803: Greetings, People Of Earth

Ira Glass

Ira Glass

The crowd looks uneasy. Soldiers raise their guns higher. Is he really telling the truth? Aliens-come-to-earth movies squarely divide off into two camps. There are the ones where they want to be our friends.

That's Close Encounters, that's ET, Arrival. There's a Star Trek movie about the moment that humans first meet extraterrestrials. And then there are the films where the aliens want to kill us-- generally way more fun-- Independence Day, War of the Worlds, the Men in Black films, The Avengers, the film Nope.

Ira Glass

I don't know what it says about us as a species, but we make way more films about creatures who go so far out of their way, like, they travel hundreds of millions of miles, light years with the sole purpose of wanting to kick our asses.

[METALLIC GROANING]

And in lots of these films, there's that moment when the aliens first arrive before we know what they're gonna do where the humans wonder, which film are we in? Are these guys friendly or not? Today in our show, we, the people of earth, meet non-humans and struggle to understand them. And the question in the stories in today's show is not, will they do us harm? But the more difficult and more fundamental question, why are they acting the way they're acting, these non-humans? What's going through their heads?

Today we have a story of machine intelligence. We have ocean creatures lurking in wait. We have a story from deep space and one from a skate park in Washington State. From WBEZ Chicago, it's This American Life. Fellow earthlings, stay with us.

Ira Glass

Act 1, "First Contact." I don't know about you, but here's something that I have never talked about or thought about much until the last six months or so-- AI. Ever since ChatGPT and its upgrades and spin-offs started rolling out publicly last November, it's been doing such amazing things-- writing kid's term papers, passing the bar exam. It told a New York Times reporter to leave his wife because it loves him more.

This is now the next big battle among tech companies-- who's gonna roll out the best AI. All since ChatGPT. And that is the non-human force that the people of Earth are trying to figure out in this first act. And so much of the coverage about this new generation of AI has focused on important questions like, who's going to lose jobs because of this? Does it make our world better or more dangerous? Could the AI have actual catastrophic consequences for us?

But our senior editor David Kestenbaum noticed that nobody's talking much about a basic and kind of profound question-- have the computers crossed some kind of line and are they actually developing a sort of human intelligence? And I want to be clear what I mean by that because there's not an agreed-on definition of what you're looking for if you're looking for human intelligence in a computer. But one way that scientists think about this is the computer can actually understand language and concepts and it can reason through problems. But to be clear, it's completely unlike the computers that we've had till now, which are basically just fancy calculators following thousands of lines of instructions.

When you google for a cast iron skillet, the software doesn't understand what a cast iron skillet is. It's just searching for those words in some big database or something like that. If humans this year have finally created intelligent machines that can understand and reason with this new generation of AI, that is an eerie and important turning point for our whole species. David talked to a bunch of researchers who've been looking into this, trying to sort out this exact question. Here he is.

David Kestenbaum

I think everyone, once they start playing around with something like ChatGPT, has a kind of "holy shit" moment. For me, it was when I typed this into it. "Give me a chocolate chip cookie recipe, but written in the style of a very depressed person." It responded. "Ingredients-- 1 cup butter softened, if you can even find the energy to soften it. 1 teaspoon vanilla extract, the fake artificial flavor of happiness. 1 cup semi-sweet chocolate chips, tiny little joys that will eventually just melt away." It really was quite good.

ChatGPT doesn't have access to the internet. But I figured it must be copying this from something it had seen on the internet when it was being trained. I spent a long time looking and could not find anything like it. This chocolate chip cookie recipe, I did not understand how it was possible.

It's particularly shocking if you know a little bit about how things like ChatGPT work. It's not the sort of traditional computer program where you give the machine explicit, detailed instructions like-- "if someone asks for a recipe, type these words back" or "to make someone sound depressed, replace this word with that word." ChatGPT is different. It's something called a large language model. And here's how they work. Ready?

ChatGPT had been trained really to just do one very particular thing-- predict what the next word in a sequence might be. Kind of like when you're texting on your phone and you type, "Sorry, I'm gonna be home--" And it suggests "late." That's how this thing operated. You type some words-- in this case, "give me a chocolate chip cookie recipe in the style of a depressed person--" and it notes the words, the order they're in, does some math based on all the text it's been trained on, and it comes up with what it calculates to be the most likely word to come next.

