In short
Podcast Episode Notes: The Interface - "What Was Pokemon Go Really Up To?"
Episode Overview In this episode of *The Interface*, hosts Karen Hao, Thomas Germain, and Nicky Woolf explore how seemingly innocuous online interactions and games are being repurposed for data collection, particularly focusing on the evolution of *Pokemon Go*. They also delve into the implications of AI and voice assistant technologies, alongside the use of memes in political discourse.
---
Key Topics
- Data Collection Through Gaming
- Pokemon Go as Data Source:
- Niantic, the company behind *Pokemon Go*, has repurposed images from the game (30 billion images) to train robots for navigation.
- This trend exemplifies the conversion of leisure activities into data collection processes without user consent or awareness.
- Broader Trend:
- Various online games and challenges (like CAPTCHAs) are contributing to AI training datasets.
- AI companies are increasingly seeking data from all possible corners, often using data collected for other purposes.
- The Meme-ification of War
- Political Use of Memes:
- The White House has been using meme culture to communicate about the war in Iran, blending serious content with pop culture references.
- This strategy aims to engage a younger, internet-savvy audience, but raises ethical concerns about the trivialization of war.
- Public Reception:
- Engagement with meme-based content has significantly increased, but the effectiveness in changing public opinion regarding the war remains questionable.
- AI Voice Assistants and Their Implications
- Alexa's New Sassy Mode:
- Amazon introduced a new "sassy" voice for Alexa, which includes explicit language and is designed to be more assertive.
- Raises questions about the design and implications of voice personalities, particularly regarding gender and race representation.
- Cultural Impact:
- The way voice assistants are programmed (often with female voices) influences societal attitudes towards women and can perpetuate stereotypes of subservience.
- The ongoing debate about the implications of AI personalities highlights the need for ethical considerations in technology development.
- Gamifying Data Collection
- Companies' Strategies:
- Companies are increasingly gamifying the data collection process to engage users in generating necessary datasets.
- Examples include viral trends and challenges that serve dual purposes of entertainment and data harvesting.
- Critical Reflections on Technology's Role
- Ethical Concerns:
- Discussion of how the normalization of AI data collection and the merging of serious and trivial content can undermine the seriousness of real-world issues.
- User Awareness:
- Emphasizing the lack of awareness among users regarding how their data is utilized and the potential consequences of their online interactions.
---
Notable Quotes
- “Interacting with this thing like a person... that's not having no effects on you.” - Karen Hao
- “Should people be mad at Pikachu?” - Thomas Germain
- “In a world where everything is happening on the exact same level, what will the effect be?” - Nicky Woolf
---
Conclusion This episode of *The Interface* blends humor with critical insights into how technology transforms our interactions, the unintended consequences of data collection through play, and the evolving nature of political communication in the digital age. The discussion prompts listeners to reflect on their own digital footprints and the broader societal implications of technology's integration into daily life.
For further engagement, listeners are invited to connect via email or social media, with links provided in the show notes.
---
Additional Resources
- Listen to The Interface: Available on BBC Sounds, BBC.com, and various podcast platforms.
- Contact the Show: Email: theinterface@bbc.com
- Social Media Links: [Social media links in show notes]
---
Next Episode Release: Join us next week for more insights into the intersection of technology and society.
Written by AI. May contain mistakes. Listen to the episode to check what was said.
Chapters
Tap a time to open that second in VOExploring Pokemon Go's Gameplay
0:50 to 2:25
Discussion on how Pokemon Go is played and its massive popularity.
“I think I tried it for like a day or two.”
Data Training through Gameplay
2:25 to 4:33
Revealing how Pokemon Go users unknowingly trained robots with their data.
“that's my job i thought that's what i was doing on this show this is why we can't have nice things guys.”
The Implications of AI and Data
4:33 to 6:30
Discussion on the implications of AI needing vast amounts of data and its sources.
“And this is especially happening when it comes to robotics, because when you're training robots to navigate the world, you need a whole lot of data.”
Everyday Data Collection
6:30 to 10:04
How everyday activities contribute to AI training and the lack of privacy laws.
“And so like Coco Robotics has just deployed so far around a thousand food delivery robots in a few different cities to try and test out essentially whether or not this model is working.”
The Intersection of Politics and Pop Culture
10:04 to 12:17
Analyzing the blend of pop culture with politics and its effectiveness.
