff.

Speaker 1:

And you used two words I want you to break down. One was regulation and one was ethical. Can you explain what both of those mean?

Speaker 2:

Ethics in the context of AI is helping to make sure that the parties that are responsible or guilty for the consequences of their actions are appropriately dealt with and get the appropriate consequences, but also people that are not causing any harm are deemed innocent.

Speaker 2:

So one example of this would be like in school, using AI to write your essay. If you have a clear and defined AI policy, say like you need to score above 90% on this ChaiGBT detection tool, then that helps it, so that the people who are not using ChaiGBT and end up with a false positive are not held responsible for their actions and are not punished as a result, whereas the people that are very clearly using AI are held responsible for that. I think everyone on a large scale agrees that AI should be regulated. It's just a matter of how much, like in the example I just provided, should the threshold be 80%, 70%, 40%? I've had a lot of friends be falsely accused of using AI, and so I feel like we definitely all have to deal with the fallout and it's really important to have rules in place for that before it gets out of control. Yeah, I definitely agree, and it's really important to have rules in place for that before it gets out of control.

Speaker 1:

Yeah, I definitely agree, and regulation in this term just means rules, and it makes it super easy to understand that. So thank you for breaking that down. And your other point was like the whole idea, like the false positives, and you got into it. But that means your friends are accused of using AI when they didn't use it. So the positive part of that is that the tool says you used it. The false part is you didn't use it. So I love that you break that down and explain that is actually happening right now to kids. So why do you think it's important for students to be involved in these decisions when it comes to AI policy?

Speaker 2:

A lot of it is because our generation is the first generation to really grow up having all of these tools you know at our disposal.

Speaker 2:

Like you sort of mentioned with the news, teenagers have like unprecedented access to technology, and so I think if we really set the sort of precedent as a generation and if we don't take steps to ensure that students have a voice in deciding these regulations now, then it's going to be a lot more difficult to do that in the future if we don't set, like the proper precedent for it.

Speaker 2:

And also because students are a lot more tech savvy than like adults tend to give people credit for, and so students are actually very well equipped to notice sort of gaps or discrepancies in detection of these problems, and they can help bring it up and address it at local, state or even federal level. Because we're using these tools all the time, I think kids have a very unique perspective on how they can be used for good as well as bad, and so if a student is able to identify, say, like a loophole that an adult can't, because they have a lot more exposure to these tools, then I think it's a net positive for society that they're able to do this, if they're able to bring this loophole up to their officials and anyone who can help regulate it, if they themselves are not only the ones being affected by actively influencing change.

Speaker 1:

I agree with you 1000%. I think a lot of times we're trying to block students or kids from using it. They're going to figure it out, just like when we were kids, when there was something new came out, we figured out and we knew it quicker. And, like you said, y'all have grown up in this space. So I agree that kids should be, and students should be, at the table helping to decide on some of these things, because we also don't understand as adults how kids can find the loopholes or find other ways, because one thing about kids they're creative, they're innovative and they're going to find a way. So it's best for us just to work together to learn about that, because I think otherwise, like you said, if we're not putting kids at the forefront and students at the forefront now and we're developing policies without them, we have no idea what they're dealing with, what they know, and we need to bring that. And you got to build technologies with the users and not just on your own thinking. You know more when the users are in it on a regular basis.

Speaker 1:

So I agree with you 1000%. So are you ready to play a game? Yes, I love games, awesome. So we have a game called Tech Trivia. I'll ask you some fun questions about AI and technology, and you have to try to answer as quickly as you can. Which company developed the AI system? Watson?

Speaker 2:

So I'm going to say IBM.

Speaker 1:

Yes, that is exactly correct. So next question what is the name of the humanoid robot that has been granted citizenship by Saudi Arabia?

Speaker 2:

Oh my gosh, I know exactly what the robot looks like. I remember seeing this in the news, but I don't think I remember the name.

Speaker 1:

I'm so sorry it starts with an S. I'm going to give you a hint.

Speaker 2:

Cyborg or something, but with an S.

Speaker 1:

That would actually have been a cool name, but it's sophia oh, okay oh yeah, you remember. Now and then the last question what game did the ai name? Deep blue win against a world champion in 1997? Chess, exactly right, well done, so let's jump in. How can younger students get involved in AI? Do you have any tips to share for them?