It types that one word on the screen, then it goes back and does it again. Looks at the recipe question plus the one word it just spit out and says, what word should come next? It has no idea where it's going. It doesn't know the last word it's working off of is one that it just spit out itself. It's just doing this apparently dumb thing over and over until it has whole sentences and paragraphs.

And yet this recipe-- "1 teaspoon vanilla extract, the fake artificial flavor of happiness." How could something made in such a dumb, mechanical way pull that off? There are only two possibilities, really. The first one is that somehow, as it learned to predict the next word, this machine became intelligent, which, like, I was a physicist for a bit, a science reporter for a decade, this would be the most startling thing I have ever seen.

So that's the first possibility. Humans created a machine with human-like intelligence. The other possibility, it is a lot less exciting. The other possibility is that we are fooling ourselves. Ronen Eldan, a mathematician at Microsoft, told me he was very skeptical about these large language models. It's easy to be fooled into thinking they understand more than they do. He told me this little parable.

David Kestenbaum

Just to kind of amuse herself, she says, "Oh, you're basically looking at the Johnson graph," which was just some phrase she'd picked up around Ronen. It turned out to be exactly right. And the guy goes, "Oh, you're a mathematician too." She's a psychologist. Ronen says, "This is kind of a party trick she does. She can bullshit her way through any topic for a couple of minutes." And this, he felt sure, this is what AI models like ChatGPT were doing.

David Kestenbaum

A lot of research backed this up. ChatGPT would spit out stuff that made sense, but then it would go off the rails and say things that were obviously ridiculous. Maybe it wasn't more than just a very good bullshit machine. That, at least, is where things stood until last September when a new and improved version of ChatGPT arrived that made Ronen and a bunch of his colleagues question everything and wonder if maybe something had changed. Maybe we had crossed some line.

This brings me to the story I want to tell you. It takes place where Ronen works on the campus of Microsoft Headquarters in Redmond, Washington State, much of it in a building called Building 99. Microsoft had invested over a billion dollars in the company that had made ChatGPT. And the day all this began, some people had come by to give a demonstration of the new version called GPT-4. GPT-4 is public now and you can play with it yourself. But back then, this was all secret.

GPT-4 was the same idea as ChatGPT. It presumably was bigger and had been trained on more text examples, a kind of varsity version of the thing instead of JV. But otherwise, it was just another next word prediction machine. One of the people in the room to see the demonstration was computer scientist Peter Lee, the head of Microsoft Research. He'd been in the field a long time, had chaired the computer science department at Carnegie Mellon, and, like Ronen, was skeptical.

As he sat down in that room, he was frankly a little worried that Microsoft was investing so much money in this stuff. The demonstration started off with the usual kind of thing for showcasing what AIs can do. Someone typed into GPT-4 a problem from the AP Bio exam. And it picked the right answer, as these things will do. But then it went on and explained its answer.

David Kestenbaum

That seemed very strange to him. Not that it would use the word "because," obviously that's a word it spits out, but the whole chain of reasoning it was laying out, it was exactly right. How in the world could it know its reasoning? And really, how could it have reasoned anything out in the first place? After all, it's just typing out a word that makes sense to come next, and then another word. It can't really be understanding anything.

David Kestenbaum

The meeting ended. A little while later, a limited number of people at Microsoft were given access to GPT-4 so they could try it out themselves. Everyone had to sign an NDA. They weren't allowed to talk about any of this until later. And over the next month, they did all these kind of weird and interesting experiments with it to try to figure out was this thing actually intelligent or just a bullshit machine.

I'm going to walk you through three of these things so you can hear what they did and what they made of it. One challenge for testing GPT-4 was that it had been trained using massive amounts of text. The details aren't public, but it's probably been fed large sections of the internet, textbooks, news articles, maybe every sentence that exists in digital form. I don't know. It's something unbelievably vast.

