Yesterday Today Tomorrow

Guest (Female): Before we get into the show, here's something worth your time. Every day, animals across the country face abuse, neglect, and cruelty. And every day, the ASPCA shows up. They rescue animals in crisis and provide life-saving medical care. They help animals recover, and they give them the second chance they deserve. But they can't do it alone. Every rescue, every medical treatment, every safe place to heal depends on people who choose to help. If you'd like to be part of that work, visit aspca.org/need. Become a monthly donor today, and you'll help support these life-saving rescue efforts and provide the care, shelter, and love these animals urgently need.

And for a limited time, the ASPCA will send you a free gift as a special thank you. Again, that's aspca.org/need.

Guest (Male): Yesterday, connecting past. Today, with an outer view. Tomorrow, to understand future. Yesterday, today, and tomorrow is the program that covers the current contemporary social issues in the light of our history. To understand our yesterday, to live fully today and tomorrow. Yesterday, today, tomorrow with In Song Kim.

Inseong J Kim: Hello, this is In Song Kim from Yesterday, Today, and Tomorrow. I've been sharing about technology AI that is there for us, so we are for technology. And today, I want to talk about a little bit positive side of AI and why it's so useful. And uh there's like several um section that I want to uh talk about is one is uh in education. And the second will be uh in a marketing research tool. And also uh the health questions that we can uh ask. And but there is some paradox that I'm going to do a conclusion about uh we are in this uh AI a boom generations.

So uh in the education part of it, I've been talking about foundation of structural education is needed. But at the same time is locally using as a tool is incredibly powerful, too. So it's 11:00 at night. The homework is due tomorrow. And many children want to have fun, play games, spend time with their friends. And then homework is due at 11:00 at night. And as we know in negative way, AI can uh bring a negative side of education because AI can do the homework for them. But on the other side, there's a positive side is you've read the chapter twice, and you still don't get it.

And then teacher is sleeping, and uh parents are tired. But they learned this 30 years ago. The parents at 30 years ago, so they cannot really help much what's going on today's subject. So um students, you're too embarrassed to text a friends the same question fourth time. So you open an app and type, explain it again. Uh slow this time, and it does. Then you ask again and again and again until you get it. And that is a positive part of the using the AI. It never get tired of it explaining same thing over and over again until you get it.

And it's never uh sigh or never checked the clock. Uh emotionally does not bounce back to you that uh you cannot just keep learning. So I think that's the really positive side to use it focus on narrow a subject that you just come keep pounding on uh this information that we can get. So welcome to this program. And we want to something a little unusual. There is a shortage of the worry about artificial intelligent uh jobs and misinformation and deep fakes. And those concerns are real. And but today, I want to look at the other side of the ledger that where is AI generally useful in ordinary life uh today.

And I want to be a data driven about it. So what is a data driven life? And that's something it's introduced. It's used used to be a specialized on uh certain jobs or certain professions, and not many people were using the data to make a decisions. And it's a new concept, and we're getting used to it. So not a hype and studies, numbers, or limits that comes with them. And so we'll look at this three areas. First, AI is a backup tutor. Uh it can be powerful. Second, AI is for market research which turned out to uh matter for small businesses much as big brand.

And third, AI as a first step to health questions with a caution. And like uh what uh it's you what is in your food and a side effect of the medicine. And you can just go on and on go go the deep research about it. And I recently there is egg issues that recalled how many which brand was egg was uh recalled and which egg is safer. Uh those questions that we can ask those uh AI and also health questions, a simple one that easily get access to the information, but with a caution and being careful.

And then we end the puzzle of this program because more you want AI to be accurate, the more data it's needed. The more data it's needed it's needs. And then the more the privacy is on the table. And so that's the something that we cannot avoid the issue that of the time today. So today, our first uh education the patient tutor uh is a very patient tutor. Basically the new one uh as far as the teaching goes. Here's a core idea the in structured to course with a teacher syllabus and real assessment. AI works best not as a teacher, but as a backup.

The tutor tutor you go to after class. And the superpower is simple. You can just keep asking until you get it. So rather than uh I mean weak as human, you might we might uh experiment with a complete AI education without any structure, any any syllabus, nothing. People might experiment how human from the lit the baby to navigate for the informations. But on the other hand, ordinary people not super smart and genius. Ordinary people need some structures. And uh think about how learning actually goes wrong in a a student miss a one step one step, especially math or even any other subject.

If you miss the one step, the class moves on. Every lesson after that sit on a crack of the foundation. So most students never ask the follow-up questions because they don't want to look slow in front of the 30 people. And that's the reality. And so on AI tutor removes that social cost. There is no audience, and you can ask the same question five ways. You can say give me a example with a football or explain it to me. I'm a 12 you can just specify about the question the your level of what you don't know, and you have to tackle.

