ChatGPT Transforms a Classroom + Is 'M3GAN' Real?

A high school teacher on how the new chatbot from OpenAI is transforming her classroom — for the better. And, “M3GAN” may be closer than you think. Plus: Why Gen Z is chasing the digital camera aesthetic.

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Speaker 00

  Your agents can talk to each other. They still can't think together. That's the gap Outshift by Cisco is closing. Outshift is Cisco's incubation engine, building the internet of cognition. Shared intent, shared memory, and guardrails that allow agents to work together from anywhere. Open source protocols and code. A system for intelligence that scales. Read the paper and experience the demo at outshift.com. That's O-U-T-S-H-I-F-T.com.

• 0:31Kevin Roose?

  Hello, Casey.

• 0:32Casey Newton?

  Hey, Kevin.

• 0:33Kevin Roose?

  We're gonna start the show in just a second, but first, I want to tell our listeners about another New York Times podcast that I think they might enjoy.

• 0:42Casey Newton?

  Why don't you tell me about a podcast?

• 0:44Kevin Roose?

  Sure. So this show is called First Person. It's hosted by Lulu Garcia Navarro, and this week's episode is a sort of hard-forky episode. It's an interview with this guy named Kyle Weens. Do you know Kyle Weens?

• 0:56Casey Newton?

  Yeah, he's the CEO of iFixit, right?

• 0:58Kevin Roose?

  Correct, yes. iFixit, the website that basically teaches you how to fix your gadgets. And he's also become a major figure in this movement that's known as right to repair. You know right to repair?

• 1:11Casey Newton?

  Yeah, you know, it's a basic idea that if you own a gadget, that you should be able to fix it, and it shouldn't be up to the manufacturer of that thing whether you can fix it or not.

• 1:20Kevin Roose?

  Right, which sounds pretty obvious, but has been a topic that has been very contentious in the tech industry for a long time. Lots of companies, including Apple, have fought this sort of right to repair movement. And Kyle has become sort of the leader of that movement. And recently, he got a big victory when New York State passed its new right to repair law, which is the first such law in the country. Basically requires tech companies to make it easier for you to fix the stuff that they sell you, so that you don't have to keep buying new phones and new phones and new phones and filling up landfills. So really interesting conversation, and I think listeners of this show will enjoy it. So go check it out. It's called First Person. You can find it in your podcast app. Let's get started with the show.

• 2:05Kevin Roose?

  I'm Kevin Roos. I'm a tech columnist at The New York Times.

• 2:07Casey Newton?

  I'm Casey Newton from Platformer.

• 2:09Kevin Roose?

  This week on the show, a high school English teacher tells us how ChatGPT has already transformed her school, why Gen Z is obsessed with 20-year-old digital cameras.

• 2:19Casey Newton?

  And our exclusive review of the hit new horror movie and living meme, Megan.

• 2:42Kevin Roose?

  Hello.

• 2:43Speaker 03

  Oh, hi. How are you?

• 2:46Kevin Roose?

  Doing well. How are you?

• 2:47Speaker 03

  I'm very good.

• 2:49Kevin Roose?

  Would you just start by introducing yourself to our listeners?

• 2:52Speaker 03

  Okay. Well, my name is Sheri Shields. I've been a high school English teacher for 30 years. Currently, I'm teaching in Oregon in a little town called Sandy. And I have been at this current school that I'm at. Did you hear my bell ringing? Isn't that nice? I'm actually at my school. There goes the bell. But yeah, and I have been teaching creative writing, college credit English, and currently advanced ninth grade English, which is just basically the regular English, but souped up a little bit. So.

• 3:20Casey Newton?

  Well, the sound of the bell means that class has begun and we are excited to pepper you with questions.

• 3:25Speaker 03

  Good.

• 3:26Kevin Roose?

  Yeah. So we wanted to talk with you, Sheri, because you are using ChatGPT, this new AI tool, in your classroom. And right now, it seems like a lot of the education world is scrambling to try to figure out what to do about tools like these. Some districts, including New York City Public Schools, have banned ChatGPT, saying that it's just a tool for students to cheat on their homework and have the AI write their essays for them. Other schools are trying to sort of adapt and make their curriculum more ChatGPT friendly. So I guess my first question for you is, how did you hear about ChatGPT? And once you learned about it, how long did it take before you started using it in the classroom?

• 4:09Speaker 03

  Oh, a matter of days. I learned about it on a Friday and I was Using it by Monday. Uh, I spent the entire weekend fooling around with it. My son actually brought it to my attention. He's a teacher as well, and he teaches tech at the high school he teaches at. And he says, hey, have you heard about this? And I'm like, I haven't. And so it literally just took a few seconds to get the basics of it. Uh, watched a few videos, but it really didn't take much of a learning curve at all to become pretty good at using it.

• 4:35Casey Newton?

  And was your first thought on using this like this will definitely be useful in my classroom? Or did you have any anxiety about what it might be?

• 4:43Speaker 03

  No, I said, this is going to be amazing. Um, we're going to use this in the classroom. So one of the first things I did was I asked it to write an essay. And to be quite frank, the essay was not very polished. It was rough around the edges. It was very generic. I had to tweak it. I mean, with my advanced questioning skills, I really had to go in there and tweak, tweak, tweak, tweak, tweak to get it to do what I wanted it to do, to present even like a, what I would consider like a, a basic C paper. So most students, I don't know if they would have the skills to go in and get it to write what they need it to write.

• 5:15Casey Newton?

  And what were you trying to get it to write? And, and what, like, how did you tweak it?

• 5:18Speaker 03

  Yeah, let me tell you. So I have two short stories that I like to compare and contrast, which is the story of an hour by Kate Chopin and the yellow wallpaper by Charlotte Perkins Gilman. And they both talk about repression in women's society in the 1880s. And also there's mental health aspects to it. And so what I want was I wanted a three to four page paper, right? And this doesn't produce a three to four page paper. It only produces at most about five paragraphs. I asked it to do a compare contrast on certain things. And I went in and asked it to do it, you know, on how the husbands treated the wives, how societies demands and expect all these different things, expectations were similar or different between the two stories. And I had to get it to write about, I'm going to say eight different essays. Um, in order to get all of the elements in the essay that I wanted it to write.

• 6:03Speaker 03

  So one of the things that we have to do is to teach students questioning strategies. And they're going to have to learn how to go in there and say, well, that didn't produce what I wanted. Now, how am I going to ask it to produce what I want? And that's not really what I'm, you know, that's, that's giving me something I don't want. But what could I do to make it give me something I want? And I'm going to tell you, you'd spend about 30 minutes trying to get it to put out sort of a decent, halfway decent, very short essay. But during that time, their questioning strategies are. You know, they're gonna have to develop them. They're gonna have to know what they're talking about. They're gonna have to ask it questions. You know, they're gonna have to be very specific. So they would have had to pay attention and read both of those stories and listen to. We do close readings in the class and take their notes and go back into the chat and say,

• 6:47Speaker 03

  according to my notes, let's see, this is in there. And then they'd have, you know what I mean? And so they'd have to do a lot of setup in order to get it to even produce an eh essay.

• 6:55Casey Newton?

  Well, so I read the yellow wallpaper in high school, and I imagine I had to write something about it. And to the extent that I learned anything from that process, it was in sort of, I guess, the reading comprehension, understanding what I wrote, trying to maybe synthesize an argument, getting it down on paper. You're talking about a world where we may be moving towards students asking a language model to just kind of make the argument for them. And I have to say, like, I hear that and like my, my ears prick up a little bit. And I say, hmm, like, are we gonna be losing something if the game is to teach students how to ask questions of the AI rather than to synthesize their own argument? So how have you been thinking about that?