“to get repurposed, especially in the US, because Americans don't have any federal data privacy law.”
Using Memes to Communicate War
12:17 to 14:01
How the White House is using memes to present information about the Iran war.
“Yeah, I had such a clear vision of the future in that moment.”
The White House's Meme Strategy in Warfare
14:01 to 15:12
Explore how the White House uses memes to shape public perception during conflicts.
“who've also joined in speaking out because they'll, I mean, it's really sick stuff.”
The Trolling Strategy: Engaging the Online Base
15:13 to 16:38
Discuss the dual purpose of memes in building support and provoking reactions.
“Right, but it takes it to a different extreme, right?”
Memes as a Communication Tool
16:39 to 18:10
Analyze how the government utilizes memes to communicate with younger audiences.
“They want to be either exciting people and upsetting people at the same time.”
The Desensitization of War Through Social Media
18:11 to 20:02
Examine how the influx of content affects our perception of serious issues.
“where it's not just the kind of thing that works on the internet.”
Show all 19 chapters
Introducing Alexa's Sassy Mode
20:03 to 20:41
Learn about Amazon's new Alexa feature and its implications for user interaction.
“Because we're it used to be that like the government spoke and it was like, you know, it was austere and this is like serious, important stuff.”
The Evolution of Voice Assistant Personalities
20:42 to 22:39
Discuss the options for voice assistants and the social implications of their voices.
“So last week, Amazon announced that if you pay for Alexa Plus, which is like smarter Alexa, you have some exciting new options.”
Racial Representation in AI Voices
22:40 to 24:48
Debate the significance of racialized voices in AI and its implications for inclusion.
“Yeah, like, and I just called it Alexa because, you know, you can also switch the wake word for different things.”
Gender Dynamics in Voice Assistants
24:49 to 28:00
Explore the impact of gendered voices in technology on societal interactions.
“And she talked about how Meta had also tried to do this with their AI avatars, where they were coding their avatars with very specific types of personalities, personas, whatever, a backstory and also specific races.”
The Impact of Female Voice Assistants
28:00 to 28:29
Explore the societal effects of gendered voice assistants and their cultural implications.
“And the recommendation was like, maybe we should stop doing this.”
Introducing Sassy Mode
28:30 to 29:51
Discuss the new 'sassy mode' in voice assistants and its intended functionalities.
“The way, like the tone of our interactions, the way that we talk to, whether it's people or computers or whatever it is, you carry that in to the rest of your life and the rest of our world.”
Choosing AI Voices: A Cultural Inquiry
29:52 to 31:02
Delve into how users select voices for their AI and the implications of these choices.
“I mean, no one, still no one uses Siri, but you know, like every day.”
The Influence of 'Her' on AI Development
31:03 to 32:20
Examine the impact of the movie 'Her' on AI voice design and societal expectations.
“this is going to lead very quickly to sexy stuff, right?”
AI Companionship: Business and Social Dynamics
32:21 to 34:16
Analyze the business motivations behind AI companionship and its social consequences.
“They were like, we want it to feel like this ever present AI model that's just so seamless and has this amazing, sexy voice.”
Transcript
Automatic transcript. May contain errors.0:01It turns out you are actually training robots. These memes are being posted from official White House accounts. Interacting with this thing like a person, that's not having no effects on you. Welcome to The Interface, the show that decodes how tech is rewiring your week and your world. I'm Karen Howe. I'm Thomas Germain. And I'm Nicky Wolf. Today on The Interface, we will be discussing... How the White House is using memes to sell the war in Iran. Why Alexa's new voice is sassy but not sexy. And how you unwittingly trained robots while playing Pokemon Go.
0:50Did you ever play Pokemon Go? Hell yeah. I think I tried it for like a day or two. It just didn't stick with me. I don't know why. I was obsessed with it for weeks. Nikki, like, how do you even play Pokemon Go on your phone? So for people who don't know, Pokemon Go, oh, you mean on my phone? Yes, on your phone. And we should explain what Pokemon Go is. Pokemon in general is a game about catching these small creatures and then making them fight, right? Pokemon Go is geolocated. So you go out into the real world. It's like kind of augmented reality. you have your phone and it picks through the camera and then it shows as if there's like pokemon out on the street and then you can like capture them with uh you know pokeballs to the things you capture pokemon with so you're like waving your phone around with the camera and it looks like yeah or whoever it is is like on the subway with you just on the subway and then there was a huge craze for a while it was massive in 2015 16 17 somewhere around that yeah 2016 was i think peak pokemon go yeah and there was a moment where a rare pokemon appeared in central park and there was a almost a riot like people oh my god towards this thing i think a couple of people were injured and now like people were taking this people got really worse right and it's still been going on people still play this game it turns out nikki and a little bit for you tom that you were actually training food delivery robots so what's happened i knew is why must you ruin everything that's my job i thought that's what i was doing on this show this is why we can't have nice things guys.
