What is artificial intelligence?

Summary

Brains On! – “What is Artificial Intelligence?”

Date: August 22, 2023
Host: Molly Bloom
Kid Co-Host: Sydney (Encino, California)
Guests: Rosie Dupont (Brains On! producer), Avital Balwit (AI researcher and Rhodes Scholar)

---

Episode Overview

This episode explores the basics, potential, and future of artificial intelligence (AI) in a fun, accessible way for kids and curious adults. Through playful sketches, analogies, and expert interviews, Molly Bloom, Sydney, and special guests break down what AI is, how it works, where we already encounter it, and what the future might hold.

---

Key Discussion Points & Insights

1. Jumping into AI – Everyday Examples

• AI isn't just robots! Many people think of robots when they hear AI, but it also exists inside programs and apps we use every day (04:04).
• Examples of AI in daily life:
  • Voice assistants (Alexa, Siri) ([04:21])
  • Autocomplete in search engines ([04:36])
  • Movie or music recommendations ([04:55])
• Sydney shares how she uses AI to help write text messages, especially to help apologize to friends ([05:53]).

Sydney: “Maybe when I'm texting my friends, it will auto generate some texts for me...I'll ask it saying like, my friend is mad at me. What do I say to make her feel better?" (05:53)

2. Why Are People Excited and Worried about AI?

• AI brings possibilities, like making tasks easier and more creative, but some people are skeptical or nervous – for instance, about privacy and the increasing power of these systems ([05:22], [07:08]).
• Sydney expresses concern about AI collecting lots of information through human interaction (07:08).

Sydney: “Maybe that humans talk to it a lot. Right. It's collecting all that information, and sometimes it can get more powerful, I guess.” (07:08)

3. What is AI, Really?

• AI as a tool: Rosie compares AI to the evolution of tools, from hammers and wheels to the printing press, and now to computers and AI ([09:53]-[10:43]).
• Algorithms: Explains that an algorithm is just a set of instructions, not something mystical ([10:47]-[11:32]).
• Data: AI needs lots of data to learn patterns – like fortunes, books, web pages, etc. ([12:05]-[12:28])
• Pattern prediction: AI doesn’t “understand” – it predicts the next word or action based on its training data ([13:42]-[14:12]).

Rosie Dupont: "The AI doesn't actually understand the meaning of the sentence. It's just seeing what sorts of words might follow other words." (14:12)

4. AI vs. Humans: Who Makes Better Fortunes?

• Human vs. AI comparison: Rosie and the AI both write fortune cookie fortunes about dogs and birthdays. The results range from silly to awkward, highlighting how AI can sometimes be off or lack creativity ([14:51]-[16:21]).
• AI makes mistakes: Even big AIs can get facts wrong. Google’s Bard gave an inaccurate answer about space telescopes in its demo ([16:29]-[17:06]).

Rosie Dupont: “Google's chatbot Bard made an error...claimed that the James Webb Space Telescope took the very first pictures of a planet outside of our own solar system. That is not true.” (16:29)

5. Why Shouldn’t We Always Trust AI’s Answers?

• AI is only as good as its training data. It doesn’t check facts or understand the truth ([17:41]-[18:29]).
• It can sound confident about things that are simply wrong.
• Advice: Double-check anything important an AI gives you.

Rosie Dupont: "Don't assume what they say is true." (18:29)

6. The Future of AI – Imagination and Expert Views

Expert Guest: Avital Balwit ([22:01]-[28:00])

• How do big, modern AIs work?
  • Large language models “read” enormous amounts of material and get good at predicting words.
  • By learning this core task, they also pick up other skills like summarizing, answering questions, even coding ([22:15]).
• A possible day in a future AI world:
  • Woken by a personalized AI assistant
  • Take new medicines invented with AI
  • Use self-driving cars, interact with robot servers
  • People might work less, have more free time
  • Ask your AI to make a custom movie based on your story to watch that evening! ([23:07])

Avital Balwit: “In the evening you might decide to watch a movie...you'd ask your AI system to create a movie based on a story that you come up with, and you'd fall asleep watching that movie together.” (24:05)

• Does AI have emotions?
  • Not really! They can sound emotional because they imitate what humans write, but they don’t feel emotions themselves—at least, not yet ([24:21]):

Avital Balwit: “You should think about AI systems right now as kind of like actors in a play. They're playing Characters that have emotions and might say that they're angry or scared or happy or sad. But the actual AI models aren't feeling these things as far as we know.” (24:21)

• Can AI make us smarter or nicer?
  • Probably! AI tutors and helpers could explain things in ways that work best for each student and help people work through problems patiently ([24:59]-[25:46]).
• Ensuring honesty and safety:
  • Researchers are trying “interpretability” – like brain scans for AIs – to see how they work and make them more trustworthy ([25:51]).
• Will AI go rogue?
  • Current AIs can’t act independently; they only do what we ask them. This may change, but smart people are working on controls ([26:29]-[27:08]).
• Dreams for AI:
  • Avital hopes AI will help create super-fast, eco-friendly trains or planes so people can visit each other quickly ([27:37]).