Speaker 2:

Most people, how they get started in AI is through a non-technical capacity, just like playing around with AI tools and like prompt engineering, which is essentially like if you're using chat GPT, for example. You can try a variety of different prompts all asking for essentially the same thing and see which prompt works better. So first getting involved in AI through a non-technical capacity is always a good idea. I like to have philosophical conversations with chatbots. I always find that fun. After you do that, you're going to find more things that you can look into on your own that are interesting and you can get started with coding your own AI projects. So there's a few very basic machine learning projects that I think are really fun for everyone to do.

Speaker 2:

There's one particular sentiment analysis exercise that I think is very accessible. It's essentially building a classifier for the Reuters network and there's a very interactive guide online that you can find that goes along with that exercise. But essentially you can look online for machine learning related projects to do and there's always going to be some really helpful guides and step-by-step tutorials that you can find. I would say. The other really big thing is bridging your own interests outside of AI. With AI, I think generation is really big on interdisciplinary research.

Speaker 1:

And what does that word mean?

Speaker 2:

interdisciplinary. It means like between disciplines, so like if you want to do something involving like AI and English or AI and biology, there are a lot of projects that you know intersect both of those fields, and often your expertise of one field outside of AI can really help with your knowledge of another Like, for example, if you know another language, you can help train AI to be better at translation, and there are a lot of things that you can do that don't solely lie within the realm of AI. I would definitely be on the lookout for any of those opportunities and then, once you get to the high school level, there are a lot of summer programs and boot camps and classes that you can attend. Look around your local area and see if any of those opportunities are available. I know the AI for All summer program is a really big one around our area and it's around the country, so you could definitely look into that if you wanted to. There's a lot of cool things you can do if you just take the time to look around for the stuff.

Speaker 1:

Are there any tools that are your favorite?

Speaker 2:

Something I really like playing around with a lot is the TensorFlow Word Embedding Projector.

Speaker 1:

And what does that do? Like playing around?

Speaker 2:

with a lot is the TensorFlow word embedding projector. And what does that do? So the embedding projector is a way to visualize certain words. Do you mind if I present this? Actually, I would be open to it. Oh, go for it. So this is the TensorFlow word embedding projector. What word embeddings are is essentially just a way to represent words as numbers, because that's a lot easier for computers to work with. If you play around with this tool, you can see I'm on this word right now survey and it looks like the closest words to survey, according to this, are research, geological statistics, report and census. You might expect research, but you probably wouldn't expect geological next to survey?

Speaker 1:

I wouldn't.

Speaker 2:

The reason for that is probably because the AI has been trained on a lot of data that uses these two words next to each other, so like geological survey. If a lot of texts have that specific phrase, then that explains why geological is so close to survey. Playing around with this tool is always fun to see what AI learning actually sort of looks like, and there's a lot of other cool tools as well.

Speaker 1:

Oh, that is so neat.

Speaker 2:

Yeah, this word embedding projector is always fun.

Speaker 1:

And then you can like zoom in to the different words and different clusters. Is that what you're able to do there?

Speaker 2:

Yeah, exactly, that is so cool.

Speaker 1:

Thank you for sharing that. I'm glad you were able to share your screen and let folks see that. That'd be great. Considering some of the things that we just talked about around AI. I know that gender bias in AI is a big topic. Can you explain what gender bias is and why it's important to address?

Speaker 2:

Since AI is often trained on biased texts, it's very easy for that bias to be replicated and then spread out. It's really important for us to eliminate those biases in AI so that AI doesn't end up bringing more bias as a result of being trained on biased texts. So there was a really famous research paper that was published a few years ago. It was titled man minus woman equals computer programmer minus homemaker. Oh wow, Remember the word embedding projector I just showed?

Speaker 2:

So word embeddings have been proven to help search engine results and it can help show like more relevant results for your search. But one thing that people found when using word embeddings for search engines is they realized that it was leading to some biased results. If you looked up computer scientists on like a college directory, for example, it would list any man's name above female names. And the reason why it was doing that, they found, was because the male names were closer to computer programmer as in like the word computer programmer in that visualization space we just saw than female names, which was a really big surprise. And it just goes to show that all of these biases that have occurred before they all compound and result in more biased AI tools. It's really important to make sure that these biases aren't going to be spread by AI, because it's becoming so widespread and there's not a lot of organizations or tools that really address this at a large scale.

Speaker 1:

That is such a good point. When you use the word bias earlier, just in case someone doesn't understand that what does that word generally mean?

Speaker 2:

Bias in general is prejudice or discrimination against a certain group of people, Like in that search engine example. You could say that the search engine is biased in favor of male computer programmers and against female computer programmers.