• 7:39Speaker 03

  Um, I think about that. Um, when we do the preliminary work for that essay, we do a lot of stuff in the classroom through discussion, through think pair share, through writing short responses by hand and turning them in, and then throwing those out to the class for further discussion. So all of that's still taking place in the classroom, all of those elements. One of the nice things about chat is that it just, um, it'll put out what we've already been talking about. It's like it's not giving them anything new, but it may be organizing it in a way that they can go, oh, I understand. This is a, like, for instance, compare contrast has a very, um, strong organizational method that they have to do. They either have to choose block method or point by point method. And so chat can actually give them an outline

• 8:23Speaker 03

  that will help them with the organizational structure. So it's not really about the arguments maybe so much as it is. Um, this is how I'm going to organize my essay. And I can put what I know from our class discussions and from my notes, that I'm gonna put that into this structure. And so that's what I'm hoping to use it for, is more as a skeleton, more as a scaffold to help them with feedback. That's what I'm hoping for.

• 8:47Casey Newton?

  Mm hmm. And just walk us through how you've used this in your classroom. What does a day's class look like where you and your students are all using ChatGPT together in the classroom?

• 8:58Speaker 03

  Yeah, so since it's so new, this is literally only about the third day, right? So it hasn't been a ton of experience with it. One of the things, one of my freshmen were just working on an essay where we were talking about Martin Luther King's "I Have a Dream" speech. And we were using certain techniques, like he uses repetition and metaphor and simile and descriptive language and all of these sorts of things. And so we already talked about that. The students identified all of that stuff themselves. They underlined we have the text, they annotated it. And then they had to go through and pick another piece by a different person. And they had to do again, another compare and contrast where they said, so which one of these two do you think have the better persuasive element?

• 9:40Speaker 03

  So they have to use their evaluative techniques to say, well, this persuasive technique, this language helped me change my mind about something. You know, I really thought differently about this after I read the speech because of these language techniques. So identifying what the language techniques are is a nice way. So they can say, oh, repetition, metaphor, simile. But then they have to go in and find them. And then they have to go in and do an evaluative process and say, why this one is better than that one. So it's gonna be a step. And the one problem, you know, that you have is, you have to say, you can only use it for this. I'm gonna give you five minutes to generate a list of techniques used in the speech, right? If that's all you want them to do, then that's it. And then they have to close their computers and then they're gonna have to go back to

• 10:25Speaker 03

  paper and pen and they're gonna have to write out the response after they get that list, you know? The other thing it does is, it's really helpful for generating like lists of things, outlines, gosh, oh, prompts, you know? Sometimes students, one of the worst things I have is having a student come up and I'll give them a topic, a general topic, and they'll say, I don't know what to write about. And I'll say, okay, well. So we go through and we have literally 20, 30 minute conversations about what they want, what they wanna do, you know?

• 10:51Casey Newton?

  Yeah, journalists have that problem too, by the way. You know, some, our editors are always asking us what we wanna write about and that's usually when we change the subject.

• 10:58Speaker 03

  And so it's really hard. And sometimes when a student sees it and they can pick it out of a list or they can pick it out of a general set, you know, they can say, oh, I didn't know I could write about that. Or that looks interesting. Maybe I'll give that a try. And so they can tweak it and say, nothing on that list looks good to me. And I'll say, well, what's the least boring thing in this list? Let's make another list, you know? And and so we can generate, gosh, you can generate lists so fast. You can generate information so fast. You can take one kernel of something you find and you can make a whole new list out of that. And then what we have to do as educators, you know, we have to put the emphasis more on the the process of writing. And then, you know, having them write the piece and then thinking about how am I going to have my students write this?

• 11:42Speaker 03

  So they don't have access to chat G P, you know, so they don't have access to the computer while they're creating or drafting. But how can I use this as a stepping point? How can I use this as a stepping stone? I don't know if you are aware of this, but it can also create lesson plans. It generates lesson plans and it can it can evaluate writing. So one of the things I asked it to do last night was I had a student essay and just for grins, I said, evaluate this essay for grammatical and sentence structure. And it did really well. And it said, here's, and it gave him the strengths. It said, here's what you're doing well. And then it said, here are some places to work on, you know, and and it even said stuff like, you know, your transitions aren't very smooth. And your introduction is lacking and there's no thesis in this whole essay. I mean, that's that's one on one feedback. Yeah.

• 12:24Speaker 03

  It would take me about a week to get through, like, I have eighty essays to do at a time. So about a week later, I get to give him feedback, but this is going to give him instant feedback. So I might give him ten minutes to, and I'll give him the prompt. I'll say, have this evaluate your essay for ideas and content or for sentence structure or for organization and then go ahead and take that feedback, that personalized feedback and improve your essay.

• 12:50Casey Newton?

  That's amazing.

• 12:51Speaker 03

  Yeah.

• 12:51Casey Newton?

  I I'm curious how good you thought the lesson plans were that it generated for you.

• 12:55Speaker 03

  Um, not bad. They have to be fairly simple. So I'm a creative writer and one of the things I asked it to do is we're going to start a science fiction unit soon. And I asked it to develop a lesson plan about how to create a cool alien character. Like, how do you go about generating an alien character? And then I put that in there and it spat back out a lesson plan that was more than I expected. And it it went far more into detail than I thought it would. And it really talked about characterization, about description and about all the different methods of how to characterize. And then to apply that to an alien, which would be unusual and interesting, you know? And I was just like, I wasn't, wow, this is not what I was expecting, but I'm I'm all for this. And then I went through and kind of retweaked it for what I needed it, but it was very little tweaking and I was ready to go.

• 13:38Kevin Roose?

  This is all a a very rosy picture of ChatGPT in the classroom. And I actually happen to agree with you. I think there are lots of amazing ways that tools like ChatGPT could be used and are being used in classes. But I want to talk about the other side of the coin, too, because I, you know, we've all been teenagers, we've all been students. And as much as I like to think that I would have only used ChatGPT in the teacher-approved classroom ways if I were a teenager today. I also know that there were days, maybe I hadn't had time to do the reading or I was feeling a little lazy or I just wanted to maybe get a better grade, where I would have used ChatGPT to do my homework and passed it off as my own.

• 14:21Kevin Roose?

  So how worried are you about students using ChatGPT and tools like it to to cheat, to turn in work that they didn't actually spend much time creating? How worried are you about that?

• 14:33Speaker 03

  Um, so not as worried as I've been the last two decades. So being a creative writing teacher, I've had video game stories. I've had the the Late Late Late Show, somebody watched it, decided that I probably wouldn't have watched it and did the the whole story of that and turned it in as their story. She Deviled, you remember that movie, She Deviled?

• 14:52Kevin Roose?

  Yeah.

• 14:52Speaker 03

  I had an entire, this student wrote this whole. Story about basically She Devil and I guess they thought that I'd never seen the movie or would never have seen it. Anyway, I'm reading it and I'm like, this is the plot to She Devil. I'm pretty sure this is She Devil. And, um, you know, and they had just watched that movie and just thought to themselves, okay, Mrs. Schultz would never have seen this movie. Of course I have, you know? Um, even video game plots, they think that, and I play video games. My son and I have both been video gamers for since ever. And they try to pass their stories off, their video game.

• 15:23Casey Newton?

  They try to say, let me tell you a legend about someone named Zelda.

• 15:26Speaker 03

  Exactly. And I'm like, oh, really? Okay.

• 15:28Casey Newton?

  A plumber and his brother.

• 15:29Speaker 03

  Or even anime stuff, which, you know, I, I kind of am familiar with all that stuff. So, um, they, they've been doing that sort of thing. I mean, in creative, in terms of creative writing or taking a poem out of some old book they got off the shelf, thinking I would never have seen it. And it's Dennisin or something, you know what I mean? And it's like, huh, yeah, I think I'm familiar with this one. So the kids have been cheating, uh, kids have been using the internet to cheat. I have seen whole things popped off the internet and, um, copied and pasted and, and kind of reworded, you know, in a way to where I'm still looking at it going, oh, you know, my, and then I'll, I'll even put, all you have to do is put a little piece of it into the internet and you can see it, right? The plagiarism checker, it'll catch, even if it's just part of it. Um, the bad thing about AI is, um, the plagiarism checkers don't catch it

• 16:13Speaker 03

  because it's all generated fresh for each student, each time they, so that's a bad part. Um, there are some new AI generators coming out. Uh, there is one that, um, I guess just came out, but crashed because it was too overused.