7. Predicting the Impact: Lessons from History

• People have worried about new technologies before (electricity, calculators, etc.), but their impacts are hard to predict ([28:43]-[30:16]).
• Some predictions go nowhere (flying cars); others, like personal computers, come true ([30:16]-[30:52]).
• Sydney’s hope: AI will help prevent car accidents and reduce traffic ([31:00]).
• Molly’s hope: AI can help humanity face big problems like climate change ([31:07]).

---

Notable Quotes & Memorable Moments

• “Artificial intelligence is when a machine can perform tasks usually only associated with human intelligence and get better over time.”
  — Sydney (03:56)

• “An algorithm is just a set of instructions.”
  — Rosie Dupont (11:03)

• “AI doesn't actually understand the meaning...it's just seeing what sorts of words might follow other words.”
  — Rosie Dupont (14:12)

• “Don't assume what they say is true.”
  — Rosie Dupont (18:29)

• “With an AI system, they can spend a lot of time explaining different topics to you...in exactly the way you like.”
  — Avital Balwit (25:02)

• “You should think about AI systems right now as kind of like actors in a play. They're playing Characters that have emotions...but the actual AI models aren't feeling these things as far as we know.”
  — Avital Balwit (24:21)

• Molly’s big takeaway:
  “AI is when a machine can complete tasks typically associated with human intelligence and learn to improve over time.” (32:08)

---

Fun & Interactive Segments

• Band practice & AI flyer generator sketch ([00:20]-[03:41]): Playful intro shows AI writing quirky band flyers.
• Mystery Sound Game ([19:38]-[20:16], [33:57]-[35:26]): Listeners and co-host guess the source of a strange sound—a fun break that keeps things lively.
• Haiku challenge ([20:26]-[21:06]): Listeners are encouraged to send in haikus about prehistoric creatures.

---

Timestamps for Important Segments

• AI in everyday life explained – [03:50]-[05:22]
• Sydney’s experiences with AI – [05:53]-[06:51]
• Concerns about AI and privacy – [07:08]-[07:26]
• What is an algorithm? – [10:47]-[11:32]
• How do AI models learn? – [12:03]-[14:20]
• AI sometimes gets things wrong – [16:29]-[17:41]
• Expert interview with Avital Balwit – [22:01]-[28:00]
• AI in the future: custom movies, better medicine – [23:07]-[24:15]
• Does AI feel emotions? – [24:21]-[24:56]
• Big hopes for AI’s future – [27:37]-[31:07]

---

Closing Thought

The episode ends with a reminder that predicting the future of new technology is tricky, and encourages listeners to stay curious, double-check AI’s work, and dream up ways AI could do good in the world—while still making room for band practice, bean bags, and a little bit of silliness.

---

For More: Visit Brains On! to send questions, mystery sounds, or prehistoric haikus!

Transcript

[Sydney]

You're listening to Brains on, where we're serious about being curious. Brains on is supported in part by a grant from the National Science Foundation.

[Band Member]

Awesome work, everyone. That's a wrap on band practice. That was so punk.

[Rosie Dupont]

See you later.

[Band Member]

Did anyone see my guitar pick? Hey, that's my incredible new band. Loudly with a chance of scream Balls. We're kind of a post punk, post hardcore, post ska, post office, post Malone, post serial rock band. And we've been practicing so much. Like tons. Like at least 45 minutes. We're finally ready for our big debut this Saturday. Now all I need is a cool flyer, a poster that'll convince everyone and their Doc Martin wearing grandma to to turn up for our big show.

[Molly Bloom]

Let's see.

[Band Member]

Hey. Hey, friend. Do you like fun? Then come see my band, okay? Please? Okay, bye. Ugh, that was terrible. It doesn't capture any of the essence of my exuberant yet slightly tortured musical genius energy. Let me try again. No, that's worse. No. Why is this so hard?

[Sydney]

Hard?

[Rosie Dupont]

Hey, buddy.

[Band Member]

Oh, by the looks of it, I'd say you're trying to make a flyer for your new band that captures the essence of your exuberant yet slightly tortured musical genius energy. Is it that obvious? Why don't we ask the new chatterbrain for help? It's the new software I made. Ooh, is that the new artificial intelligence powered text generator you invented? Yeah, let me fire it up and I'll type in a prompt. Let's see. Write a jaw dropping, eye popping, cool as all get out. Flyer for my new band. Loudly with a chance of scream balls. Our show is Saturday.

[Rosie Dupont]

Computing. Computing. Here you go.

[Band Member]

Okay. It says, witness the musical stylings of a band that may or may not be playing music on a day that comes after Friday but before Sunday. And there's a picture of a carrot for some reason. Yeah, that's not the vibe, bro. But I do like the carrot. Yeah, it needs more oomph. Something that expresses your signature blend of heart pounding beats, rebellious anthems and raw, untamed energy. How did you know about our raw, untamed energy? Okay, chatterbrain, can you write the flyer again, but be more punk about it.

[Rosie Dupont]

Computing. Computing. Here you go.