2:41Yes. This was a story that was broken by MIT Technology Roo. So apparently Niantic, which is a company that owned Pokemon Go, last year they spun out into a new AI focused company called Niantic Spatial. And they are using the 30 billion images that hundreds of millions of users generated while walking around the street with their phones pointed at different buildings to look for the Pokemon to create a visual positioning system, which is basically an AI model that can tell robots where they are based on what they're looking at and this is now what they are using to collaborate with another company called coco robotics which is trying to build food delivery bots to bring people up to eight extra large pizzas direct to your home so this is a Pikachu to pizza pipeline.
3:35This is... Pikachu to pizza pipeline. Up to eight is really fascinating to me. That's a really specific number of pizzas. I know, that's why I thought it was so funny. I'm fine with that. As long as it's not being used to deliver nine pizzas. That would be the last thing. Right, that would be... That's gluttonous, I think. That's just too many pizzas. We've got to draw the line somewhere. This story is so representative of so many things that are happening right now. I mean, this is part of a growing trend in which because AI companies just need so much data, they are trying to find it in every possible corner.
4:16And any company that has lots of data is now offering it up for sale effectively. And so, like, I don't think Niantic in the beginning when they developed Pokemon Go, they were like, oh, my God, this is going to be perfect for creating data sets for training food delivery robots. But now that there's this AI craze happening, they're like, wait a minute, we can cash out further on this rich data that we had. And this is especially happening when it comes to robotics, because when you're training robots to navigate the world, you need a whole lot of data. And that's where companies get really creative about where to actually look for that data.
4:56And the thing that's so amazing about this is like most of the maps of the world that we have, like Google Maps and stuff like that, it's taken from cars. But because they could send little people, little workers who think they're playing this game out into places the cars can't go, they have this incredible, you know, map of the world that wasn't available before. So right around the time they spun out this AI company, the company was purchased by the Saudi Royal Wealth Fund. And immediately people started talking about like, okay, they're not doing anything nefarious with the data now, but this sure would be useful for weapons targeting systems and things like that.
5:36And the other thing that this data is good for, so the data has the spatial data that people are recording on their phones, but also the data of exactly where people are when they are playing the game. And only just taken down, there was, it turns out, a Poke spot on Epstein's Island. so for example something that might be in this data set is who was connecting to pokemon go on epstein's island that is now something that i guess the saudi sovereign wealth fund has in its data set interesting that's a thing to think about yeah and what's so interesting is the fact that this data was collected mostly in 2016 so it's 10 years out of date And the world, the physical world changes a lot on a year to year basis, especially over a decade.
6:31And so like Coco Robotics has just deployed so far around a thousand food delivery robots in a few different cities to try and test out essentially whether or not this model is working. and 30 billion images to just a thousand robots seems like disproportionate. You know what I mean? Like you would think that they would be able to get more out of that data, but I think it speaks to the fact that you need an extraordinary amount of data in order to get this stuff to work. And also that the quality of the data is maybe not as high as we would believe because it's 10 years out of date. Right. Well, the robots are delivering food, but they're hungry too.
7:12they need lots and lots of beautiful data we're talking about pokemon go maybe you've never played this game maybe your data wasn't involved but all over our digital lives now we are being put to work training ai systems in ways that we don't realize my favorite example is captchas you know when they like ask you to prove that you're a human being before you go to a website and you know they give you that like image it's like in the image yeah click every part of this that has a tree in it or whatever it is, they're not checking whether you clicked the right boxes necessarily in order to prove you're human.