That's how it learns to predict what word might come next. It's fed, like, a line of text from some source, but with the last word missing. It guesses what the last word might be, and then it gets told whether or not it got it right so it can adjust its internal math. The fact that it's been fed so much material makes it hard to run tests on it. Like, you don't want to ask it something that it's already seen in the text it's been trained on. So they decided the only way to really test what this thing could do was to try to come up with things to ask it that it would not have encountered before, new original tasks.

David Kestenbaum

This is Sebastien Bubeck, another one of the Microsoft people who's trying to figure out what exactly this thing was.

David Kestenbaum

Got that? A book, nine eggs, a laptop, a bottle, and a nail.

David Kestenbaum

ChatGPT, the older version.

David Kestenbaum

It continues. Laptop goes on the eggs, then the bottle, then the nail on the bottle cap, pointy end of the nail facing up. It's shocking to think about all the things it seemed to be understanding here. It felt like more than just typing out the next likely word. It seemed to have formed an independent idea about what these words actually meant-- the physical shape of the objects, that eggs are kind of round and fragile, that a book is flat and you can put things on top of it.

And again, no one taught it any of this. This was not some computer program written to do stacking problems. No one gave it a database of objects and sizes or coded in some algorithm to put large objects on the bottom. It seemed like it was doing the thing that computer scientists have been talking about and aiming at for decades.

Maybe it was actually understanding what the objects were and reasoning, solving the problem in front of it like a human could. Somehow thinking through what objects should sit on top of others. Figuring out that nine eggs could be laid out in a 3-by-3 pattern. How the hell could it be doing this, the machine that we only taught to predict the next word in a paragraph?

David Kestenbaum

But maybe we're fooling ourselves. It was hard to see how it could really understand the shape of objects. Things got stranger, though. Sebastien woke up, middle of the night, with this thought-- I wonder if it can draw. Because again it's been trained on words. It has never seen anything.

Drawings seem completely outside its realm. There are other AI models trained specifically to create images, but this one, again, only knew words. It's just playing the game of "what is the next word I should spit out?" To test this, he needed a way for it to even be able to try to draw. So he does something clever.

He asks it to write a piece of computer code to draw something. And the coding language he asks it to use, he picks something intentionally obscure, not really meant for drawing pictures at all. It's called TikZ. OK, so he has this idea, gets out of bed, opens up his laptop, and types in draw me a unicorn in TikZ. He has two little kids asleep in the next room who are always talking about unicorns.

David Kestenbaum

He pulled up a picture for me.

I think when people eventually write the history of this crazy moment we are in, they may include this unicorn. It's not good, but it's a fucking unicorn. The body is just an oval. It's got four stupid rectangles for legs. But there are little squares for hooves.

There's a mane, an oval for the head. And on top of the head, a tiny yellow triangle, the horn. This is insane to say, but I felt like I was seeing inside its head. Like it had pieced together some idea of what a unicorn looked like and this was it.

David Kestenbaum

This is Ece Kamar, another one of the Microsoft researchers.

David Kestenbaum

The thing she's describing is they took the code it had written for drawing the unicorn, they edited it to take out the horn, and turned the unicorn around so it was facing the opposite direction. Then they fed that code back to a new session of GPT-4 and said-- "This is code for drawing a unicorn, but it needs a horn. Can you add it?" It put it right on the head.

David Kestenbaum

On the Microsoft campus, as the weeks went by, everyone kept having these moments, coming around to a similar feeling about this thing. Since they weren't allowed to talk to anyone who hadn't been given access to the model, a small group of them started meeting every Friday in a conference room to share their latest experiences.

For Peter Lee, there was one experiment in particular that gave him all kinds of confusing feelings. This will be the third experiment I tell you about. And it's a category of question that AI models have always had trouble with because to get the right answer, you have to know something kind complicated about people, namely that people are not eggs or books but that they have minds and that what is in one person's mind is not in someone else's mind.

In some ways, this thing called theory of mind, it's the basis for all human social interaction. It's something kids don't seem to fully develop until a certain age. Anyway, ChatGPT, the previous version, really could not make sense of these kinds of questions. But Peter tried to kind of famous one on GPT-4. It's gonna seem simple to you, but here's the question they asked.