• 16:25Casey Newton?

  You're talking about the AIs that try to detect AI-generated writing?

• 16:28Speaker 03

  Mm-hmm. Yeah. Also, a watermark is being developed. So when students copy and paste anything that's AI, it's going to put a watermark. And I don't know what that's going to look like, if it's going to turn it a different color or there's going to be something behind it or over it that says, this is AI. I sure hope so, because then they're going to have to do something in order to get that watermark removed, which is going to be probably rewriting the whole thing in their own words, you know, so.

• 16:50Casey Newton?

  I, I want to ask the sort of, um, a, a question similar to what, what Kevin did, but from another angle, which is maybe you accept that like, okay, by the time that kids get to high school, um, there are a lot of really productive ways to, to use this, uh, technology. But I wonder if you think at a younger age, it might be a little too dangerous to give them technology that can write a five paragraph essay that's better than, you know, 90% of, of nine-year-olds. Is there, do you, do you think about there sort of maybe being like an age threshold that we want kids to reach before we, uh, have the AI start writing their assignments?

• 17:25Speaker 03

  Absolutely. Yes, I I do. I think they should keep it out of the elementary school for sure, out of the even the middle school. Um, and then maybe even the early grades, you know, ninth grade start introducing them to limited bursts where it's all controlled. You know, it's kind of like, hey, um, here's what we're gonna use it for. And I mean, here's the thing, if you don't show them that it can do all these other things. Most ninth graders won't go out of their way to check that out. There's gonna be a few savvy ones that will. But most of them are just gonna go, my teacher just showed me how to do this, so I'm just gonna do that, right? So, uh, it's kind of like if we don't show them, maybe they won't really.

• 18:00Casey Newton?

  Well, I mean, but but just to argue against myself, though, you know, one of the ways that you've been talking about using this technology in your classroom is essentially as a personalized tutor for your students, right? They can show it their work and it can say, hey, you know, you need to work on your transitions or or these other issues. And I think, man, if you're a nine or 10 year old, i- i- and you have a personalized tutor, that's probably really helpful, right? And maybe you want to ask it questions about science, you know, you want to understand the Kreb cycle or how osmosis works. That seems like actually a great thing that we probably would want to have in kids' hands, right? I just get a little nervous when I think about it, auto completing their assignments.

• 18:34Speaker 03

  Yes, I substituted the other day for a Spanish class and they were doing verb types, right? And I'm not really in, I don't know the verb types in Spanish class. And so they're all doing this lesson about um, how to how to like say that you you take a shower, I take a shower, we, you know, that that they, you know, that whole thing. And I didn't know. And so I got this, I said, hey, go to this website. And I didn't say anything about it. I go, it's called ChatGPT. And would you ask it to write a report about how to um, do these verb types in a conversation? And she did. She read the report and she said, okay, I'm good now. And she completed her worksheet. Now, it didn't give her the answer, but it gave her how to do it. And she said that explained it better than than the teacher has ever done. So.

• 19:17Kevin Roose?

  That's amazing.

• 19:19Speaker 03

  Yeah, there you go.

• 19:20Kevin Roose?

  One, one thing I've been thinking about is I'm really glad that this chat GPT tool did not exist during the era of COVID remote learning. Because it seems to me like a lot of the solution to the cheating problem is going to be less at homework and more in class work. Does that strike you as, as being plausible as a way that teachers are going to respond to the threat of people doing their homework with chat GPT is just you're going to do your homework in class instead?

• 19:53Speaker 03

  Yes, and in fact, our department was having this discussion and they said, well, maybe we need to have shorter bursts and so maybe we should move away from the standard essay, you know, as a summative tool to test for whatever. And we should move into shorter bursts of writing. So they might take one main idea, develop a paragraph about that. And so they would write it right there in class and turn it in. And then they might do a connection where the next day they might connect that thought to another thought, you know what I mean? And so it's going to be smaller bursts of writing where they, they utilize the concepts with like a topic sentence and developing supporting details and then transitioning into another, possibly another paragraph at another date.

• 20:35Speaker 03

  So maybe having those smaller skills and then sitting them down at some point and putting all those skills together and into a larger piece of writing, you know, like a full essay. So we were talking about that. We were actually speaking about that.

• 20:49Kevin Roose?

  Are there people at your school who disagree, who think this is too advanced and we need to just ban it entirely?

• 20:55Speaker 03

  The first reaction was, oh, my gosh. Yeah, the first reaction was it can do what now? It can have what now? Yeah. And yeah, well, we spent a long time getting over the shock because I actually went in to our department meeting and said, did you know it can do this? Did you know it can also do this? And we, I spent a good 15 or 20 minutes just listing all the things it could do. And I've been on the, the videos to look, to see about all the different things it can do. Some of it's awesome, some of it is, yes, absolutely, a student could work this out and turn it in. One of the teachers got on there and said, well, I just looked up all my questions for Great Gatsby and it only got two right. And I was like, so sometimes, and this, they've even said it, the information is incorrect. Which means that if you got a page full of answers and two of them are right and the rest are

• 21:40Speaker 03

  wrong, obviously that came from somewhere and it was not correct and the student didn't read the book, you know, so there's that.

• 21:48Kevin Roose?

  That's interesting. It almost makes me wonder if in the future, tests or take home assignments will have to include, you know, like the CAPTCHAs that you have to, you know, to do before you log into certain websites to prove that you're a human. Like it might include, it might include some trap questions that like an AI would get wrong, but a student who had actually done the reading would get right. So I'm, I'm thinking about like, you know, if you asked a question on a midterm or something that was like, you know, explain Jay Gatsby's role in, you know, Tender is the Night, or just some book that he wasn't in. And if you ask that question of an AI model, it might actually respond in some like confident but totally wrong way. Whereas a student who had done the reading would say,

• 22:33Kevin Roose?

  actually, he's not in that book, he's from another book. So do you think you'll have to start including trick questions like that on exams?

• 22:38Casey Newton?

  And would you look forward to sort of being a trickster who's constantly trying to trip your students up?

• 22:43Speaker 03

  You know something about us English teachers, we have had to deal with Cliff Notes for decades. And Cliff Notes has all the answers. And so we've had to design tests and evaluations that doesn't include, we read the Cliff Notes and we read all the Spark Notes online and we go, all right, what, what's not here? What can I put in my test? So we've been, we've been masters of that for quite some time now. So this is just one more thing that we've got to, you know, maybe test around. And I think that's where those in class, I think having, you know, discussions, having presentations, having shorter burst assignments, working with questioning strategies, tweaking things, you know, having students be able to go in and learn how to use this informational technology. Why shouldn't

• 23:28Speaker 03

  we be teaching them how to do this and learn from it and use it correctly? And then have a unit where We take all the technology away and say, now you're going to have to demonstrate that all of this that we've been doing is going to have helped you somehow make connections in your writing and be better at coming up with examples and smoothing out your sentences and things like that, that you've been shown one-on-one with the tutor help of this. Hopefully that'll translate, but we've been hoping for that for, you know, all kinds of different methods for a long time. So we'll just have to see.

• 24:00Casey Newton?

  I'm curious, did you use Cliff Notes?

• 24:02Speaker 03

  No, I read everything. I was a reader. I read everything.

• 24:06Kevin Roose?

  I definitely used Cliff Notes in high school.

• 24:09Casey Newton?

  I did too. And like, as I'm thinking back, I'm sure there were times when it was a substitute for reading at least part of something. Plus, I remember, you know, and English was like one of my favorite classes. I was an English major in college, like mostly as an excuse to read novels. But there's so much reading and trying to keep all of it in your head when you're sort of like coming up on a big midterm is tough. And so having someone who has sort of already taken the notes for you and can sort of jog your memory about, you know, these are sort of the major themes, these are the major characters, and having that available at a glance, I didn't think of it as cheating so much as like a resource to use me as I like tried to make it through the class. And I'm wondering if, you know, in short order, we'll come to think of something like ChatGPT the same way.