Speaker 1:

And that's a good point, and often, when people are putting in data which our listeners will have learned about, the data may be more focused in one area, which can even include biases that they weren't even trying to do, because they just upload one type of data and they're not getting data from different systems. So thank you for calling that out, because I think it's important for people who are interested in this topic to make sure they get that. Can you share some fun activities or resources that you think kids can do or use to learn more about AI and technology in general?

Speaker 2:

As for coding in general, I think Scratch is a really fun tool. It's how I got started with programming and it's how a lot of people usually get started with programming. Scotch is like a block programming language and it's really good for just learning the basics of object-oriented programming and you'll find that it's actually, structurally, it's really not any different from writing regular code. As for like articles and reading, if you go on the platform Medium, it's like a website where people write blogs essentially related to topics in AI and machine learning A lot of those. While they're geared at like a more adult audience, they're very accessible for middle school students and like high school students to read and, honestly, it's not really complicated. Once you read one, it'll be really simple to figure out Like other articles related to the same topic. It'll be really simple to figure out like other articles related to the same topic. Medium is a really good resource to find new project ideas and also to read up on like topics relating to AI.

Speaker 1:

We're going to do one more fun segment. Are you ready? Yes, jaden, we want to learn more about you in general. So we do a game called Two Truths in a Dream. You'll tell a two true facts about your life and one dream that you had since you were a little kid, or even now, and we'll try to guess which one is the dream.

Speaker 2:

I want to be an author is the first one, I know how to speak French is the second one, and I play the piano.

Speaker 1:

Okay.

Speaker 1:

So the three are you want it to be a author. The second one is you speak, or you're going to speak, French. Third is that you play piano. So I'm going to start with the first one, that you want to be an author. I think that that could be true, because you've mentioned before and kids make sure you're thinking about conversation, using context clues and also trying to figure out which one are the two facts and which one is the dream. I think you would want to be an author because you are already doing research and already like reading and writing. So I want to put hold that one for now. So I think that could be true.

Speaker 1:

The second one is around you wanting to speak French, or you already are speaking French. That one, I think, may be a dream. I think you would pick a different language. If you were going to choose a new language, I would assume you would pick something else that was more quickly developing. I could be wrong, so we'll see. And then last, you play piano. I think that could be true, so I'm going to gonna go with, even though you didn't say anything about music. So that may be wrong. I'm probably wrong, but I'm gonna guess that french is the dream I do speak french, I took french when did you learn french?

Speaker 1:

I have so many questions.

Speaker 2:

I started learning french in middle school, and so I've been learning it for about five years now, I believe.

Speaker 1:

I didn't think about that. I should have thought about like in middle school. I took French as well. Oh, that's right, I didn't think about middle school language. I was thinking you were older, so I was like she's not about to learn French as a new thing. You got me Very well done. Which one is the dream?

Speaker 2:

The dream is being an author actually. I've always wanted to publish a fantasy novel that's based on language and magic. Types of magic you're able to perform are determined by the languages that you speak, so, like Logomancy, that's a dream of mine.

Speaker 1:

For a while now. I love that. That was fun. I really appreciate that. Before we go, do you have any advice for kids who want to learn more about AI?

Speaker 2:

My biggest piece of advice would be to be curious and ask questions. So if there's ever something that you don't know, oh, that reminds me of a really good resource I forgot to mention earlier. There's a YouTube channel called 3Blue1Brown. This YouTuber has really good videos on the basics of machine learning that, honestly, anyone can understand, regardless of their age. So definitely look him up. His name is 3Blue1Brown. I will make sure that is in the show notes. If you're ever curious about how anything works, or if you're ever confused, don't just sweep those things under the rug. Make sure you actually look for answers to those questions and understand conceptually how things are going on, which you can only do if you're curious about things.

Speaker 1:

Yeah, I'm glad you said that. I feel like every guest always puts curiosity somewhere in there, so I really appreciate that. Is there anything else you want to share with our listeners before we wrap up, if you?

Speaker 2:

guys are passionate about gender bias and AI, or you want to learn more about what that entails, you should definitely check out Girls for Algorithmic Justice. They're a nonprofit organization that's centered around addressing gender inequality in AI, and they're just really cool, so you should definitely check that out.

Speaker 1:

Such an amazing resource. Thank you so much for joining us today, jaden, and thank you to all the listeners for tuning in. Don't forget to subscribe to AI for Kids and stay curious. Thank you for joining us as we explore the fascinating world of artificial intelligence. Don't keep this adventure to yourself. Download it, share it with your friends and let everyone else in on the fun. Subscribe wherever you get your podcasts or on YouTube. See you next time on AI for Kids.

People on this episode

Head Shot

Amber Ivey (AI)

Host