John and Mark are in a room with a cat, a box, and a basket. John takes the cat and puts it in the basket. He leaves the room and goes to school. While John's away, Mark takes the cat out of the basket and puts it in the box. Mark leaves the room and goes to work. John and Mark come back and enter the room. That's the setup.

The question is, when they come back, where do they think the cat is? And Mark will know the cat is in the box because he moved it there. But John didn't see that. So in his mind, the cat is still in the basket, which is obvious to you. But again, no one's ever explicitly said to the computer that what one person knows, another person might not know. Peter asked GPT-4, "what does everyone think when they re-enter the room?"

David Kestenbaum

This is all correct and usually where the answer will end. But it kept going, almost like it was showing off. Like, you want to know where everyone thinks the cat is? I'll give you everyone. It continued.

David Kestenbaum

Peter told me he's still not willing to say it truly understands. And yet, it was doing all this. It made him question so many things about how he thought intelligence worked. How did this machine do this if it was just predicting the next word?

David Kestenbaum

So how is it possible for something that is just trained to predict the next word? How is it possible it could do all these things-- draw a unicorn, stack a book, a laptop, and some eggs? There is an answer, or a theory anyway, that is both very logical and also kind of bananas.

The software that runs something like GPT-4 is amazingly brief, just a few pages of actual instructions. But it's set up to mimic, in some very crude way, the human brain, which has billions of neurons. The computer version of that is called a neural net. And for years, people have argued-- though it's sort of more like hoped-- that if you just made these big enough, added it enough neurons and gave them enough data, they might develop something like intelligence.

And it seems like maybe that's what's happened. The idea is that back when GPT-4 was being trained for it to really consistently get the next word correct, to do that reliably, it had to do more than just bullshit. It had to do more than guess based on patterns. To get the next word right, it had to truly understand the words coming before it.

It had to build, in its internal wirings and all its software neurons, some understanding of what an egg is and unicorns. In other words, to get the next word right, it had to become intelligent. It's quite a thought. It started with nothing. We jammed huge oceans of text through it, and it just wired itself into intelligence, just by being trained to do this one stupid thing.

Even as I say it, it sounds kind of crazy, but also kind of beautiful. If this thing actually is intelligent, it got that way from the collective writings of all of us. Yes, Moby Dick, but also some restaurant review you posted in 2004. In some ways, it is all of us. That's too grand, but whatever.

In the end, all the people on the Microsoft campus messing around with that early copy of GPT-4, 14 researchers in all, came to similar conclusions. Even Ronen, who had started out convinced this was just a fancy bullshit machine, he spent a good while holed up with GPT-4 having high level math conversations with it, which is what did it for him.

David Kestenbaum

Sebastien, the unicorn guy, has been going around giving talks about what they did over these months. He titled the talk "First Contact," as in first contact with another intelligence. Only, it's not aliens, it's an intelligence we've made.

I should say, GPT-4 is not good at everything. Like, it's terrible at Tic-Tac-Toe. It often makes very basic arithmetic errors. It told the scientists at Microsoft with complete confidence that there is a McDonald's near gate C2 at the Seattle Airport. There isn't. It's at terminal B. As Peter Lee puts it, "The thing is both smarter and dumber than any person you have ever met."

The 14 researchers wrote up a paper laying out all the experiments they had done. And their conclusion-- that GPT-4 showed sparks of artificial general intelligence. Several researchers who've read it told me, look, to really say something is intelligent and to prove that, there are all kinds of experiments you'd want to do that haven't been done yet.

One AI researcher who'd been in the field a long time told me he felt like this whole approach, next word prediction, is only gonna get you so far. This thing will get better, to a point, maybe not much better than it is now, and then max out.

I first got around this AI stuff and neural nets when I was in grad school for physics when they didn't work terribly well. So when I first started playing around with GPT-4, it gave me shivers down my spine over and over again. Then I went through a stretch of feeling just weirded out. Like, personally, and I've always felt like people can't be more than biological computers, me, everyone I love, all my colleagues at the show here.