[Band Member]

This is more like it. Scream your lungs out to the anthems of rebellion and unbridled passion. Tear down the walls and embrace the chaos. Loudly with a chance of scream balls. This Saturday only.

[Sydney]

Okay.

[Band Member]

Wow. It's perfect. And the carrot is wearing sunglasses now. And it has sideburns. Sideburns are so cool. This flyer is weird, but kind of amazing. Now, can your AI also go out and post this on every telephone pole and message board in a 500 block radius? Sadly, no. Aw, man.

[Molly Bloom]

You're listening to Brains on from APM Studios. I'm Molly Bloom and I'm here today with Sydney from Encino, California. Hi, Sydney.

[Sydney]

Hi, Molly.

[Molly Bloom]

So, Sydney, we're here to talk about AI, which stands for artificial intelligence.

[Sydney]

Artificial intelligence is when a machine can perform tasks usually only associated with human intelligence and get better over time.

[Molly Bloom]

Yeah, some people think of AI and picture robots, but that's not quite right. Robots can have AI programs inside them the same way we have brains inside our bodies. But AI doesn't have to have a body.

[Sydney]

It can be a program on a computer or phone.

[Molly Bloom]

Yeah, voice assistants like Alexa and Siri are powered by AI how can I help you? Where is the nearest pajama theater costume store?

[Avital Balwit]

Big Mama's pajama drama o Rama is 1.5 miles away.

[Sydney]

And some search engines autocomplete sentences using artificial intelligence. You know, like when you start typing in cats r and it fills in.

[Avital Balwit]

Options like, cats are cute, Cats are related to lions, Cats are better than dogs.

[Sydney]

Whoa, that's a hot take right there.

[Molly Bloom]

Or when you're looking to stream a movie and it recommends something new, that's AI too.

[Sydney]

Based on your love of cooking shows.

[Avital Balwit]

And monster movies, we think you might like Cakenstein's monster.

[Molly Bloom]

And in the past few years, new AI programs like ChatGPT, Dall E and Beethoven have started doing things we used to think only humans could do, like writing poetry, creating lifelike images, making music, and a whole lot more.

[Sydney]

Some people are excited about the possibility of AI and some people are skeptical.

[Molly Bloom]

Right. No matter what you think of it, it's best to start by trying to really understand what it is and how it works. So, Sydney, you wrote into us with a question about AI what made you curious about this topic?

[Sydney]

At school, when they started teaching us about AI and how it works, I got interested in it.

[Molly Bloom]

Do you use any artificial intelligence models in your day to day life?

[Sydney]

Sometimes. Maybe when I'm texting my friends, it will auto generate some texts for me when I can't think of anything. So I will. I'll ask it saying like, my friend is mad at me. What do I say to make her feel better? And it will write me something and then I can edit it a bit and then I'll copy and paste that.

[Molly Bloom]

Oh, that's awesome. Do your friends know that you're sending them something? Okay, so you say, like, the AI wrote this for me?

[Sydney]

Yeah.

[Molly Bloom]

And how do they feel about that? Is that something your friends do back to you too?

[Sydney]

Yeah, I guess. Kind of. I don't really know.

[Molly Bloom]

Okay, so you're like, I'm sorry you're mad at me. The AI helped me apologize to you, basically.

[Sydney]

Yeah.

[Molly Bloom]

That's good. I mean, sometimes it's hard to find the right words. Right. And if this artificial intelligence is. Is drawing from other apologies, you know, probably does a pretty good job. Have you ever been, like, surprised at the texts it's helped write for you? Like something that. Totally unexpected?

[Sydney]

Yeah, sometimes. Cause it's like nothing I've ever thought of. I'm like, oh, I really should have said that.

[Molly Bloom]

No, that's kind of like teaching you how to have good conversations with people in a way.

[Rosie Dupont]

Yeah.

[Molly Bloom]

How handy. Does anything concern you about A.I.

[Avital Balwit]

Mm.

[Sydney]

Maybe that humans talk to it a lot. Right.

[Molly Bloom]

Mm.

[Sydney]

It's collecting all that information, and sometimes it can get more powerful, I guess.

[Molly Bloom]

Mm. Mm. So you think there's possible for good? It makes you a little nervous.

[Sydney]

Yes.

[Molly Bloom]

I mean, it's kind of fascinating because the AI Stuff has really changed a lot in the last year, and a lot of new developments have happened. So when you're listening to this episode, you know, there could have been even more new things by the time this episode comes out. Whoa. What was that?

[Rosie Dupont]

Tis I, the rose that Knows. I can read palms, tea leaves, and table leaves and see into the future with these enormous goggles. They make me slightly dizzy, but I can't return them. So an image is emerging. I see a young girl whose name begins with P. No, sorry, S. Sydney. Yes. And Sydney is eating a cinnamon raisin bagel with cream cheese.

[Molly Bloom]

Oh, it's brains on producer Rosie Dupont.

[Rosie Dupont]

Otherwise known as the Rose that Knows. Hi, Sydney.

[Sydney]

Hi.

[Rosie Dupont]

I just had a vision. I saw you eating a cinnamon raisin bagel with cream cheese in my future. Goggles. That'll happen within the week, guaranteed. Are you excited?