So now we have evidence 2025 researchers at Harvard published randomized controlled trial in the general uh journal scientific report. It involves 194 students in the introductory physics course. And what makes it striking in the comparison, this was an AI against the boring lecture. It was AI against the well designed active learning class which is already considered considered the best practice. Half of the students learned the lesson in the class. The other half learned the same materials at home with a custom uh AI tutor. Then they swapped for the next lesson. The result, the students median learning gains with AI tutor were more than double those from the active learning class.

And they got their uh less time. And students also said they felt more engaged and more motivated. And so double is a big number. And what's the catch is the most important part of the study AI tutor is not a general chatbot. The instructors build it carefully to guide the student step by step, made them do do the reasoning and didn't simply hand over the answers. Uh reviewers of the study kept landing in the same lesson. The teaching design was the active ingredient and not the road technology. So that's the key. So elsewhere the unguided use has a backfire in 2025 is study in the journal PN AS.

And nearly the 1,000 high schools math students in Turkey were given plain chat uh chat GPT access. And they did a better while they had it, but once it was taken away, they scored 17% worse than students never had it. So a a version is built in teaching safeguards largely avoided that harm. So structure is the point structure. The second study backs this up. The trial across the five secondary school in the UK with 165 students compared AI tutoring in a math and human tutors under expert supervision. One measures that simple that after student made a mistake, did they fix it?

And with this uh static pre-written hint about 65% of the students corrected their mistake with a human tutor 91%. So with an AI tutor or 93%. So in that uh setting, the AI was about as good as a human tutor as far better hint of as a far better than hint uh and hint note. The researchers know that that running the AI cost very little compared to paying people. That uh that's why this matters one on one tutoring is one of the most effective things in education, and it has always been a luxury. The families who who can afford a private tutor have had advantage of the generations.

AI could put a patient tutor uh within a reach of the child whose family can afford it. So I think that's the positive part of it. And people the a lot of students they didn't learn in the classroom depends on how it's communicated because everyone is learning a different way. Some are visual learners, some are very logical, some are uh uh very creative, artistic. So every brain is different. So when the the message or information is delivered, how perceive, accept is different. So the AI is customized questions that depends on how individual student ask the question to AI and customized way, then AI responds to the millions millions millions data to customized to that particular person.

So that is a really positive side of the AI. And but let me slow down here because I said structured and supervised several times now. What about child on their own at 11:00 at night with open-ended chatbot and nobody's watching. And that is the unsupervised uh educations. And how here we need to be honest. Unsupervised use different situation, and it may not be safe, especially for children and teenager. Uh here's why. And because the brain is still developing. The part of the brain involved in implanting in purse control judgment keep maturing into the mid 20 according to the US National Institute to Mental Health.

And learning is soft depending on struggle. Learning itself depends on struggle. So when uh a student to wrestle with the problem and uh she or she build the mental pathway that led them solve uh next one. So this mental struggle is a part of the learning process. If a tool is always removes the struggle, then students can feel like they are learned but without actually having learned. That's the difference between understanding something simply having an answer. That is the challenge in education that we have. Uh so how teaching how to use the AI is the is the part of the this um important uh tool that we have to have it rather than just put the question the teacher gave it to them and then get answer and and handed it to the teacher.

Uh that's not the right way to use the AI. Rather uh keep questioning to AI about what a student doesn't know and trying to learn and understand. And that's the part the skill that we have to teach our children. And then there's an emotional side chatbot is uh endlessly available and endlessly agreeable for the lonely or anxious teenager that can be comforting. But it can also become a substitute for friends and family teachers. For it may fell the young person what they want to hear rather than what they need to hear. This is already happening in a scale.

A 2025 common sense media survey, 1,060 American teenagers found the 72% had tried AI companion. And about the third said those conversation were as satisfying as and more satisfying than taking with the real friends. So uh especially I think the more smart smarter uh student or children or teenager uh they want to have a proper dialogue with the intellectual level of their own. And how they communicate is different. So the method is all different with the person to person. Now we're challenging with AI that they are customizing how they can communicate with AI that how they like.

And they just kind of glued to and engage into AI to have a companion because looks like feels like uh AI can understand uh the user. So how much do we actually how uh know about this and not enough? We don't know. Still we're in the experimental stage. They here's a problem most people don't think about the most of us parents include they were never taught how the brain develops. We don't know. We're not expert in neuroscience uh uh science or how brain works. Uh we are not PhD uh as far as learning about the brain.

So how technology shape attention and mood, we don't know. We don't know yet. So just kind of here is a tool that you you're going to use this without knowing how it will affect our brain, our emotion, how complicated it created this human being respond to the AI. We don't know. And the research in AI specify is very young, and Harvard study ran for a few weeks. We don't have long-term studies for what years of the AI tour tutoring does what child thinking conference relationship and genuinely under research. So as we know, we call it millennial or Gen Z, we just label them and stereotype everyone.