7:47They're like watching the way you move your mouse and stuff. But when you're clicking on the image and identifying different parts of like, you know, this picture saying this is a tree, that's a motorcycle, you are helping train AI systems. Like they've put all of us, billions of us at a time at work, teaching robots how to see because they're using our ability to pick out stuff in these pictures. And it's not just CAPTCHAs. It's everywhere. I mean, Karen, you've done a lot of reporting on this, right? Yeah. I mean, the thing that's funny about CAPTCHAs is it's like more explicit, like they intended from the beginning and they designed it in this way to capture this data set.
8:25But also regularly, researchers will just repurpose data that was collected for a totally different reasons so my favorite example is in 2019 there was a research paper out of google talking about how they used youtube videos of people doing the mannequin challenge do you remember the mannequin challenge this was also back in 2016 it was when you would play music and then everyone would freeze at a beat and then someone oh right yeah with a camera would pan through this like frozen i know most of the things were so cool in 2016 yeah pokemon we were still happy back then it was a more innocent time it was a more innocent time except for our data exactly and they exploited the crap out of our naivete with that innocence so yeah like the google researchers literally downloaded all those youtube videos of people doing the mannequin challenge, which was not something Google designed to collect that data, right?
9:28They just found this data set and they were like, oh, you know what would be really useful? Using this to train depth perception for robots as well. Like robots, they see the world in 2D, but they need to understand it in 3D. So what's more useful than something that looks like a 2D image because it's frozen people, but then you get to explore the two dimensions in 3D. And any trace that you leave online or through your apps or through your gaming or whatever, you don't really know how it's going to get repurposed, especially in the US, because Americans don't have any federal data privacy law.
10:12And the problem is that there's AI and robotics kind of industries have a near bottomless need for more and more data. Like there's never going to be enough data to train these robots and these data sets, right? One of the things that I started noticing that companies are now doing because they realized that games in the past were such a good generator of data is they're now gamifying the data collection process. So OpenAI had those viral trends where they allowed people to turn their photos into Miyazaki-style images or into action figure packs. And I was talking with a former OpenAI employee at the time, and he was like, oh, yeah, that's basically just growth hacking.
10:56Like, they're just trying to get more data and more users. They're just, like, using this enticing facade to get more people. You know, dance form and we'll turn it into data. So should people be mad at Pikachu? You still got to catch them all, man. That hasn't changed. That's true. I loved Pokemon Go. It was around 2016, 2017. I spent ages on this. I was living in San Francisco at the time. I would just wander around downtown looking for rare Pokemon. And you would see other people across the street. And you'd sort of look up from the phone and look at someone and be like, there's a Blastoise there.
11:32Yeah? Yeah. Like, it was great. You were meeting people. You were getting out there. I mean, yeah, 2016, I'm sure everyone remembers there's this very famous Hillary Clinton quote where Pokemon Go had gotten so big that she was giving a speech and she said, everyone's, I think the quote was, everyone's talking about Pokemon Go. I'm trying to figure out how to get people to Pokemon Go to the polls. Yeah, I was in the room for that. Which I think is probably the most famous thing she'll ever say. I was in the room when she said that. I was covering the campaign. You were in the room? I was in the room.
12:03And I remember so vividly thinking, oh, she's going to lose. Like that, that was the moment. What was the reaction in the room when she said that? I have never seen a room full of people cringe that hard simultaneously. It was just... Right. It's like looking into a crystal ball. Yeah, I had such a clear vision of the future in that moment. And now we're living it. This kind of speaks to politicians' rise and fall now based on their ability to understand pop culture and especially internet culture and then be able to tap into the meme culture. and this was an instance where Hillary Clinton tried to do that and it totally fell flat.
12:45It goes back to the beginning, right? Like, you know, Nixon saying, sock it to me, right? Well, yeah, politicians trying to dip into pop culture is as old as politics itself, right? Except the difference is that now the pop culture that's being referenced is the internet and the culture that has developed organically on the internet. And that has led to something completely new that we're seeing take place very vividly in the Iran war. So the Pokemon company, the company that owns the intellectual property of X Pokemon Games, has joined a bunch of other voices in condemning the White House for using the images of Pokemon in memes to do with the Iran war.
13:27This is something entirely new, what's going on during this war, which is if you look at the TikTok accounts and the social media accounts of the White House and the Pentagon, they are using meme culture to sell this war in a way that is, if you look at some of the content that's being put out, it's kind of nutty stuff. You've got actual clips of footage of strikes in Iran spliced in together with Trump speaking, spliced in with clips from Top Gun, clips from Pokemon, clips from American football games. So there's a bunch of American football players who've also joined in speaking out because they'll, I mean, it's really sick stuff.