• 24:48Speaker 03

  And I said that exactly when we were having a discussion in my department. I said, try not to use the word cheating because the teacher was like, well, they're gonna have all these ways to cheat. And I said, maybe substitute the word cheating for assistance. You know, they're gonna have more assistance in their corner. You know what I mean? They're gonna be able to have. And so maybe we just need to re-look at that word exactly and say, this is assistance. This isn't necessarily cheating. And maybe what that's gonna look like is gonna be different in the future because already, students can just whip out their phones. Every time I teach a class, I get at least four or five people who whip out their phones and verify what I just said. And then I get the little hand that goes up, yep, she's right. Or, you know what, Michelle, you forgot something. This is also true. And I'm like, you know, I've been dealing.

• 25:31Casey Newton?

  Oh, that sounds so annoying. Send that kid to the principal's office.

• 25:34Speaker 03

  Every single class period. Yeah, no way, man. They verify everything that I say. And it's already, it's already like that. They've already accepted that the internet is, you know, all-knowing and all-powerful and whatever it says is true and right. Whatever Michelle says, it's gotta be verified.

• 25:50Casey Newton?

  You know, speaking of this tool as an assistant, I wonder if you work with students who are English language learners and if you've sort of considered the promise that a technology like this might have for those folks.

• 26:01Speaker 03

  Yeah, I know it speaks different languages, you know, and in terms of being able to translate questions and instructions, I mean, that's really helpful. If you just put the instructions in and then ask it to translate it to English, of course, that would be extremely helpful and it's going to be right there, the exact thing that we're working on in class. If I give a handout and they put that handout into the chat and then translate it into Spanish, they'll have both things right in front of them. I can't see how that would be a bad thing. I mean, you know, because then they'll have both things. Hopefully they don't become dependent on that and then don't move over into learning English, but just keep translating everything into their original language. I think that's really helpful. One of the things I was telling you about, when you tweak the chat and you

• 26:45Speaker 03

  tweak something that's written, it can say rewrite this, the above essay, rewrite the above piece into a sixth grade level. So if it's like sitting at a 12th grade level and you ask it to revise it into a sixth grade level, that would be really helpful for language learners to be able to modify those pieces of reading that might be too difficult for them right then and there so that they can understand the gist of it, use different vocabulary words, and then they can see, oh, this word means this and this word over here in this advanced one means that, and these are the two differences between these two words, but I understand this one.

• 27:19Casey Newton?

  Yeah, there's a lot of technology that we would love to understand at even a sixth grade level that we're not quite there about. We talk about quantum computing sometimes on this show and we don't know what it is, but we're confident that maybe someday ChatGPT will tell us.

• 27:31Kevin Roose?

  I need that at like a second or third grade level, though. Sherry, I'm with you on all of the potential classroom uses for this. I'm so glad that there are teachers out there like you who are helping students understand these systems because I do think that they will need to be able to work with and around these generative AI models as adults. And so it makes a lot of sense to start them on that now. But I do want to just sort of sound a note of, I don't know if it's like melancholy. Colly here, or just, you know, I'm remembering back to my high school English classes, which were some of the most transformative classes, you know, of my life, made me want to become a writer. And part of what I loved about English class was just being bad.

• 28:14Kevin Roose?

  Like that, that kind of like, you know, you write something, a first draft, it's bad. Your teacher, you know, helps you polish it or reshape it. They explain something that maybe you didn't overlook. There's a certain value, I think, in struggle and improvement. And having to kind of do all of that manually yourself and kind of the mental processes that that triggers. You know, I don't think that ChatGPT should be banned in schools, but I'm also trying to acknowledge the fact that it does feel different to give, you know, a computer a prompt and have it write your essay, rather than you sitting there and slaving over every word and every sentence.

• 28:55Kevin Roose?

  And, you know, through that process, kind of gradually becoming a stronger and stronger writer. So I guess I'm curious what your thoughts are about that. And sort of when we gain all of these new powers with ChatGPT, is there anything that students are at risk of losing?

• 29:11Speaker 03

  Oh, yeah. I mean, so I have two different classes. One is like the standard English class with a lot of essay writing. The other one's creative writing. So we do poems, we do short stories, we do all kinds of fun, creative writing projects. And one thing chat doesn't do is it doesn't do creative very well. The poems are bland and flat, the writing lacks description. So I'm not very worried about students turning in this AI-generated creative stuff, because it's, it's going to be very telling. And in fact, they're not going to be likely to turn in because it's horrible stuff. Most students who are in my creative writing classes really want to learn how to be descriptive and how to develop characters and setting. And so that's what we spend our time on. One of the things I like to do is sit in class and write that out. And then we share with each other.

• 29:56Speaker 03

  And then they take those pieces of writing and they expand on them, just like you said. And we talk about how to develop sentences and how to make characters more rounded. And that's a really fun class. I don't see that this program's really going to interfere with the creative type class. So I'm not too worried about the creative nature, especially when kids want to be creative. You know, some of the kids that they're just in the class and they just need to get these poems out so that they can pass the class, they may use it to write some pretty terrible poems and then try to turn them in. You know, I guess. It's like anything else. They can go and find an old literary magazine and copy those poems and try to turn them in. So I think it's all out there. All the ways

• 30:37Speaker 03

  that we have that students can take that stuff, the students who want to do it, I think it's up to the teacher to inspire them to do it. And that's where good teaching comes in. Good qual-- like you were saying, those classes that inspired you, that's going to be where you have discussions with your teachers and your fellow students. And, you know, you lead the class and everyone's still talking about what you were talking about in class. Those are the moments that you live for. And that's what you have to focus on as a teacher, are those moments that we have with our students in our classes that inspire them to make and write better than what like an AI can just generate for us. And so that's why I like AI just for the bare bones, you know, the scaffolding part of it. For me, that's really important.

• 31:19Kevin Roose?

  Sheri, Ms. Shields, thank you so much for joining us. Is there anything else we should talk about?

• 31:25Casey Newton?

  Well, I just want to say, if you're planning on cheating in Mrs. Shields' class this year, don't, because it will get back to us and we will talk about it.

• 31:32Speaker 03

  Oh, good. Okay, I hope so. I hope so.

• 31:35Kevin Roose?

  Yeah, let us be your enforcement arm.

• 31:36Speaker 03

  Yeah, I will, because, yeah, definitely. Well, as this process grows, I may change my mind. I will say that, like, I don't want them to ever ban it out of my school because it would be a huge resource loss to me, you know, just things that I have been starting to use it for. I'm starting to become sort of a chat fiend, you know, like, wait, let me go to chat. Like, my old sources are like old hat now. And I'm like, no, let's see what chat says. And I'm starting to really start to rely on it and look at it.

• 32:07Kevin Roose?

  Thank you so much for spending your period with us, and we'll let you get back to class now.

• 32:11Speaker 03

  Period seven. It's not quite over. I'm gonna walk in there. They're all gonna go, Ms. Shields! And I'm gonna be like, yeah, yeah, hi. I'm excited to see him. Yeah.

• 32:21Kevin Roose?

  All right. Well, thank you for your time. We really appreciate it.

• 32:23Casey Newton?

  Thank you.

• 32:28Kevin Roose?

  When we come back from the break, New York Times reporter Callie Huang will tell us why teenagers may be hitting you up for your old digital cameras.

• 32:51Speaker 00

  Your agents can talk to each other. They still can't think together. That's the gap Outshift by Cisco is closing. Outshift is Cisco's incubation engine, building the Internet of Cognition, shared intent, shared memory, and guardrails that allow agents to work together from anywhere. Open source protocols and code, a system for intelligence that scales. Read the paper and experience the demo at outshift.com. That's O-U-T-S-H-I-F-T dot com.