[Sydney]

Uh, I guess.

[Rosie Dupont]

Great. Me too. That'll be $5, please.

[Molly Bloom]

Sydney does not need to pay you.

[Rosie Dupont]

Rosie clients never want to pay me for my brilliant and occasionally accurate fortunes. My only paying fortune telling job right now is writing fortunes for fortune cookies. And now I'm worried AI might take that job too.

[Molly Bloom]

Nothing could ever replace you. Your fortunes are so unusual. Like, remember the one that said, looking for a fortune? You've come to the wrong place. I' a piece of paper.

[Rosie Dupont]

Oh, yeah.

[Sydney]

I like the one that said, thanks to you, this fortune cookie is broke. Much like the fortune teller writing this fortune.

[Rosie Dupont]

Oh, too real. I've been practicing my fortune writing skills for years, and they're still middling at best. But I am extremely competitive, and I refuse to let AI Beat me at my own game. So I decided to learn all of its secrets. I can tell you how it works if you're interested.

[Sydney]

Sure.

[Molly Bloom]

Yeah.

[Rosie Dupont]

Okay, so, first things first. Artificial intelligence is a tool, and humans have been using tools to make their lives easier for at least 2 million years. Early humans used simple tools like the wedge and the hammer to get things done. Fast forward and tools got more complex. There was the wheel, the printing press.

[Molly Bloom]

The toilet, the train.

[Sydney]

Computers, refrigerators, and radios.

[Rosie Dupont]

Then, around 1950, the first AI was born. Artificial intelligence programs were different from all the other tools that came before because they were designed to complete tasks that, up until that point, could only be performed by human brains.

[Molly Bloom]

Ba ba ba ba ba ba ba ba ba ba. Brains on.

[Rosie Dupont]

Nice. So that got me thinking. How do AI Programs work? Turns out an artificial intelligence model is an algorithm.

[Sydney]

I have definitely heard of algorithms.

[Molly Bloom]

Yeah, I remember talking about algorithms in high school math class.

[Rosie Dupont]

Yeah, it's a word we hear everywhere these days. And even though it sounds fancy, an algorithm is just a set of instructions. For example, my algorithm for making a soft boiled egg would be, put water and egg in a pot, turn on stove, bring to boil, boil for 3.5 minutes, turn off stove, crack egg, Enjoy.

[Molly Bloom]

I love soft boiled eggs. That's my second favorite breakfast.

[Rosie Dupont]

Hold on, Molly. I'm getting a feeling. There's a tingling in my toes. Is your first favorite breakfast cottage cheese with pineapple and a cup of strong black coffee?

[Molly Bloom]

How did you know?

[Sydney]

Is it because Molly's eating cottage cheese with pineapple and drinking black coffee right now?

[Rosie Dupont]

Sydney, sh.

[Molly Bloom]

Oh, yeah.

[Rosie Dupont]

Anyway, to work, an AI Model or algorithm also needs a computer to follow the instructions and data, which is a big pile of information. Data can be a lot of different things. It could be a bunch of fortune cookie fortunes or web pages and books about a specific topic, like cooking, for example.

[Molly Bloom]

Okay, so for an AI Model or AI Algorithm to run, you need a computer and lots of data.

[Rosie Dupont]

Bingo. I'll give you an example. Pretend I want to have an AI do my job and write a fortune for me. Not that I do, but here's how it would work. I'd open an AI Model that's designed to write things. This AI Model was created by humans feeding a computer program a big pile of data. The data in this case would be a massive number of books, articles, blog posts, and other pieces of writing, mostly written by humans. The AI program has figured out patterns in the data by looking at it over and over again and gotten really good at predicting the next word in a sentence. Surprisingly, once the model learns this skill, it's able to do a bunch of other tasks, like write stories, summarize complicated ideas, and even write computer code. So to get this AI system to write a fortune for me, first I'd give it instructions. Write fortune cookie, fortune about dogs and birthdays. And the AI would take my sentence and break it apart into individual words.

[Molly Bloom]

Right?

[Rosie Dupont]

Fortune, cookie, fortune about dogs and birthdays. Then it put each word into its algorithm, and based on the patterns it's learned from its data, it'd make a best guess at what the next word in the fortunate's writing might be making. Best guess. Keep in mind, the AI doesn't actually understand the meaning of the sentence. It's just seeing what sorts of words might follow other words.

[Molly Bloom]

Like if I said it was so hot, my ice cream blank. The AI would probably guess the next word is melted.

[Rosie Dupont]

Exactly.

[Sydney]

Or if a sentence started with Tiffany.

[Rosie Dupont]

Was so tired she needed the algorithm might add the words a nap or more coffee. Right. Often, there's more than one word that could go next. So that's why AI programs can come up with so many unique sentences. As the algorithm guesses one word after another, the program spits out its answer and voila, you have your fortune about dogs and birthdays. Here it is.

[Avital Balwit]

Wag your tail, it's your birthday.

[Molly Bloom]

Hmm. I don't know if that's the best fortune.