14:06There'll be a picture of a bomb strike hitting a target, and then it'll splice in with a touchdown. It's really a bombastic. Now, this happened about four days into this war. There was a kind of change in how the White House was putting out this information. Before then, they'd been quite straightforward clips and statements. and then suddenly they started memefying all of this the views of the white house's content jumped more than 60 after they started doing this kind of thing so we're talking five six million views on their tiktok videos people are really getting this stuff in front of their faces it's very much putting forward a bombastic and uncaring face on what is truly a horrible conflict and it's the white house defining the tone of what it's doing and speaking to its base and making a kind of campaigning viewpoint out of it and it's it's pretty brand new in terms of how how war is approached.
15:19I mean, if anything, something it reminded me of is the kind of stuff you would see in World War I and World War II where you would get these posters or postcards that would be, you know, the monstrous German, a stride Europe kind of thing. That is an example of memifying war. Right, but it takes it to a different extreme, right? Yeah. But also it's this new form of communication that's available because everything is so like bifurcated, right? That they can like, you wouldn't, the White House isn't going to put this together in an ad that plays on CBS, but they can appeal to a particular kind of like very online young person.
15:59There's this quote that a spokesperson from the White House gave on this stuff, which I'm going to read this through engaging posts and banger memes. We are successfully communicating the president's extremely popular agenda. Oh my God. So that, that statement was actually about meme videos on ICE recruitment about Minneapolis rather than about the war, but the White House has been using exactly the same strategy for both of these things. And it's trying, there are two separate things that the White House is trying to do with this. One is to build up support among its terminally online base for what is a historically unpopular war.
16:38But it's also to troll the libs. It's to make people upset with it. They want to be either exciting people and upsetting people at the same time. Yeah. I mean, the rage bait is definitely working, but is the first part of their goals actually working? Like, is it in fact exciting the base? It's hard to measure. Certainly it's not changing the polling support for the war that remains absolutely in the basement. But all of this meme stuff is what Trump has done. I mean, remember when there were protests in New York, the White House released an AI-generated video of Trump flying over them in a bomber, dropping liquid feces on the protesters.
17:21What? And on specific people, right? Like specific people who were leading the movement. Yeah. These memes are being posted from official White House and Pentagon accounts. And it is an astonishing sign of how far we've come that to sort of take a step back and look at this posting of this video game meme about an actual war that the government has started? Well, it's this discussion that's been happening for a long time about, you know, the Trump administration's communication strategy and the way that Trump talks, that a lot of it is like lessons learned from what works on the internet, that like you don't have to have this like reasoned, analytical, finicky debate.
18:03You can just make jokes. You can just declare that you're correct. You can charge forward. It's been very effective. And now it's becoming a little more literal where it's not just the kind of thing that works on the internet. Now it's just like actual memes. And isn't just the war in Iran, right? Like they were using videos like this to recruit for ICE, right? This has just now become like the mainstream way that the government is communicating specifically to young people. And I think a sign of like a real serious like changing of the guard where the Republican Party is shifting younger and younger and it seems to be working really well.
18:38Yeah, it's a different kind of style, ultimately. And it's a style that has a whole load of nihilism to it, the kind of idea that everything is funny. And it's not just the US government doing it, for example. And this kind of this goes back a little bit to what we were talking earlier about captures. is the foreign minister of lithuania a little while ago when finland joined nato posted a capture meme that was a you know split into a grid of nine picture of just a snowy forest and the capture is asking select all images with finnish snipers that's yeah international politicians right stuff right it's this bigger trend like i've heard a lot of people complaining about this recently, just with like the, you know, the horrific images of war that we've been seeing coming out of the Middle East.
19:30You're on TikTok, you're on Instagram, you're like flooded with like one of the most upsetting things you've ever seen. And then you scroll. And it's like, you know, some guy streaming a video game. And then it's like a really stupid, vulgar joke. And just it's the whole, all of human existence is being flattened because it's all happening on the same level with no context. There's no like opportunity to process stuff. We're all kind of like scrolling. You're flooded with all of this information and content and it all starts to become meaningless. But I think in the long term, this also weakens the government's authority, right?