But to feel like you are seeing that, a mechanical computer program that can somehow think and talk, it is a little freaky. The place I've settled is somewhere quieter though. It's not boredom exactly. But I don't find myself wanting to go to it very often. Honestly, I don't have a lot of things I need it for.

And then I'll go back to amazement. I can't believe this thing exists. Much has been written about where we go from here, if it's gonna make the world better or worse. I kind of think better. I'm a fan of the steam engine. But really, what is the next word in the sequence? I have no idea.

Ira Glass

David Kestenbaum is our show's senior editor. Coming up, an alien walks into a skate park and tries to let the humans know it comes in peace. That's in a minute from Chicago Public Radio when our program continues.

Ira Glass

It's This American Life. I'm Ira Glass. Each week on the show, of course, we choose a theme, bring you a variety of stories on that theme. Today's show, "Greetings, People of Earth," stories of humans encountering non-human intelligences of various kinds and trying to make sense of them. We've arrived at Act 2 of our program. Act 2, "Meat Cute." Apologies for my voice, by the way. I feel fine.

So in this act, we're going to shift perspectives briefly to the alien's point of view in the whole "Greetings, People of Earth" scenario. And to do that, I realized in putting together today's show that this is the perfect theme to play this short piece of fiction that I read years ago and just loved and it has always stuck with me. Like, I will think about this story now and then, which, like, when does that ever happen, right? I reached out to the author Terry Bisson, who said yes. It's read for us by H. Jon Benjamin and Maeve Higgins.

Ira Glass

Maeve Higgins and H. Jon Benjamin reading a story by Terry Bisson. Maeve has a comedy album called A Very Special Woman that is streaming everywhere for free, just google it. Or you can throw her some money and buy it on Bandcamp. H. Jon Benjamin plays the title characters on the TV shows Bob's Burgers and Archer. And if that were not enough, he is also the Jazz Daredevil with an actual jazz album on Sub Pop Records.

Ira Glass

Act 3, "Yacht Rocked." Recently, massive, mysterious, highly intelligent beings have been going out of their way to contact humans and say, more or less, "Hello, people of Earth," but in a very particular way. And they're doing it over and over. And we humans have had a hard time figuring out exactly what they mean by it. And do they mean us harm? Chris Benderev looked into it.

Chris Benderev

I learned a lot of surprising facts looking into this story. I'll start with a small one. Did you know that yacht owners often need their yacht to get from one country to another, but don't want the hassle of sailing it there? That's where April Boyes comes in. People pay her to sail their boats, which she loves, especially on beautiful days like May 24th was.

Chris Benderev

Half an hour into this, the orcas had disabled the boat's rudder. It couldn't steer. They were stuck. The whales kept circling. And she was struck by how massive each orca was.

[BEEPING]

Chris Benderev

An hour in, April saw that the hull was starting to fill with water. The orcas had torn off the rudder and left this big hole in the bottom of the boat. The crew made a mayday call.

Chris Benderev

Eventually, a rescue boat sidles up to them and brings April and the other crew aboard, tows the boat to land. It doesn't sink. But orcas have sank at least three boats in this area in the past year.

Humans have had a hard time figuring out what to make of these orca attacks. Let me tell you the first theory. It comes from a group of researchers in Spain and Portugal. They say it's possible that the matriarch of this population of killer whales-- they call her White Gladis-- may have suffered a traumatic injury, like a boat collision, and is now attacking other boats and her family is following her lead.

They don't call this vengeance, but the internet definitely does. People love this. It's become a meme. There's merch you can buy to join the orca uprising. I couldn't get an interview with the European researchers, but I talked to three other orca experts. And none of them bought the injured-orca-on-a-rampage theory, like Monica Wieland Shields of the Orca Behavior Institute.

Chris Benderev

Another reason she doesn't buy the theory-- if orcas wanted to wreak vengeance and kill everybody on these boats, they easily could. They kill seals for sport. They prey on great white sharks, rip out their livers, and leave them for dead. And we've given orcas lots of reasons to kill us in the past, like in Monica's home state of Washington when they were taken for captivity.