• 33:23Speaker 04

  Your sales team is at risk of missing quota. Don't panic. Just ask Rippling AI. Built on your real-time people and business data, Rippling AI can pull metrics from Rippling and your CRM into a meeting-ready dashboard showing quota attainment, headcount trajectory, and revenue to quota by region. In seconds, you'll see what's behind your quota risk and fix it before it's missed. Head to rippling.ai/hardfork to get the only AI built to give you full visibility and take complex actions across your whole business. That's R-I-P-P-L-I-N-G dot AI slash hardfork. Sign up today.

• 33:57Speaker 05

  Today's AI builders need infrastructure that scales with their ambitions. Iren is a vertically-integrated AI cloud platform, delivering data centers, compute, and software for AI training and inference, underpinned by its expansive portfolio of land and grid-connected power in renewable-rich regions globally. Iren's AI cloud gives direct access to Nvidia GPUs, scalable from hundreds to thousands. For builders, that means direct access to AI compute, rapid scaling, and predictable pricing. Iren, AI cloud for the next big thing.

• 34:32Kevin Roose?

  All right, all right. Are we ready? We're rolling here. Casey can confirm red showing. Red confirmed. Twelve plus hours of recording time? That's great. You brought lunch, right? We're doing 12 hours.

• 34:44Speaker 06

  Forget my deadline.

• 34:46Kevin Roose?

  Beautiful. Kelly Huang, my colleague at The New York Times, welcome to Hard Fork.

• 34:51Speaker 06

  Thanks for having me.

• 34:52Kevin Roose?

  So you had a story, it was titled "The hottest Gen Z gadget is a 20-year-old digital camera." And I saw this story everywhere. It was in all of my group chats, all of my elder millennial friends were texting each other about it, saying like, "I gotta dig the camera out of the old junk drawer and sell it to some zoomer for many times what it's worth." It really seemed to touch a nerve. And I'm also curious, because you are our Hard Fork unofficial Gen Z correspondent.

• 35:30Speaker 06

  Yeah.

• 35:30Kevin Roose?

  How old are you?

• 35:31Speaker 06

  I'm 22.

• 35:32Kevin Roose?

  22, wonderful. So how did you notice that this was a trend?

• 35:37Speaker 06

  It started on Instagram, where a lot of things for young people start. I saw a lot of people I know posting photos that were obviously not taken with an iPhone. Okay, weird, people being quirky. And then I saw it being done by the biggest influencers, people like Kylie Jenner, who I think set the tone for a lot of people. And yeah, I was out at bars and I went to parties where people were bringing around digital cameras. And on TikTok, the hashtag has hundreds of millions of views, I think.

• 36:16Kevin Roose?

  And what is the hashtag?

• 36:17Speaker 06

  Literally just "digital camera."

• 36:19Kevin Roose?

  Wait, are people actually taking video with these things? Or they're just, they're posting bad photos on TikTok as videos?

• 36:25Speaker 06

  Yeah, people are on TikTok saying, "Here's how you get this new aesthetic."

• 36:32Speaker 06

  And it's a digital camera from when they were like four years old.

• 36:36Kevin Roose?

  And these are not, to be clear, these are not like high-end SLR digital cameras. These are like the Nikon Coolpix with one megapixel from like 2003, right?

• 36:47Speaker 06

  Yeah, the Sony Cybershot or the Nikon Coolpix. The worse quality, the better. We're not going for anything professional here.

• 36:56Kevin Roose?

  It's hard to think of two more 2000s sounding names than Cybershot and Coolpix.

• 37:03Kevin Roose?

  Yeah. As I was reading your story, I was thinking, A, do I have any digital cameras collecting dust in my closet somewhere that I could dust off and resell on eBay? And which it sounds like people are doing. There was some data in your story about how the searches for, you know, words like Nikon Coolpix are like way up year over year. And my second thought was like, how much of this is sincere, right? Because like, I think every kid, every teen, every young person goes through a phase where they get like into something that is out of step with their generation. So for example, when I was in my late teens, I thought it would be cool to like buy a typewriter. And I started, I can't believe I'm admitting this on this podcast.

• 37:45Kevin Roose?

  There was a period of maybe six months where I would like bring out my typewriter and like type a note to someone and like give it to them.

• 37:54Casey Newton?

  Wait, what kind of notes?

• 37:55Kevin Roose?

  You know, like, uh, you know, would you like to go to the dining hall? And, and get some chicken fingers with me?

• 38:01Casey Newton?

  And you would sort of like deliver these in what, like an embossed stationery or?

• 38:05Kevin Roose?

  Yeah, I was very annoying. I would just slip them beneath people's doors. But it was, it's not a period I'm particularly proud of. But, you know, this lasts for six months and eventually it's like kind of annoying because I have to like refill the ink and like, there's only one supplier that has the ink for this kind of typewriter. And I realized like, oh yeah, email was invented for a reason. Text messages were invented for a reason. This is way worse than just doing the thing that everyone else does. So how much of this is kind of like zoomers being like into the act of taking digital photos and actually the aesthetic of it? Or how much of it is just kind of novelty value and kind of retro Y2K nostalgia?

• 38:44Speaker 06

  I think a lot of it is the novelty. Like it feels refreshing to like have to take a camera out with you instead of just having everything on your phone. And I think a lot of people are encountering the, this is annoying. I have to put ink into my typewriter in that like, some people don't know how to upload photos onto their phone. I mean, if you look at a TikTok video featuring this, people are like, how do I get it on my camera roll? Because it is sort of like a foreign item.

• 39:10Casey Newton?

  You need like a laptop, right? Like you need like an SD card. You gotta hope you have a computer that has the right slot for an SD card on it. Yeah.

• 39:16Speaker 06

  It's really complicated. And like, a lot of today's laptops like don't even have a slot for an SD card, right? So it's sort of like, for some people, the complications that you have to go through to get to. An objectively bad photo are kind of fun.

• 39:33Casey Newton?

  Right. And then there is like sort of the vinyl comparison, right? Where like vinyl in a lot of ways is a bigger hassle than any other way you could listen to music. But that becomes part of the charm. So the digital cameras mostly are much worse than the cameras that we have today. But like, what else defines this aesthetic? And is there anything else about the aesthetic that you think is appealing to younger people?

• 39:52Speaker 06

  I think part of it is that you have a lot less control over the photo that you're producing. I think part of it is also this call back to the Y2K era. When you think of someone like, I don't know, Paris Hilton or whoever was popular back then, like, it makes you feel cool when you are like emulating them. And I think part of it is sort of the like performance of being casual and looking silly or like effortlessly pretty or whatever. And being like, I don't care that I'm posting these on Instagram when I'm blurry, I'm like washed out. Maybe it's a little unflattering. I think that performance of authenticity is like a big appeal for some people.

• 40:33Casey Newton?

  And I feel like we've seen this before on Instagram, right? Where people will start to embrace a look that is not perfectly airbrushed, right? There's sort of been wave after wave of people rebelling against the sort of beauty standards on Instagram. It seemed like there's just kind of a fundamental tension on that platform where people are tired of having to dress up so much.

• 40:53Speaker 06

  Yeah, I think there's a tension in particular among young people to decide like how curated you want your online personality to be, how to portray the level of curation that you are seeking. Like, it's an intentional choice to like post a photo where you look bad and washed out. You know, like Paris Hilton was not picking which photos the paparazzi publish, but like Kylie Jenner is picking which washed out photo she's posting.

• 41:22Casey Newton?

  Yeah, it's interesting. A few weeks ago on the show, we were talking about this app, Lensa, which a lot of people were using to create these magic avatars, the app calls them. And on that app, you know, you upload a few higher quality photos of yourself and you come back looking like an astronaut, a wizard, a god, right? And it's sort of the opposite of what we're talking about here, right?

• 41:41Speaker 06

  It sort of strikes me as like also wanting a time when social media didn't exist or wasn't as pervasive as it is now. So I think part of it is like wanting to return to, I heard a lot of like wanting to return to simpler things.

• 41:57Kevin Roose?