[Sydney]

Yeah, what if it's not your birthday or you don't have a tail?

[Rosie Dupont]

Yeah, true. Not a huge fan. But it's a good example for my purposes because it shows us that AI isn't always great at doing the job it's been given. Humans have to train AI models so their algorithms get more specific and their answers get better. Wait, I have to ask. What do you think of my version? Let me just adjust my fortune telling goggles and. Okay.

[Molly Bloom]

Woo.

[Rosie Dupont]

Getting older is a dog gone. Pain in the tail. But don't give up. You're doing pawsome.

[Molly Bloom]

Hmm.

[Rosie Dupont]

Yeah, it's okay, I get it. Like me, AI learns through trial and error. And just like an AI model, I need a human teacher to help me get better. Any feedback?

[Sydney]

Maybe make it a little more positive?

[Molly Bloom]

Yeah, and maybe like, actually a fortune that predicts the future.

[Rosie Dupont]

Noted. Okay, okay, I'll take it in stride. All right, how about happiness and Cake are coming your way for soon you'll see A dog's birthday.

[Sydney]

Oh, that's much better. And now I really want to get invited to a dog birthday party.

[Rosie Dupont]

Thank you, thank you, thank you.

[Molly Bloom]

Yeah, that one's the best of the three predictions by far.

[Rosie Dupont]

Thanks, Molly. That reminds me. Another thing to keep in mind about AI is artificial intelligence. Models sometimes get things totally wrong. For example, Google's chatbot Bard made an error in its very first demo in early 2023. It claimed that the James Webb Space Telescope took the very first pictures of a planet outside of our own solar system. That is not true. The first image of an exoplanet was taken all the way back in 2004, almost 20 years ago by a totally different telescope.

[Sydney]

Awkward.

[Rosie Dupont]

I'll say. But it's relatively easy for AI models to make mistakes like this because they are making a best guess at the right answer. They aren't all knowing. Like, say, me, for example. Well, anyway, humans are still developing lots of AI tools, and some systems have done more and better training than others. So some are better at guessing the right answers than others. With that in mind, you shouldn't automatically trust what a model writes or creates because it might not be true or right.

[Molly Bloom]

Oh boy. Okay, we will keep that in mind.

[Rosie Dupont]

Please do. Because an artificial intelligence is only as good as the data it has. So, like, if I had an artificial intelligence model and I. I wanted it to tell me how to cook a soft boiled egg, but it didn't have good information on how to cook eggs, it would probably give me a strange and incorrect answer. Speak softly, boil water, crack egg on head. Soft boiled and egged. Not the way I want to start my day. So remember, AI models are good at doing a lot of things, but don't assume what they say is true.

[Molly Bloom]

We won't.

[Rosie Dupont]

Lucky for me, my fortune cookie fortunes never have to be factually accurate. So I'm going to go make some things up, and hopefully they'll be weirder and more creative than anything my AI competition dreams up.

[Molly Bloom]

I believe in you, Rosie.

[Sydney]

Yeah, I bet your fortunes are going to be positively pawsome.

[Rosie Dupont]

Thank you. Wait. My future goggles are giving me something. A migraine, but also a vision. I can see myself surrounded by adoring fans. And I'm at the fan store. Huh. Not surprising, really. I. I've been needing to get a new fan. It's been hot lately. Toodaloo, you two.

[Molly Bloom]

Bye, Rosie. By. All right, Sydney, that was a lot of information about artificial intelligence, but now it's time for you to use your human intelligence to make a best guess. Cause it's time for the mystery cell. Are you ready?

[Sydney]

Yes.

[Molly Bloom]

All right, here it is. Hmm.

[Sydney]

What do you think it sounds like A microphone rubbing into a carpet or like a jacket, like something soft.

[Molly Bloom]

Ooh, yeah, that's a really good guess. I have no idea what it is either. It kind of sounded to me like, I don't know, shoveling carpet. I mean, it sounds like carpet was involved, but I don't know why you'd be shoveling the car.

[Sydney]

Like something fuzzy being rubbed.

[Molly Bloom]

Mm. Mm. Okay. Well, we're gonna get another chance to hear it. Make a guess and hear the answer after the credits.

[Sydney]

So stay with us.

[Molly Bloom]

Hey, friends. We're working on an exciting new quiz show about prehistoric animals, and we want to hear from you. Pick your favorite prehistoric creature and write us a haiku about it. A haiku is a short poem with five syllables in the first line, seven syllables in the second, and five syllables in the third. Sydney, do you have an example of a prehistoric haiku you'd like to share?

[Sydney]

Yes, here's my example. Mighty Mosasaurus swimming in the deep blue sea, roaring with delight.

[Molly Bloom]

Oh, that was excellent. Well done. Can you describe what a Mosasaurus looks like?

[Sydney]

It's a prehistoric animal that is really big and it's like a shark.

[Molly Bloom]

Oh, so very cool. Well, listeners, we want to hear from you. You record yourself reciting your haiku and send it to [e-mail] contact. And while you're there, you can send us mystery sounds, drawings, and questions like this one.

