The 404 Media Podcast (Premium Feed)

The rise of artificial intelligence is the next industrial revolution. Oh,

woah. Okay.

Hello, and welcome to the four zero four Media Podcast where we bring you unparalleled access to hidden worlds both online and IRL. Four zero four Media is a journalist founded company and needs your support. To subscribe, go to 404media.co, as well as bonus content every single week. Subscribers also get access to additional episodes where we respond to their best comments. Gain access to that content at 404media.co.

I'm your host, Joseph, and with me are four zero four Media co founders, Sam Cole. Hey. And Jason Kebler.

Hello. Hello.

Alright. Let's get into it. The first one here, the headline is students boo commencement speaker after she calls AI the quote next industrial revolution. What we'll actually do before we talk about it, let's just play, the audio first of this particular speaker.

Next industrial revolution.

Oh. Woah. What happened? Okay. I struck a chord.

May I finish? Only a few years ago,

AI was not a factor in our lives.

Okay. A lot

of booing today. A lot of clips of people booing, students booing. Yeah. That clip was obviously it's pretty uncomfortable to watch. I'm sure she was very uncomfortable in the moment, but it's, you know, like, of graduates from University of Central Florida's College of Arts and Humanities and the Nicholson School of Communication and Media who are graduating on May 8.

And the commencement speaker was Gloria Caulfield, and she is the vice president of strategic alliances at Tavistock Group. Why she was booked for this particular commencement, I have no idea. Feels like a setup from the the organizers. It's almost you almost feel bad for her because she's so off the mark as far as what these graduates want to hear or need to hear during this graduation. So she she's, she's coming from, like, a a private equity sort of background.

Sometimes that's someone who's, like, connected to the college in some way, like they are an alum or they are, sometimes they're like a famous person who the school is giving them like an honorary doctorate or something like that. As far as I know, this person is not connected to the school. The one at my school is like extremely stupid as well. Like sometimes it's just a rando and it's unclear why. And, yes, in this case, they chose like a private equity person.

On the point of it being, like, people being connected, like, ostensibly to the school, the group, Tavasoc Group that Gloria works for, she her part of her job description is she oversees the health and medical partnerships and business development for Tavistock's visionary Lake Nona community. And Lake Nona is a planned community in Florida, and Lake Nona has a partnership with UCF. And, like, that's where the university's health and sciences campus is. So I assume probably what happened, and, again, we don't know for sure, but, UCF was like, we gotta give, this, like, in exchange for helping us set up the health and sciences campus on Visionary Lake Nona, we're gonna give the the person who helped make that happen a slot in a commencement speech. And, you know, unfortunately for her, the slot was the College of Arts and Humanities.

So that's maybe the connection. We don't really know for sure.

Yeah. Well, that actually leads to my next question, which is that, I mean, all sorts of students may boo AI, you know, like it's no, definitely not a universal position, obviously, but as we're going to see, there's several clips where this has happened, recent commencement speeches, but the fact that these were humanities students specifically, can you just talk about that for a second? Like, what were these people studying as far as we know and how does that relate to AI?

Yeah. I mean, we're talking about students who've spent the last, you know, several years of their lives studying, like, library sciences and, all of the the sorts of, like, history and things like that. And it's the the jobs and the sectors that we've documented pretty thoroughly, especially with libraries and research and, social sciences and things like that where, like, AI is coming in and just making people's jobs way harder. It's not actually creating any kind of, like, great productivity boost. It's just muddying the waters for people like archivists and historians, researchers who are trying to get papers published.

The peer review process is just bogged down incredibly by AI generated slop. And at the same time, you have people like Alex Karp, who's the CEO of Palantir, outright saying that AI will destroy humanity's jobs. And it's like the people developing AI are saying, it will destroy your job. And we then see the CEOs and the executives at these companies blaming AI very publicly for things like layoffs. It's like Meta announced that it was laying off 10% of its workforce this month.

Pinterest, Dell, Microsoft, all these companies that are very visibly and publicly and vocally blaming AI, whether or not that's true or not, you know, whether or not AI is to blame for these layoffs. They're saying, hey. AI is doing this to the jobs that you used to have or the jobs that you're entering into, like the job market that you hope to have as a graduate. And people see that, and people aren't, you know, like, dumb. People are paying attention to the news, paying attention to how AI is being talked about and how people are experiencing AI in their fields.

And that's who she's talking to is people who are trying to get a job right after college. You know, they spend a bunch of money on this degree. They got a bunch of loans. And the market that they're entering into is just really bleak, and a lot of it is AI related or being said that it's AI related, whether or not it is. But, yeah, it's just like read the room.

It's kind of the the message.

Yeah. And you're right about sort of the there's almost a difference between the actual impact of AI on jobs and then what is being said about it. But at the end of the day, these are people who are about to enter the workforce hopefully in some form. And even if they are just saying it and it turns out maybe they're gonna be okay, like, you're still saying it, and this is gonna be the obvious reaction Right. To it.