  Yeah, this this sort of neo-Luddite sort of feeling. I mean, there was a great story in The Times a few weeks ago about about these teens who just reject all technology. I think they're in Brooklyn, obviously. But it's like, it does feel like there's a sort of niche cohort of young people who are just saying like, screw all of this. Like, I'm gonna read books and go outside and take walks and use my cool pics to capture my moments with my friends.

• 42:22Casey Newton?

  You know, I don't, I sort of have the opposite perspective, where I think all of this is just a way to be on social media and look slightly different and feel like you've adopted an aesthetic that is just a bit ahead of the curve. So to me, this feels like just as plugged into social media as anything else we talk about on the show.

• 42:38Kevin Roose?

  That's true, actually, because it's like, it's not like people are nostalgic for the distribution of the early, you know, no one's like posting these on Flickr.

• 42:45Casey Newton?

  Right.

• 42:46Kevin Roose?

  Right? So that you're taking your photo with your old-school digital camera, but you're posting it on TikTok and Instagram. And I don't know if it's, there's a way to do it on BeReal or not, but like, it is not true that people are nostalgic for the actual ways that we consumed information back then. It's just the camera.

• 43:05Casey Newton?

  You know, you brought up BeReal, and to me, BeReal is kind of a way of doing this, too, right? Where the whole idea is you're gonna take this photo at a random time each day, you're probably not gonna look that great, you're probably not doing something that interesting, and so it may come across as more authentic.

• 43:21Speaker 06

  I mean, people are not doing photo dumps in Facebook albums, right? Like-

• 43:26Kevin Roose?

  Thank God.

• 43:28Speaker 06

  Right. No one's like drunkenly or like hungover posting like 30 photos from a night out on Facebook. So I think part of this is the presentation of authenticity. Part of it is certainly, this is cool, interesting, maybe you're not looking at your phone as much because you don't need to to take photos. But I think the flip side of it is that it is still going online and you are making- A choice of how you take the photo, which photo you're putting online, what kind of appearance you're projecting.

• 44:03Kevin Roose?

  Mm hmm.

• 44:04Casey Newton?

  Here's my prediction, that if this thing sticks around for like a few more weeks, I think it just sort of becomes a filter on Snapchat and Instagram, right? And like, here's your sort of, you know, 2000s cool pics filter. And then we can start leaving the cameras at home again.

• 44:17Kevin Roose?

  But I think there is something, I can imagine it being very cool to like be in a bar, talking with your friends and then just like pull out this like hunk of plastic from the early 2000s.

• 44:28Casey Newton?

  I mean, has it really been so long since any of us have seen a camera? That this would be an event if one came out of a purse? I just can't believe that.

• 44:36Kevin Roose?

  I don't know. I mean, it sounds like it's, it's sort of a thing to pull it out when you're out with your friends.

• 44:41Speaker 06

  I think it makes people excited, like you're so used to just like someone taking a photo with you. I mean, this makes me sound like someone who just came out of the womb. But I think it makes people excited to like have an actual camera because it's more of like an event. Yeah, I mean, I think you see the same thing happen with film cameras. Like Instagram certainly has filters that you can use that make it look like you took something on film. VSCO does, but there's something about the actual act of like taking out the camera, I think, that is appealing to people.

• 45:20Casey Newton?

  Let's see, what will we be nostalgic for in 20 years? Like, what's something that's like sort of painful that our grandchildren will scramble to do again?

• 45:28Kevin Roose?

  Hmm. Callie, any thoughts?

• 45:29Speaker 06

  I would not be surprised if it were something like an Apple Watch or like something involving the way that we listen to music. I think a lot of this comes for like the weird little gadgets that we use now. I don't know. I mean, it's hard to imagine like a futuristic way of communication, but I bet our iPhones will look like Brex.

• 45:55Kevin Roose?

  Yeah, there'll just be like some hipster teens like carrying around their iPhone 13 Pros like. Showing it off to all their friends.

• 46:02Casey Newton?

  You know, if you sort of, not to bring up the metaverse, but if you believe in a world where we are wearing some sort of virtual reality headsets or maybe using virtual displays inside of those headsets, I can actually see laptops becoming kind of a nostalgia item, right? Where people are walking around with laptops. Or maybe laptops will still be around, but people will sort of want them to be styled like laptops in the early 2020s or the 2010s.

• 46:26Speaker 06

  Yeah, I mean, Kevin used typewriters. Maybe the people 20 years from now will be using laptops.

• 46:31Casey Newton?

  I can't wait till my grandson asks me for a 2013 MacBook Air.

• 46:37Kevin Roose?

  Kelly, thank you.

• 46:38Speaker 06

  Thank you.

• 46:47Casey Newton?

  And when we come back from the break, Meghan.

• 46:54Speaker 07

  Companies are pouring money into AI, but few leaders can explain how those AI investments work with the systems you already have. You know, the ones that actually run your business? Without Systems, you don't have to start over with AI. You can choose the projects where AI will have the most impact. And Outsystems gives you the platform to make it agentic on your terms with complete control every step of the way. So you can trust AI with the mission-critical systems that run your business. Own your agentic future. Learn more at outsystems.com/NYT. That's outsystems.com/NYT.

• 47:29Speaker 04

  Your sales team is at risk of missing quota. Don't panic, just ask Rippling AI. Built on your real-time people and business data, Rippling AI can pull metrics from Rippling and your CRM into a meeting-ready dashboard showing quota attainment, headcount trajectory, and revenue to quota by region. In seconds, you'll see what's behind your quota risk and fix it before it's missed. Head to rippling.ai/hardfork to get the only AI built to give you full visibility and take complex actions across your whole business. That's R-I-P-P-L-I-N-G dot AI slash hardfork. Sign up today.

• 48:03Speaker 05

  Today's AI builders need infrastructure that scales with their ambitions. Irene is a vertically integrated AI cloud platform, delivering data centers, compute, and software for AI training and inference, underpinned by its expansive portfolio of land and grid-connected power in renewable-rich regions globally. Irene's AI cloud gives direct access to Nvidia GPUs, scalable from hundreds to thousands. For builders, that means direct access to AI compute, rapid scaling and predictable pricing. Iron, AI Cloud for the next big thing.

• 48:39Casey Newton?

  All right, Kevin, it's time to talk about my favorite movie-going experience of the year so far, Megan.

• 48:44Kevin Roose?

  Megan.

• 48:44Casey Newton?

  I would say the premise of Megan is that a young girl is given a sentient robot doll by her aunt, who is the doll's inventor. And the movie is about what happens when the child begins to depend more heavily on the doll and the doll develops some ideas of its own. And you also saw Megan last night.

• 49:07Kevin Roose?

  Yes.

• 49:07Casey Newton?

  What was your feeling about this movie?

• 49:10Kevin Roose?

  So I am not a scary movie guy, like at all. I am a huge baby. I don't even really like cliffhanger sort of action movies because they make me too anxious. And I've got enough of that in my life. So I was really resenting the fact that you were making me watch this movie to talk about it on the podcast. But I actually, I enjoyed it. I thought it was fun. And I will say, I am a terrible person to watch a movie about anything tech-related with because like, I just couldn't stop thinking about like AI and robotics and which parts were realistic and which parts weren't.

• 49:46Casey Newton?

  Were you furiously taking notes on your typewriter about this?

• 49:52Kevin Roose?

  No, but I was taking notes. So what did you think?

• 49:56Casey Newton?

  Well, so listen, if you show a gay man a life-sized doll who dances and kills people, you've got a great movie. Well, you can stop writing the screenplay right there. I have all that I need. So I really enjoyed the movie. It's extremely fun. It's funny. It really is a comedy. Like, yes, it's a horror movie, but it's more of a comedy than anything else.

• 50:15Kevin Roose?

  Yeah, I was expecting to be very scared and I was, I was not. It's campy enough that it, it makes the scary parts less scary.

• 50:22Casey Newton?

  Right. But so why are we talking about it on Hard Fork? This is not a movie review podcast, but this movie is about a lot of things that we've been talking about lately. So it's about artificial intelligence. It's about what happens when people begin relying on artificial intelligence over human beings, right? And it's about fears around screen time, about parenting, right? There's a surprising amount of the modern condition, I thought, in this movie. And I think, you know, if you listen to this podcast, you're probably going to find something to think about in Megan.