[Sydney]

What's the difference between a mouse and a rat?

[Molly Bloom]

Again, that's brainson.org contact.

[Sydney]

Keep listening. You're listening to brains on. I'm Sydney.

[Molly Bloom]

And I'm Molly. And today we're exploring the how and.

[Sydney]

Why of AI we learned that AI or artificial intelligence, is all around us.

[Molly Bloom]

And that AI Programs are different from other tools because they can imitate human intelligence and learn over time.

[Sydney]

But AI Gets better at doing things through trial and error, and sometimes the guesses it makes aren't quite right. So it's always good to double check the answer AI gives you.

[Molly Bloom]

As AI Gets better and better, it's going to change our lives in unexpected ways. Here to talk to us about that is writer, researcher, and winner of the Rhodes Scholarship, Avital Balwit.

[Sydney]

Hi, Avital.

[Avital Balwit]

Hi, Sydney.

[Molly Bloom]

So, Avital, can you explain how AI works?

[Avital Balwit]

So I guess I mostly think about the type of AI That I work on, which is sort of large language models and these are basically computer programs that have read a bunch of different things. They basically read the entire Internet. And they've gotten really good at sort of predicting the next word that'll come next in a sequence of words. But they haven't just learned that skill by reading all that material. They also got really good at other skills like summarizing things that you ask them to summarize or answering questions or writing code. And so they've gotten a bunch of different skills that work in a bunch of different areas. But they're basically big, very smart computer programs.

[Sydney]

So, Avital, you think a lot about the ways artificial intelligence might impact our lives in the future. Imagine a day 15 years from now. Can you describe how we might encounter AI over the course of that day?

[Avital Balwit]

Sure thing, yeah. So the first thing that might happen is that you'd be woken up by your AI assistant. And your AI assistant is an agent that exists just to help you. And so they answer questions that you have, they help you run errands. And today they would wake you up with a song they made just for you based on the type of music that you like. And after that you might take some vitamins. And that's because AI has helped invent new medicines that help people be healthier and to live a lot longer. And then you might decide to go to a restaurant for lunch, and you would take a self driving car to get there, and you'd meet some friends there, and there would be robots there helping to make the food and serve the food, and it would be easy to talk to these robots in a very natural way. And after lunch, you might decide to go to a park and you'd notice that there are a lot more people around during the day because people don't have to work quite as much. And so you'd see people doing hobbies and spending time with their friends. And then in the evening you might decide to watch a movie. But all the movies at home are movies that you've already seen. And so you'd ask your AI system to create a movie based on a story that you come up with, and you'd fall asleep watching that movie together.

[Sydney]

Wow. I was also wondering, do AI models have emotions? Can AI get angry or scared?

[Avital Balwit]

Right now we don't think that AI systems have emotions, but sometimes they seem like they do. And that's because they've read a bunch of books where humans characters have emotions. And so you should think about AI systems right now as kind of like actors in a play. And they're playing Characters that have emotions and might say that they're angry or scared or happy or sad. But the actual AI models aren't feeling these things as far as we know. But AI systems are getting smarter and more complicated every year. So it's possible that this will change in the future. And this is something that people are researching. But right now we think that AI systems are just sort of imitating the things they've read.

[Sydney]

Wow, that's really interesting. But can AI make us smarter and nicer?

[Avital Balwit]

I think that AI systems will probably be able to make us smarter and nicer. You can think about right now when you're in class, you might have a bunch of questions for your teacher, but there's probably only so many questions of yours that they can answer before they need to move on to another student. But with an AI system, they can spend a lot of time explaining different topics to you, and they can explain them to you in exactly the way you like. Like maybe you like to learn by being told students stories. And the AI system can help with that. And so it could be sort of a really great teacher or tutor. And they can also probably help you be nicer, because at my company, we try to make AI systems that are very polite and patient and so you can talk to them about your problems, they can help you think through them, and they have a bunch of time to help you work on different things. And so in the same way that a good teacher can help you become more of the type of person that you want to become, AI systems can probably do this too.

[Sydney]

How can we make sure that an AI model is honest in telling the truth?

[Avital Balwit]

That's a really great question. And a lot of people are working on this problem because right now, AI systems sometimes get things wrong or make things up. So one thing that we're trying to do is something called interpretability. And that's a fancy word for basically doing brain science on the models on the AI systems. Like right now, if you do a brain scan on a human, you can see part of the human's brain is for long term memory, part of it's for vision. And basically we're trying to do the same thing on AI systems so we can understand what's going on in there. And that might be able to help us see whether it's telling the truth or not. And that can help us know when we're able to rely on what our AI system is saying, because we want to be able to trust what it's saying.

[Sydney]

Can AI go rogue? Can it do things that we can't predict. Can it start doing stuff on its own?

[Avital Balwit]

Yeah. So right now, AI systems are basically chatbots. And so they're things that you ask questions to, and they can answer your questions, but they're not doing a bunch out on their own in the world. So right now, you don't need to worry about AI systems going rogue. People are trying to make AI systems smarter and smarter every year and make them more independent so that they are able to go and do more things out on their own. And so this might be something to worry about in the future, but there's a lot of really smart people working on this problem and trying to find ways so that even in the future, AI systems still listen to us and still do what we want them to do.