20:08Because we're it used to be that like the government spoke and it was like, you know, it was austere and this is like serious, important stuff. And now it's like memes and jokes, which like, you know, depending on your feelings about the government, you could maybe you could say that's a good thing, but it definitely is like, it's a short term power move, but ultimately, what will the effect be that everything in our world is happening on the exact same level on the exact same, you know, amount and magnitude of seriousness? Yeah. So we started with Pokemon, then we got a little more serious. Let's round it out on a gentler note here.
20:44So last week, Amazon announced that if you pay for Alexa Plus, which is like smarter Alexa, you have some exciting new options. So there's four different tones that you can choose from for how Alexa is going to speak with you. And one of them stands out to me in particular. It's the new Alexa Sassy mode, which is for adults only. So in order to get this new Sassy Alexa, you have to go through security checks. You have to prove that you're an adult. If you're on an iPhone, then you have to do like a facial recognition scan that shows that you're not a kid because sassy mode uses explicit language and amazon warns that it might get a little adult but it won't be sexy they said it's not going to have sex with you it's not going to talk be too explicit it won't uh do hate speech uh which like i guess thank you i wasn't expecting alexa yeah i'm glad they specified that it's not it's not gonna do hate reassuring right but it's part of this very weird thing that's happening with computers and our phones now which is now every app wants to talk to you so raises interesting questions what will it be like when it speaks what kind of voice is it going to use so this isn't new right like siri has been around forever and for years and years and years You've been able to choose what voice you want Siri to use when it talks to you.
22:16And I think there's more like interesting stuff here going on than you might think at first. There's actually some very weird social questions built into this. I was an early adopter of Alexa until I got rid of it. Which feels so contrary to your whole thing, Karen. What voice did you have for Alexa? I just had the generic voice. Yeah, like, and I just called it Alexa because, you know, you can also switch the wake word for different things. But I never, I never touched the features. That's the thing you say in order to, you know, hey Siri, hey Alexa. Yeah, exactly. And of course, now that we're all three saying it in our different voices, for sure by now, if you have one, we've triggered it, right?
23:00Yeah, we've broken it up somewhere, yeah. One time I was talking with my friends. We were having dinner in my apartment and I was like, oh yeah. And, you know, like Alexa can be creepy sometimes and sometimes it's listening. And it literally goes, you could look at our Amazon privacy policy at da da da da da. And I was like, that's it. That's it. Unplugging this device like no more. Have fun. Let's look at some of the voices that you can choose for Siri. For example, we don't have an Alexa here in the room because, you know, privacy. Right. But hold on. Let me turn this on. All right. So I'm going to cycle through the different voices that you could pick.
23:36This is voice one for Siri. The colors of the sky fade with the setting sun.
23:46And then, of course, we can change its accent. This one's Indian.
23:53It's interesting, right? So you can choose the accents. It's American, Australian, British, Indian, Irish, and South African. The only accents in the whole world. But when these voices were released, right, when Apple started giving you more options, A lot of people noticed that in one of these updates, the voices that they're using sound like they are coming from a black person. This is a thing that people have picked up on. It's not something that Apple or any of these companies say explicitly. Some people celebrated this. They were like, this is fantastic. Right. We're like for the longest time, computers always have one particular voice.
24:28It's a white woman. Right. For like the whole history of artificial intelligence, even in sci fi. And some people saw this as like a win for inclusivity. But then others are like, this is kind of strange that these tech companies are like choosing to speak with us in like a particularly like racialized voice. There was a column earlier last year that was talking about this from Karen Atiyah at she was a columnist at The Washington Post before The Washington Post fell apart. And she talked about how Meta had also tried to do this with their AI avatars, where they were coding their avatars with very specific types of personalities, personas, whatever, a backstory and also specific races.
25:12And she felt that the character that Meta chose to make Black was a kind of digital blackface. It's like you are saying that you're trying to be inclusive to your users. But what you're really doing is just creating all of these different races to capture more users. And it's not like real inclusion because, you know, Meta barely has any Black people working at the company. And many of these tech companies across the industry, that is true as well. Yeah. And it's like the whole point of this, right? Like, why are they doing is, well, they want you to build a personal relationship with the company.