• 50:54Kevin Roose?

  Totally. I mean, I think, you know, as a relatively new parent, it, I think, gets to the heart of this question of like, what amount of technology is good for kids and when does it start becoming a problem? I thought it was a little heavy-handed. There's literally a scene where they're like talking about like, you know, screen time and how much the little girl should get and it's decided that like Megan is a good form of technological interaction and healthy and that obviously, you know, gets complicated later on. But I will say, I think it really does illustrate just how anxious people are when they give their kids technology. Not necessarily that it's going to melt their brains, but that the technology itself is just going to like, take on capabilities we didn't know it had.

• 51:36Casey Newton?

  Yeah, I mean, so I think at their best, horror movies take some sort of fear in the culture and they make it tangible by turning it into a monster and then having the monster kill people. And in a real way, like, screen time is the monster in Megan, right? There is a girl who might have otherwise been interacting with her parents who, you know, I think we can say are tragically killed, like the opening seconds of the movie. She could be interacting with her aunt, who has sort of taken her in and become her guardian, but her aunt is very busy at work and her aunt never stops looking at her phone either. And so what does the little girl turn to? She turns to screen time and that screen time later goes on a homicidal rampage.

• 52:15Kevin Roose?

  Yeah, I thought it was actually interesting. The other movie I've watched in the last couple of weeks is Glass Onion, the new Knives Out movie, in which the tech world, I would say, is not portrayed positively. There's sort of like an evil, like Elon Musk figure at the center of that plot. And then in Megan, we have this kind of, it's not really the central part of the plot, but it is a part of it where this doll, Megan, is actually like being sort of rushed into production by this, I guess, toy company, but sort of tech company, you know, the office looks very Silicon Valley, like it's, you know, it's run by people who like talk the way that tech people talk. And it's all about like, speed and how fast can we ship this thing?

• 52:58Kevin Roose?

  And like, we're not going to program any safeguards into it because we're just going to like, throw it out to the world and like, who cares if- If it misfires sometimes, like that's the price of progress. And it's just such a, we seem to be in such a moment right now of sort of cultural backlash to the way that tech companies have been operating.

• 53:13Casey Newton?

  Oh, yeah. As a Silicon Valley satire, I found it surprisingly on point. There's a very funny moment where the doll begins talking to the people who built her when she's not supposed to be. And one of the engineers says, like, didn't you encode like parental controls? And then, you know, the protagonist is like, well, I didn't have any time, you know? And like, the number of times that we've heard that from tech executives over the years made that, I think, a really satisfying moment to watch.

• 53:38Kevin Roose?

  Totally. And another question I was finding myself thinking about, again, I'm a horrible person to go to a movie with, but I was thinking about just how, how close or far we are in real life from the technologies that we saw in the movie.

• 53:51Casey Newton?

  Yeah, well, so, you know, there was an article in the New York Times recently about scientists who are trying to do just this, who are trying to make robots that exhibit some signs of consciousness. And while I don't think anything is quite to the level of Megan yet, particularly in terms of like interpersonal relations, it's clear that we're able to develop devices that are at least in some ways self-aware, have some concept of themselves, and we assume that's only going to accelerate quickly.

• 54:17Kevin Roose?

  Totally. I was actually impressed. They must have had some AI experts consulting

• 54:22Casey Newton?

  Yeah.

• 54:23Kevin Roose?

  in the writing of the script for this movie because they did use phrases like probabilistic inference to describe how Megan learns, which is like a real thing that exists in AI.

• 54:35Casey Newton?

  Yeah. And we want to say, good job, Megan producers. You did your homework.

• 54:40Kevin Roose?

  If you were the AI consultant for Megan, please come on Hard Fork and tell us about your process. I was thinking, as I watched the movie, about Moravec's paradox.

• 54:50Casey Newton?

  No, I don't know what this paradox is.

• 54:51Kevin Roose?

  Okay, Moravec's paradox, it's a sort of well-known principle in artificial intelligence and robotics. And it basically says that things that we, that humans find very easy are very hard for machines to do and vice versa. Things that are very hard for humans sometimes are very easy for machines to do. So like a classic example would be like predict which of these. Ten loans is most likely to default. Like, that's, that's kind of hard for a human, pretty easy for a machine. On the other hand, something like move this cup from one part of the table over to another part of the table, which like any human toddler can do, is actually incredibly hard for a robot.

• 55:34Kevin Roose?

  So there are all these moments during the movie where I was thinking like, so one, one thing that Meghan does is like, explain condensation, the concept of condensation to the little girl. And this is not a spoiler, by the way, this is not a major plot point in the movie. But I was thinking like, oh, that's, that's easy. Like Siri can do that, like, you know, Chat GPT can do that. That's not hard at all. And then you have these scenes where Meghan is dancing or moving in some lifelike way. And I'm thinking like, oh, that's going to take 20 more years and several billion dollars more of R and D before robots can do that.

• 56:06Casey Newton?

  Yeah, the, the fluidity of Meghan's movements is maybe one of the less realistic things in the movie. But, you know, I, I do think that when it comes to, will we be able to use a Chat GPT like tool in some sort of doll and that doll has a voice that is fairly human sounding and maybe has some emotion in it? That doesn't feel all that far away at all. You know, there's this kind of secondary toy in the movie that's like the, the kind of, the, the bridge, the first thing that the inventor builds before she invents Meghan. And it's just like this little fuzzy, like furry creature that like talks. And it's like, well, we're like basically already there.

• 56:45Kevin Roose?

  Yeah. And I also think it's realistic to think that when these dolls with sort of AI built into them exist, like children will love them. When I was growing up, I had this doll called Teddy Ruxpin. Do you remember Teddy Ruxpin? Who was like a, just a teddy bear, but you were like-

• 57:01Casey Newton?

  By the way, I've heard Gen Z is taking Teddy Ruxpin out to bars just to impress their friends, but go on.

• 57:07Kevin Roose?

  They're selling for thousands of dollars a piece. So Teddy Ruxpin, for those of you who don't know, was basically a teddy bear that you would like put a cassette into and it would read you a story and the mouth would move a little bit. And it was like not advanced by any modern measure, but I freaking love Teddy Ruxpin. I was so attached and when it would run out of batteries, I would cry and like- That was almost 30 years ago. So I can only imagine what today's kids are going to think when things like ChatGPT and these large language models are starting to be built into the toys that they use every day.

• 57:42Casey Newton?

  Yeah, and it just becomes really powerful. You know, it's like as powerful as ChatGPT is as a text interface, it's like, you know, you put it into like the shape of a doll and make it talk. It starts to feel like something very different. And all of this AI stuff we've been talking about so much just becomes like kind of infrastructure for a brand new set of products that might kill us.

• 57:58Kevin Roose?

  Totally. And they also could help us. I mean, one of the things that Meghan actually does in the movie is to basically befriend this little girl, hear her problems, basically act as a therapist.

• 58:12Casey Newton?

  Yeah.

• 58:13Kevin Roose?

  And I think that's a question that is sort of on my mind is like, when this kind of thing actually exists, when it's capable of not only like talking at kids, but responding to them, will parents let it into their lives and to what extent?

• 58:27Casey Newton?

  Yeah.

• 58:28Kevin Roose?

  It really also, the parts in the movie that were the scariest to me from like an AI perspective are when Meghan starts displaying capabilities that she never had before, when she's sort of learning on her own. And I really feel that, like, this is a very different example, but like, sometimes my Alexa devices in my house will, like, I'll ask it a question and then it'll answer and then it'll say, by the way, did you know I can also help you store recipes?

• 58:57Casey Newton?

  If you're the Amazon people working on this, you have to stop. When I ask the weather, don't tell me that it's time to shop for Father's Day gifts, okay?

• 59:04Kevin Roose?

  It's really freaky. And like, that's a pretty mundane example, but it's like, I don't want you to get smarter.