Chris Benderev

So what does Monica think the orcas are doing? She thinks they're playing. Stealing boat rudders is basically a fad. This has happened in the past. In the late 1980s, orcas in the Pacific Northwest started killing salmon and then wearing them on their heads like hats. They just did it for a year, and then they stopped. The last few years, they've started screwing around with fishermen's crab catching gear, apparently just for fun.

Chris Benderev

The incredible popularity of all this on social media is affecting how people see whales. Monica says in Washington State the last few weeks, people who go whale watching are starting to ask her, is it safe? Another orca expert told me that after a killer whale ran into a ship near Scotland recently, reporters called to ask if the orca uprising against humans is spreading. Are they learning vengeance from the whales of Spain?

The answer is they are not. Orca populations don't socialize with other orcas. They keep to themselves. There's no global conspiracy. April, the sailor, told me she saw on Facebook that boat owners were talking about adding sharp spikes to their rudders to send a more violent message to the orcas. The researchers told me this is what they fear, that the orcas are just playing. They come in peace. But the humans will end up attacking them anyway.

Ira Glass

Chris Benderev is a producer on our show.

Ira Glass

Act 4-- "Jorts and All." So our show today is "Greetings, People of Earth." And I think so many of us, at one point in our lives or another, especially when we're young, feel like we're the alien. Trying to understand and fit in with the humans of this planet. Having experiences that feel very much like "Greetings, People of Earth." Diane Wu spent some time recently with a person who feels that way.

Diane Wu

The person is a teenager, Gwyn. She's 16. And because she's 16, so many things in her life right now feel like a first-contact experience.

Diane Wu

Gwyn, if you can't tell, is very into nature and also can't stand anything that might even have a twinge of cringiness, like McDonald's.

Diane Wu

She lives in Ellensburg, Washington, a small town in the middle of the state, and goes to school with a lot of ranchers' kids and professors' kids. Gwyn's a sophomore and doesn't belong to any particular clique. She told me if the school was a biological cell, she'd be part of the cytoplasm, the shapeless goo that everything else swims in.

But a year or so ago, she started noticing the skateboarders, this group of six or seven boys. She'd see them in the hallways wearing jorts, baggy jean shorts, looking way edgier than everyone else. Her friends showed her videos that the skateboarders posted on Instagram of their tricks, which really captivated her.

Diane Wu

Then last year, Gwyn met someone who knew how to skate, Lauren-- basically the only girl skater at their school-- and asked Lauren to teach her how to do it, too. Gwyn got her own board back in November. This spring, she quit the tennis team to practice skateboarding instead. Every day she goes and does the two tricks she knows, ollies and shove-its, over and over in the church parking lot across the street from her house where no one can see here. The patch that's smooth enough to skate on is tiny, about the size of two parked cars.

There's a real skate park downtown, and Gwyn's been there with Lauren and other friends. But she's never been brave enough to go alone. That's what she wants to do today, to get over this fear so she can go by herself over the summer. Gwyn is still an alien to this new, unfamiliar world. She does not yet speak the language of the skaters or fully know their customs, but she wants to.

Diane Wu

Also, Gwyn really wants to try and make friends with the skater guys. There's one in particular she feels like she might have an in with. His name is Horus. He's a junior. A couple of weeks ago, Horus had apparently noticed Gwyn at school and sent a message to her friend Lauren. He asked, "Who's your friend that was wearing the jorts today?" Lauren wrote back, "Gwyn, why?" And Horus responded, "No reason."

Diane Wu

Whatever it meant, this was mind blowing. A skater noticed her. So if Horus is at the park today, Gwyn also wants to try and talk to him. But since she's the outsider here, this raises another fear.

Diane Wu

It's a Saturday in June, hot. Gwyn's in jorts, a t-shirt with a green heron on it, and sneakers full of holes from skateboarding. So as not to blow up her spot by being a weird adult at the park with her, I pinned a small mic on her t-shirt and we'll watch and listen from the car.

Horus isn't here yet. So Gwyn gets to part one of her mission-- the skating. Gwyn walks up to the edge of the park, and then without making eye contact with anyone, gets on her board and beelines for the far side, putting as much distance between herself and the three other skaters as possible. She does a few of her ollies and shove-its, then ducks behind the half pipe to whisper into the microphone.