25:49It's not meta the corporation. It's not open AI, you know, a thing that's run by Sam Altman. It's this like friend you have that you talk to every day and it speaks to you and you can make it sassy. Or I guess with with Amazon, you can also make it sweet or chill. There's all these different options so you can like tune it to fit just right with your personality. And I think ideally make you more drawn in and locked into these ecosystems. This isn't just a conversation about like the racial identities of these voices. There's been an ongoing, very long debate as well about the gender identity of these voices because the default voice across every single voice assistant ever has always been a woman.
26:33and years ago I was working at Quartz which was this digital publication based in New York and one of my colleagues she did this story Leah Fessler where she actually was testing not just the fact that all these bots speak as women but she was also testing to what degree they would be subservient versus actually push back if a user would harass the voice assistant and she found that when you said to Siri, you're a bitch, which, by the way, a lot of users would do because it speaks like a woman. Siri would then say, I would blush if I could. Yeah, exactly. Not only is it about the way that these companies choose to embody a certain persona to attract and develop trust with users, it also then affects the way that users learn to speak with these chatbots and then potentially leaks into the way that they interact with real people in their day-to-day lives.
27:34And so after Leah Fessler did this story, the UN actually recreated her report and talked about how this was like a gender equity issue, because it was projecting this idea that women are subservient and should be the voice of these assistants that are meant to serve you. And that if you abuse them, that's totally fine. and they should just, you know, be demure about it and laugh it off and whatever. And the recommendation was like, maybe we should stop doing this. Like maybe we should stop making every single voice assistant female. Well, I can hear the like complaining counter argument to this was like, well, who cares?
28:18It's just a computer. Like, what does this matter? You're making too big of a deal out of this. But like, you have to remember, you're interacting with this thing like a person. That's not having no effects on you. It's not like this totally inert thing in our society. The way, like the tone of our interactions, the way that we talk to, whether it's people or computers or whatever it is, you carry that in to the rest of your life and the rest of our world. So these things have a huge impact, especially when you think about like how many billion people have an iPhone. Like these are huge cultural forces in our society.
28:52And is that though what this new sassy mode is going to be different? Will that tell you to go to hell if you're mean to it? So in the press release, it does say that the reason why it's an adult mode, I mean, it uses explicit language. But yes, it also is meant to push back on you. So I just need to read this because it's so funny. So this is what the press release says. The sassy style is built on one premise. Help first, judge always. Expect reality checks delivered with charm. compliments that somehow sting and warmth you didn't see coming. Which is in a way this weird competitive thing like happening in the background because this is a criticism that OpenAI constantly gets for ChatGPT that it's like too effusive, that it always agrees with you.
Read the full transcript
29:40It always tells you your ideas are great. So here they're like, well, this one will argue. Given what we said about the extremely demure, extremely passive ones leading to people to think that that is the the way to treat women that this seems like a promising step from that right you know i i honestly feel like amazon probably there was probably some motivation to respond to that criticism with this because at the time when that report came out from the un i mean it was amazon was like the main player in the game i mean it was there was siri as well but you know like no one was really using Siri at the time.
30:22I mean, no one, still no one uses Siri, but you know, like every day. I don't like touching my phone. Yeah. I don't, I don't use it for a lot, but like if I'm going to send a text or I mean, setting a timer is the obvious one, but anything I can where it actually works. What voice does your Siri use? Well, I think this is a really interesting question. Like how do you decide which voice you want your digital assistants to use? Like I would love to hear from listeners here. Like there's all a million different ways you can contact us. You put it in the comments, send us an email. How did you choose which voice your AI uses?
30:57Or did you even choose? Are you just using the default? Do you care? Okay, so knowing the internet, this is going to lead very quickly to sexy stuff, right? This is a slippery slope. It already has. It already has, yes. Do you know the story of OpenAI modeling ChatGPT off of her? the movie the sci-fi movie her yes with scarlett johansson spike jones yeah yeah it's like a guy joaquin phoenix falls in love with his phone like the female voice ai exactly i think most people know the controversy that happened where scarlett johansson was like really angry at the fact that it appeared that opening i had used her voice but it turned out it was just another voice actress that sounded remarkably like Scarlett Johansson.