• 59:10Casey Newton?

  Exactly.

• 59:11Kevin Roose?

  Like, I don't, I don't, I bought you because you do kitchen timers and you, you know, you tell me the weather and whether I need to bring an umbrella that day. And like, that is what I need you for. I don't want you getting smarter. And I think we're moving into an era where everything in our lives, from our cars to our kitchen appliances, to our, you know, large language models, is just going to be getting better all the time in the background. And I think that experience is going to freak people out more than the actual capabilities. It's, it's the improvement in the background when we didn't ask for it that's going to be a lot of people's sort of first scary moment with this stuff.

• 59:47Casey Newton?

  Yeah, pretty soon you're going to open up your refrigerator and it's going to say, you know, I could teach you Spanish. And you're just going to unplug it. You're going to say, I don't want this in my life. We need to go back to basics. Go get my Coolpix. I'm out of here.

• 59:59Kevin Roose?

  I also thought it was interesting, I mean, without giving away too much of the plot, one of the sort of themes of Megan, and I think this goes, you know, not back, not just to other movies about sentient robots, but all the way back to Frankenstein, is this idea that you can instruct a robot to do one thing, like protect this little girl, and it will take that instruction very seriously and will accomplish it in ways that maybe you didn't intend or want. So in AI research, this is called the paperclip problem sometimes. Have you heard about the paperclip?

• 1:00:35Casey Newton?

  I've heard about the paperclip problem, but tell our listeners.

• 1:00:37Kevin Roose?

  So the paperclip problem is this thought experiment that was proposed decades ago by Nick Bostrom, this philosopher. And it's basically says that, you know, you could build an AI and tell it to make paperclips. That's its only instruction. And if you give it no further instructions, it will do that. And it will do that. It will, you know, first, it will use all of the metal in the world to create paperclips. So it will take it from factories. It will destroy things. It will destroy cars to get the metal to make paperclips. And then it will effectively kill all the humans on the planet to keep them from fighting their acquisition of metal to make paperclips. And so this robot that you just told to make paperclips ends up destroying the world.

• 1:01:21Kevin Roose?

  And there's a similar plotline in Megan where this robot's only instruction is to take care of this little girl, and it does so in increasingly violent and scary ways.

• 1:01:32Casey Newton?

  Yeah. And, you know, God, I don't even know what I have to say about that other than the paperclip problem seems unsolved, man.

• 1:01:41Kevin Roose?

  Well, I am glad that this kind of thing is making its way into pop culture. Obviously, it's ridiculous. Obviously, you know, we are a long way away from killer humanoid robots. But I do think this is the moment to start thinking more broadly as a culture about AI. And I think one of the reasons this is really striking a nerve right now is that there is so much anxiety about these tools that just seem to be appearing from these tech companies out of nowhere. With sort of very rudimentary safeguards in place. And it just feels like a moment where something is being let out of a bag that might be hard to get back.

• 1:02:15Casey Newton?

  Yeah, also, it was like about a month ago that the San Francisco Board of Supervisors was considering letting armed robots kill crime suspects. And this was not a sci fi story. This was just up for debate at my local city council was whether we should use murder robots in police work. So, you know, some of this stuff is very sci fi, but I think what is most interesting about it is that in a lot of ways, it doesn't really feel that sci fi.

• 1:02:41Kevin Roose?

  Right, no, a lot of this stuff is technically possible. It exists like the Boston Dynamics robots that we've seen videos of like, there was this one scene in particular where someone starts pushing Meghan with like this big jousting stick almost just like, like trying to knock her off balance and see how she recovers. And like, that's almost shot for shot from one of these Boston Dynamics videos where the researchers are like poking these robots with sticks, trying to throw them off balance. Basically, all you need is for someone to like, take the large language models and like jam them into the Boston Dynamics robots. And then you essentially have Meghan.

• 1:03:14Casey Newton?

  Yeah, and by the way, there's a moment.

• 1:03:16Kevin Roose?

  Please don't do that, by the way.

• 1:03:19Kevin Roose?

  Very bad idea.

• 1:03:20Casey Newton?

  The moment when Meghan gets down on all fours and like runs through the forest and looks like a Boston Dynamics dog is like one of the greatest transformations in cinema history, as far as I'm concerned.

• 1:03:30Kevin Roose?

  Yeah, that was terrifying.

• 1:03:33Kevin Roose?

  Yeah, I feel like sci fi movies used to take place like 40, 50, 100 years into the future. And now like Meghan just seems like, yeah, that could happen next year. Google or some other company could come out and say like, this is our new, I mean, Tesla is literally building a humanoid robot, right? That's like one of Elon Musk's pet projects is like, he wants to make a lifelike android robot that can talk and listen and do therapy and move in realistic ways and do all these things. He's building Meghan.

• 1:04:05Casey Newton?

  Yeah, and he must be stopped.

• 1:04:25Speaker 05

  You

• 1:04:38Speaker 07

  Companies are pouring money into AI, but few leaders can explain how those AI investments work with the systems you already have. You know, the ones that actually run your business? Without Systems, you don't have to start over with AI. You can choose the projects where AI will have the most impact, and OutSystems gives you the platform to make it agentic on your terms, with complete control every step of the way. So you can trust AI with the mission-critical systems that run your business. Own your agentic future. Learn more at outsystems.com/NYT. That's outsystems.com/NYT.

• 1:05:13Speaker 04

  Your sales team is at risk of missing quota. Don't panic. Just ask Rippling AI. Built on your real-time people and business data, Rippling AI can pull metrics from Rippling and your CRM into a meeting-ready dashboard showing quota attainment, headcount trajectory, and revenue to quota by region. In seconds, you'll see what's behind your quota risk and fix it before it's missed. Head to rippling.ai/hardfork to get the only AI built to give you full visibility and take complex actions across your whole business. That's R-I-P-P-L-I-N-G dot AI slash hardfork. Sign up today.

• 1:05:47Speaker 05

  Today's AI builders need infrastructure that scales with their ambitions. iRen is a vertically-integrated AI cloud platform, delivering data centers, compute, and software for AI training and inference, underpinned by its expansive portfolio of land and grid-connected power in renewable-rich regions globally. iRen's AI Cloud gives direct access to Nvidia GPUs, scalable from hundreds to thousands. For builders, that means direct access to AI compute, rapid scaling, and predictable pricing. iRen: AI Cloud for the next big thing.

• 1:06:22Casey Newton?

  Hard Fork is produced by Davis Land. We're edited by Paula Schuman. This episode was fact-checked by Caitlin Love. Today's show was engineered by Alyssa Moxley. Original music by Dan Powell, Alicia B. Etup and Marian Lozano. Special thanks to Hannah Ingber, Nell Galogly, Kate LaPreste, Jeffrey Miranda, Daniel Bartel and Dan Savage. He gave us a nice shout-out this week in his Savage Love column. He tried to write his column with ChatGPT. It's a really fun read. Check it out. Thanks, Dan. We love you. You can email us at [email protected]. That's all for this week. We'll see you next time.

• 1:06:58Kevin Roose?

  See you down the dusty trail.

• 1:07:00Casey Newton?

  I want to come up with something else, but I know you told me to stop using it. And I tried to think of an alternative one, and I couldn't think of one.

• 1:07:08Kevin Roose?

  I'm sorry I brought that stress to your life. I love your dusty trail, and I love you. And I'm sorry I criticized you.

• 1:07:16Casey Newton?

  I thought about saying something like, "That's all for this week. The solution to today's hurdle was Zidane."

• 1:07:56Speaker 00

  AI isn't just opening new business opportunities, it's changing the way businesses get work done, the questions they ask, the decisions they make, and the routine tasks they can hand off. And nobody understands that better than NetSuite. That's why they put together a free expert guide you need to grab today. Aligning for the agentic era: How AI is Changing Everyday Work. Download it and you'll see exactly how AI is reshaping work across every part of your organization and how to make sure you come out ahead. Get your free guide at netsuite.com/NYT. netsuite.com/NYT.