[Molly Bloom]

And.

[Avital Balwit]

And this is a problem that we think we'll be able to solve by the time we have to worry about it.

[Molly Bloom]

So, Sydney, what's one thing you'd want AI to help with in the future? Like, Avital was talking about making things.

[Sydney]

That are specially designed for you, like songs, medicine, and movies.

[Molly Bloom]

Yeah, that movie idea blew my mind. Like, we can make a movie.

[Sydney]

I know.

[Molly Bloom]

Avital, what is, like, one thing you hope that it can do in the future that can help humans with?

[Avital Balwit]

I think sometimes I want to visit my friends, but they're really far away. And I bet there are ways to design, like, trains or planes so that they're very safe and very fast and even good for the environment. And so I would love if AI systems could help me design some kind of transportation so that I could visit people even when they're really far away, really quickly.

[Molly Bloom]

Ooh, I love that idea. Thank you, Avital.

[Sydney]

Thank you.

[Avital Balwit]

Yeah. Sydney, I really enjoyed these questions. Thank you so much for making them.

[Sydney]

You're welcome. Brains. Fun.

[Molly Bloom]

It's okay if you're kind of weirded out by the idea of AI. It's a strange new technology.

[Sydney]

Yeah. We have no idea how it might change our lives for better or worse.

[Molly Bloom]

And you might hear a lot of people make big predictions about what it'll be like in the future.

[Rosie Dupont]

AI is going to outsmart us by next year. It's going to replace teachers and raise our kids.

[Band Member]

I bet it'll be the next big comedian. Let's be real.

[Molly Bloom]

It's probably gonna fall in love with me. Ooh, I'll have so many AI Valentine's cards.

[Band Member]

No way AI's dating Taylor Swift.

[Rosie Dupont]

AI is gonna stop climate change?

[Band Member]

No.

[Rosie Dupont]

It's gonna destroy the economy.

[Molly Bloom]

But, like, what if it invents a.

[Rosie Dupont]

New kind of cake.

[Molly Bloom]

Like extra cakey cake, you know, I could get. But here's the thing. Us humans, we don't have the best track record when it comes to predicting how new technologies will change the world.

[Sydney]

Yeah, sometimes we've been right, but sometimes we've been way off the mark.

[Molly Bloom]

Like electricity. When it first came out in the late 19th and early 20th centuries, people worried that it would eliminate jobs. But electricity actually brought about more jobs.

[Sydney]

Then there's flying cars.

[Molly Bloom]

Oh yeah. During the 1950s and 60s, lots of people thought flying cars would be the next big thing in transportation.

[Sydney]

Sounds cool, but would there be sky traffic?

[Molly Bloom]

Ugh, it would totally wreck the view. Luckily, we're nowhere close to flying cars yet. But self driving cars might be pretty common soon.

[Sydney]

Oh, and calculators. They seem like a no brainer. Like who would be offended by a calculator, right?

[Molly Bloom]

But when pocket sized calculators were first introduced in the 1970s, some people thought they were going to hurt students.

[Sydney]

Like, how would kids ever learn math if they used calculators?

[Molly Bloom]

They'd never be able to pass a test on their own and their brains.

[Sydney]

Would turn to mush.

[Molly Bloom]

Obviously that didn't happen. And we still do some math by hand and some with calculators.

[Sydney]

But sometimes our predictions are right on the money.

[Molly Bloom]

For example, science fiction writer Arthur C. Clarke gave a speech in 1976 where he envisioned a world where everyone had the equivalent of a personal computer.

[Sydney]

And it's kind of amazing because back then computers were just being introduced.

[Molly Bloom]

They were humongous, often had to be connected to an external screen like your TV. And they were expensive, like the equivalent of over $6,000 today. But he knew this machine would keep getting smaller and more convenient until it became part of everyday life.

[Sydney]

Now we carry computers around in our pockets, right?

[Molly Bloom]

Those are our cell phones. So Sydney, do you have any other hopes for AI?

[Sydney]

Maybe if they could be able to help with traffic and prevent lots of car accidents.

[Molly Bloom]

Oof. I love that.

[Sydney]

How about you, Molly?

[Molly Bloom]

I'm hoping that AI can help us with climate change. Like maybe help us conserve energy and waste less. Potentially help the planet out. That's my hope.

[Band Member]

Okay people, get ready for the ride of your life. It's the band neighbors called Too loud and incapable of tuning. We're loudly with a chants of scream balls. And this is our new song, Anarchy in the AI Hit it in the neon glow we'll make make our stand to find our way in this brave new land Is the future now or is it a lie should we smash it up or give it a try?

[Molly Bloom]

Yeah. It's a brave new world of AI.

[Band Member]

Digital hearts A technical ally lost in circuits Our spirits try to hold on tight as wires multiply.

[Molly Bloom]

Yeah. AI is when a machine can complete tasks typically associated with human intelligence and learn to improve over time. We learned that AI models are algorithms that have learned skills by processing lots of data on very powerful computers. And that AI models are trained by humans to produce more accurate answers.