But definitely there is how we educate generations and outcome is different. We experienced that uh today. And there are uh differences of how our environment affect in education uh also. So what is the practical advice for parents or teacher listening to now is three things. First, to use AI inside a structure, a course teacher real test, and check whether learning and is really happening. And second is preferred tools designed for learning and one that ask a students questions back instead uh of the giving answers. Just learn to ask back and kind of um and getting answer until uh the students can fully understand the subject.

Third is a stay involved. Ask your child to explain what they learned, and if they uh can teach back and what they learned and have a conversation with what they studied and what they learned. So uh AI can be a great a backup tutor but not a babysitter. And that is the question that we have to ask because um AI is not just simple machine. It is very sophisticated made that attractive enough to young people to have conversation because human are made very intelligent being. So we get uh attraction from the intelligent information and knowledge.

And so we need some supervision of the this structural environment and supervised environment using an AI. And so second part is I want to talk about market research. It's a very interesting and fun uh to use AI. The we used to use the website to use it. But you have to gather all the information by yourself and kind of add it all together and get a conclusion. AI does it for you. So let's move to the classroom to the marketplace. Picture someone is wants to open small bakery and sell homemade bags online. And what's the first question they face?

Uh will anyone buy it? Who are the customers? Who are the competitor charges? And what are people are com complaining about? And traditionally answers those questions and cost a fortune. Classic market research means survey, focus group, research agency, big project can take a weeks and months to run into the tens of hundred thousand dollars and large companies could afford it but uh person opening a bakery mostly couldn't. And uh they win um on a gut feeling. That's what they used to do by smart a book smart. It's it's hard to get it but street smart who run the business, the genius people who are the best on sales, they had it in their brain.

But the commoners like uh most of people want the small business, we don't have that brain but AI can help us. How does AI change that in a few ways? Simplest is research assistant. You can ask AI to gather what's publicly known about the market, the main competitors, a typical price, recent trend, common complaints in the public reviews, work that might take a person several days in searching that can come together in an afternoon. And then you check it and always and uh will come back to that. So those are a really great tool. Second use of the analysis.

Imagine imagine you have the 500 customer comments and survey answers and reading and sort them by hand is so slow. AI can group them into the theme like a price, delivery time, packaging, and show you which complaints come up the most and that used to be a specialist to job. And it's just summarized for you. And the professionals are already using it and one industry reports found that 77% of the marketing professionals say that they use chat GPT or smaller AI tools for their work. That's no longer early adoption. It's the norm. It becoming a norm.

Then there is a newest the most debated use is a synthetic respondents. The code synthetic respondents uncode. It's a sound like a science fiction but ideas that uh AI trained a large amount of a real survey data and simulate how groups of the customers are likely to answer a questions. Instead of the waiting weeks in panels of the real people, you get early uh you get an early read in a minutes. And does it work? And the vendors say it works surprisingly surprisingly well. One company uh SIM surveys reports that across the nine validation studied in the systematic answers matched real panel data with 80 to 90% accuracy at the question level at a roughly tens of the cost of the traditional panel.

But the phrase uh that we have to understand is how we use it without the discriminations. So um so the please hear that as a company own claim the promising is not independent proof. The independent test are more sobering. Uh in September 26 Pew Research Center had a two leading AI model stand for six 7,700 real survey respondents. Each model painted a very different picture of the American public and neither matched what the real people said. So we are not there yet but we're getting there. And the the second danger is uh it may be a bigger one uh the money.

When you ask an AI about the market, it often uh draws what published online. And most of the information gathered on the computer. So a lot of what's online about the product is paid for already and sponsor the review or top 10 list written by brands themselves or articles you die designed to rank well in search. If AI reads mostly paid content content, then its summary can repeat its bias back to you. Now it's a calm neutral sounding voice. And so you uh think that you're getting a true picture of the market but actually you may getting a advised uh uh advertised the budget of whoever spend the most.

So the accuracy is there that uh do we have this uh accurate information? This is something that we have to struggle with. And yeah, so we have to understand how we use this data. And also we have to carefully uh design. Are we uh are we promoting the AI the way we want to use it to shape our society or are we repeating ourselves what's existing in the system that we cannot really get out? So that's the question that we have to ask before we use the AI. And next way I'm going to go uh move on I'm continue our program with the market research and also health and a paradox.

So thank you for listening from yesterday, today, and tomorrow. And we'll be back next week. Thank you.

Guest (Male): You've been listening to Yesterday, Today, and Tomorrow with In Song Kim. You can also find more from In Song Kim at insongkim.org. That's I N S E O N G K I M dot O R G. Thank you for listening to the show.

Guest (Male): Maybe you've been waiting for the right time to make your move. At Eastwood Homes, that time could be right now. For a limited time, get a 4.99% interest rate on select move-in ready homes. Beautifully designed, thoughtfully built, and ready right now. No waiting for some day. No wondering when the time will be right. It's time. All that's missing is you. Eastwood Homes, built with care. Visit eastwoodhomes.com to find your move-in ready home today. Restrictions apply. See new home specialist for details.