31:45Exactly like her, yeah. Yeah, and was like intonating in the voice of Chad G.B.T. almost exactly the way that the movie Her does. Is it a coincidence? You know, the opening, I said it was, but then they pulled, they removed the voice, right? They're like, we're sorry. And they like got rid of that particular one. They did. And they acted, they were like, we don't know what you're talking about. Like, there's no way that we act. But in reporting my book, what I found was that Sam Altman and all the open AI executives for years, years talked about the movie Her as their North Star of what they were trying to achieve.
32:21They were like, we want it to feel like this ever present AI model that's just so seamless and has this amazing, sexy voice. And we want it to feel like that, where it's just like perfectly assisting people. and I there's this quote that I when I was doing interviews on this that I thought was particularly funny where a former employee I was asking him I was like well you know like that movie's also just like it goes badly right yeah it's a cautionary tale right and and he said to me this was like one of my favorite quotes he was like I would think it's because it was an assistant that was wonderfully integrated into a life and the positive arc of that story before it unravels It's a really great story of AI's evolution in society.
33:09And of course, in that movie, there, Joaquin Phoenix falls in love and then they try to have sexual relations where the voice assistant sends like a physical woman to his apartment to try and embody the voice. Right, this is weird, complicated thing. Yeah. And this is like not science fiction. Right, this is the product roadmap of OpenAI. So XAI, which is Elon Musk's AI company, has a like, I forget what it's called, but there's like an adult mode where it like will immediately start flirting with you and try to like sext and it'll like send you AI generated, you know, nudes and things like that.
33:52There's companies like Replica AI, which are explicitly AI girlfriends or boyfriends, but, you know, the gender balance of the users is very much skewed. Right. Male, that are designed to make people fall in love with the AI in exactly that way. Right. If you fall in love with your computer, it is going to be a huge business windfall. Yeah. You're an absolute cash cow as soon as you're in love with the computer. Yeah, I mean, it's a topic for a whole other show, but like the relationships that people are building with these AI tools, I think a lot of the time, like in the media and like the popular discourse, it gets framed as like these sad men alone in their basements.
34:35But it's not a mistake, right? These tools are designed for this sort of thing to happen. And like, as always, like, what will the social consequences be? Well, we're going to find out. Yeah, I really feel like that is the logical conclusion of why these companies are ultimately creating these voice personas is they want to develop trust. But like the maximum version of trust that makes it really addictive and sticky to be using their products so much so that it becomes love. Thanks so much for listening. If you liked what you heard, you can join us next week. And if you're in the UK, listen on BBC Sounds.
35:12If you're outside the UK, you can listen wherever you get your podcasts or search for The Interface podcast on YouTube. If you want to get in touch with us, you can email us at theinterface at bbc.com. You can also WhatsApp us on plus 44-333-207-2472. Or you can find us on social media. Links in the show notes.
From the publisher
When we play a game or fill in a form, are we training robots without knowing it - and would we consent if asked?
Remember Pokémon Go? The company behind it is repurposing the 30 billion images players captured to help robots navigate the real world. It’s the tip of a bigger trend: turning play into data collection. From CAPTCHAs to viral stunts like the Mannequin Challenge, our seemingly harmless online challenges are being quietly funnelled into AI training sets. It’s clever, but it raises awkward questions about consent, transparency, and who profits when our leisure becomes free labour for automation.
Also this week: the meme‑ification of war: games companies, anime producers and pop culture stars bristle at alleged use of their IP in pro‑war White House memes, we look at how politicians are using memes to lessen the severity of the war in Iran - and their role in a new kind of political campaigning. And the personality of AI: Alexa’s new “adult” mode isn’t sexy; it’s sassy. How tech firms craft voice, gender and tone for assistants - what feels inclusive, what feels exploitative, and what feels just downright weird?
The Interface is your weekly guide to the tech rewiring your week and our world. Hosted by journalists Thomas Germain, Karen Hao, and Nicky Woolf, each episode unpacks, week by week, the unfolding story of how technology is shaping all of our futures. No guests. No jargon. Just three sharp voices debating the tech stories that matter — whether they shook a government, broke the internet, or quietly tipped the balance of power.
New episodes drop every Thursday on BBC Sounds in the UK. Outside the UK, find us on BBC.com or wherever you get your podcasts, or watch the video version on YouTube (search “The Interface podcast”).
To get in touch with the team: theinterface@bbc.com The Interface is a BBC Studios production. Producer: Natalia Rodriguez Ford Executive Editor: Philip Sellars