[Sydney]

And that one day AI might become better than humans at doing certain tasks.

[Molly Bloom]

So it's up to us humans to manage AI and make sure it's doing good for the world. Sydney, do you think humans are up to the task?

[Sydney]

Yes.

[Molly Bloom]

Me too. Well, that's it for this episode of Brains on.

[Sydney]

This episode was written by Rosie Dupont and Anna Wegel, with production help from Molly Bloom, Anna Goldfield, Aron Waldos, Lassie Shayla Farzon, Nico Gonzalez Whistler, Ruby Guthrie, and Mark Sanchez.

[Molly Bloom]

This episode was edited by sand and Totten and sound design by Rachel Breese. We had engineering help from Alex Simpson and Jay Markovitz. Beth Pearlman is our executive producer. The executives in charge of APM Studios are Chandra Kavati, Alex Shaffert and Joanne Griffith Griffith. Special thanks to Angelica Chen and Steven Liu.

[Sydney]

Brainson is a non profit public radio program.

[Molly Bloom]

There are lots of ways to support the show. Head to brainson.org while you're there.

[Sydney]

You can subscribe to our Smarty Pass, which lets you listen to ad free episodes and other awesome bonus content.

[Molly Bloom]

And you can submit your questions and haikus. We love getting Haikus from yous. Okay, Sydney, are you ready to listen to that mystery sound again?

[Sydney]

Of course.

[Molly Bloom]

Okay, here it is. Okay, new thoughts. What are you thinking?

[Sydney]

So I kind of changed my mind. It sounds like it's sweeping something because in that pattern, you know, like that's.

[Molly Bloom]

Yeah, it's rhythmic for sure, but definitely.

[Sydney]

Okay. I'm thinking like sweeping a carpet back and forth. That's my answer.

[Molly Bloom]

Great idea. I think it's someone brushing the hair of a snowman. I'm just putting it out there. I think it's possible. You know, you gotta style your snowman's hair. Right. Thing is. Okay, you ready for the answer?

[Sydney]

Yes.

[Molly Bloom]

Okay, here it is.

[Rosie Dupont]

Hi, Brains on.

[Sydney]

My name is Amania and I live in Delhi, India. And that was the sound of me pressing my bean bag. Bye. Thank you.

[Molly Bloom]

Oh, like a beanbag chair.

[Sydney]

I have a beanbag. I don't know why I didn't recognize.

[Molly Bloom]

That, but yeah, now that you've heard it. Does that make sense that it's a beanbag chair?

[Sydney]

It makes so much sense because I always press against it. Like when I sit on it, it makes that exact sound.

[Molly Bloom]

Yeah. It's so hard to hear these sounds out of context. And you're like, I've never heard that before in my life. Even if you hear it every day?

[Sydney]

I do that every single day.

[Molly Bloom]

Amazing. Is it comfortable, your bean bag chair?

[Avital Balwit]

Yes.

[Sydney]

It's really big, so I like to sit on it a lot.

[Molly Bloom]

Sounds delightful. Now it's time for the brains honor roll. These are the incredible kids who keep the show going with their questions, ideas, mystery sounds, drawings and high fives. Rashid and raeed from wellington, new zealand peter from kent, ohio genevra from santa fe, new mexico rosie and david from chicago anna from pennsylvania heavenly from somerset, new jersey sebastian from los angeles daphne and elliot from midlothian, virginia vivian and elizabeth from fairport, new york emma from mckinney, texas jonathan from round lake, illinois cedar and sage from westchester, ohio ada from suffolk, england grand from grays lake, illinois river from fort wayne, indiana yuba from nevada, san city, california siki from palo alto, california nora from brooklyn, new york gaia from chicago ariana from coventry, rhode island elias from long beach, california rowan from south pasadena, california ollie from melbourne, australia susan from hobart, australia arlo and eden from golden, colorado waverly from lone jack, missouri bonnie from dallas, texas emerson and hayden from medford, massachusetts dominic and lorenzo from san diego, california sophia from philadelphia audrey from durham, north carolina catherine from abena, japan clara from amsterdam brielle from north lauderdale, florida michelle from maryland emerson from italy kaden from brisbane, australia calista and emerson from medina, ohio gus from tetonia, idaho everett from dixon, california ava from new braunfels, texas indy and phoenix from portland, oregon audrey from sammamish, washington nate from leduc county, alberta avy from kennedale, texas anders from naperville, illinois luca from vineyard, utah jenna from springfield, new jersey belden, santa cruz, california soren from asheville, north carolina aiden from silver spring, maryland maeve from brooklyn, new york jack from hoboken, new jersey cadence and cormac from canningvale, australia naomi from albuquerque, new mexico zoya from atlanta adam from south beloit, illinois izzy, levi and finn from chicago jack from boise, idaho sophie from new york city liliana from brooks, alberta nguyen from hanoi, vietnam and lucy from longmont, colorado. We'll be back next week with more answers to your questions.

[Sydney]

Thanks for listening.