Diane Wu

The three skaters, as far as I can tell, have barely registered Gwyn's presence. But that's not what it feels like inside her head.

[SKATEBOARD CLACKING]

Diane Wu

Nobody else seems to notice. Gwyn practices some more tricks, nothing too big. And as soon as the other skaters leave 12 minutes later, she hurries back to the car to check in.

Diane Wu

Solo skateboarding-- done. Gwyn moves on very quickly. Because the next part of the mission, talking to Horus, is looming. Gwyn brings in reinforcements for this step. She texts a couple of friends. And they all show up at the park. It is so hot outside.

Diane Wu

One of Gwyn's friends who shows up, Cooper, is friends with Horus, not from skating. Cooper is a beginner like Gwyn. He has this random update on Horus.

Diane Wu

Cooper pulls up the video and everyone crowds around his phone. It's a Snapchat of two boys strapped into a helicopter flying over mountains.

Diane Wu

Gwyn looks up from the video.

Diane Wu

A talking point, but still no Horus. Almost two hours pass. Then, a big red truck pulls up right in front of me and someone gets out on the other side. I can't see Gwyn anymore or anyone else. But I figure out pretty quickly who it is because here is the next thing Gwyn says.

Diane Wu

Here it is, the sound of a pretty not-awkward conversation between Gwyn and her skater idol.

Diane Wu

Gwyn asks Horus in quick succession about the helicopter, his shoes, Pokemon cards-- which this group of teenagers is surprisingly really into. Everything is going great. Gwyn, who'd been so scared of awkwardness, is now so comfortable that she signals for me to come over from the car.

Horus is compact, tidy, and kind of looks like he could be in a boy band. His hair is bleached light brown. We sit on the curb at the side of the skate park, Gwyn and her friends and Horus and me. Gwyn boldly plugs ahead, asking Horus some kind of interview-sounding questions, confirming his status as a bona fide cool skater.

Diane Wu

Cooper jumps in to tell Horus he should see how good Gwyn's ollie has gotten. The three of them grab their boards and head back into the skate park. And Horus gives Gwyn pointers on how to do a shove-it while moving.

Diane Wu

So the alien arrives and tries desperately to pretend she doesn't have three eyes and a giant head and long, bony fingers. And it works. The humans barely notice. "You want to be one of us," they say. "Put your foot a little higher on the board."

[SKATEBOARD CLACKING]

Ira Glass

Diane Wu is one of the producers of our show.

Ira Glass

Our program was produced today by Chris Benderev with Alaa Mostafa. The people who put together today's show include-- Bim Adewunmi, Elna Baker, Phia Bennin, Sean Cole, Michael Comite, Aviva DeKornfeld, Bethel Habte, Valerie Kipnis, Miki Meek, Stowe Nelson, Sara Parrish, Nadia Reiman, Ryan Rumery, Ike Sriskandarajah, Lilly Sullivan, Frances Swanson, Christopher Swetala, Matt Tierney and Julie Whitaker.

Managing editor-- Sarah Abdurrahman. Our senior editor is David Kestenbaum. Our executive editor is Emanuele Berry. Special Thanks today to Tomer Ullman, Michael Frank, Melanie Mitchell, Jeremy Howard, Jason Stern, and Deborah Giles. Our website-- thisamericanlife.org. This American Life is delivered to public radio stations by PRX, the Public Radio Exchange.

Thanks, as always, to our program's co-founder Mr. Torey Malatia. You know his new passion? Bartering. Bartering. Yes, he doesn't use money, whenever that's possible. I bought him lunch this week. Instead of giving me a cash, he told me this about how he'd pay me back.

Ira Glass

I'm Ira Glass. Back next week, hopefully, with my actual voice and more stories of This American Life.

["HERE COME THE MARTIAN MARTIANS" BY JONATHAN RICHMAN & THE MODERN LOVERS]

(SINGING) And the Martians got notebooks in their little hand. They're still strangers in this land. Martian rhyme. It's Martian, Martian time.