Is something that we've we've talked about like so much how AI is going to impact jobs or how it's already impacting jobs. We've talked a little bit less about how it is affecting or going to affect new graduates and also the types of things that they study. And I don't know, maybe it won't impact like we talk a lot about like how it's changing school and all that. But I think that in journalism, which is like all we can really speak to, it has really hollowed out the entry level market in a big way. Like a lot of the stuff that is being automated is stuff that previously either interns would do or entry level people would do, where you just sort of like have someone come into a job and you teach them on the job by having them do like, and it's not always easy stuff, but you start you start them out with such stuff that is like pretty straightforward.

And that is the type of thing that AI is pretty good at or is getting better at. And I think that's what's happening with coding. It's what's happening with a lot of jobs where like the entry level stuff is going away first. And also, there are a lot of layoffs or there's been a lot of layoffs, but something else that's been happening is a lot of job openings have been closed. So like prospective jobs that, you know, one could get, they're just not high like companies aren't hiring for those roles.

They're closing those roles. And so I think it is very scary time right now to be graduating into this job market. And I'm sure that these students know that. And as Sam said, they just spent like a shitload of money, spent four years studying and, yeah, then they're then they're going into this like very uncertain space.

Yeah. To have someone from private equity then tell you this is the next industrial revolution. But there was another clip that we didn't write an article about, but Evie, our social media manager, she cut it up and pull it straight online. And it was a really, really another good example. The sort of text that went with that, almost the headline is former Google CEO Eric Schmidt Boo- for praising AI in commencement speech.

Let's just play that as well.

Architects of artificial intelligence. Interesting. You will help shape artificial intelligence. We do not know we do not know the precise contours if if you'd let me make this point, please. If you don't care about science, that's okay, because AI is gonna touch everything else as well.

Whatever path you choose, AI will become part of how work is done. You can now assemble a team of AI agents to help you with the parts that you could never accomplish on your own. When someone offers you a seat on the rocket ship, you do not ask which seat. You just get on. The rocket ship is here.

Let me give you some advice. First, find a way to say yes. Listen, you ran the judgment.

Sam, so could you tell us where this was and when you listened to it before, did any differences jump out at you about this clip or is it very similar to the other one? Like, what what do you make of it? Is it more of the same? Or

I mean, so it's Eric Schmidt who is probably way more used to being in front of a big audience than Gloria Caulfield. So his reaction her reaction is immediately, like, she turns away. She's like, oh, I don't know what to do. He just maybe also because that clip of Gloria going viral, everyone's seen it. So maybe speakers are getting used to being

booed now. I think Eric Schmidt is also used to being booed. He's been

used to being booed, for sure. So he just keeps talking. Like, he leans into it. What he's talking about is, you know, he's he's kind of addressing, like, preempting the booing. He's like, there's a fear in your generation.

The future has already been written. The machines are coming, which starts to feel like, okay. He like, someone's saying someone's acknowledging that there is something is going on here. Like, obviously, are angry, people are scared. And he says, you know, I understand that fear.

And then he goes on to say that the future is unwritten, and that the graduating class of 2026 has power to shape how AI develops, which people boo again. Because you if you notice in these clips and there are several others that we'll get to, but even when they're trying to kinda sell this, like, you still have control over your future. You still have the ability to say how you want AI to exist in your lives. AI will exist in your life, And it is inevitable, and it is a given. And no matter what you do, you're gonna have to figure out how to cope with AI in your life.

And that's what the audiences boo again. They boo they boo to begin with. It's it's definitely he's again, he's he pushes through it and so did Gloria. But I do also wonder if getting booed I mean, thank God graduation season's almost over. But if getting booed at a commencement speech is gonna become a thing that people want to do, that they try to do, if you're like an AI executive or if you're a a PE person and you come in and you get booed at this massive he's a Eric's at ASU.

I don't know if we said this already. ASU, which recently, launched ASU Atomic, which I wrote about, which is stealing lecturer's work and chopping it up into AI slop. But he's speaking at this huge, huge, huge university and so are the others. And it goes mega viral, and it's massively viral online. And then what you're saying reaches the boosters who are on your team, and they are like, yeah.

He's they're not ready for what he's saying.

Sorry. And I'll suggest to clarify. When you say boosters, you mean people who are boosting something, not people who are boo.

Boosters. Boo-

urns. Like boosting Dude, sounded like she was saying the Boosters is

really that's not a thing.

That's the head hackers. Dude, I'm not even I'm not even making the pun. I'm clarifying it for the listener. Alright? But, yes, that's what you mean.

Yeah. Yes. Yeah. Yeah. Like, people who are boosting AI, and they're pro AI or pro whatever Eric Schmidt and Goyer are saying.

And I'm saying,

you know, they don't these kids today, they're they're not they're not ready. They're not entering this. Like, they just need to deal with it, which is what another speaker, Scott Borchetta, CEO of Big Machine Records, he spoke at Middle Tennessee State University, which is a much smaller school, obviously. But, he started with AI as re rewriting production as we sit here. And then he goes on to say, deal with it.

They start building him and he says, deal with it. It's a tool. There's a tool here for sure.

Right. Right. Yeah. Oh, that's a good one. So wait.

I think I think there was one more example just before I asked my last question.

Yeah. Well well, so I was gonna say, I remember, again, I graduated college a long time ago. I don't have the best of memory, so this is maybe useless to say. But I remember we had a weird person, at my commencement speech who was talking like, he just was not super inspiring and people were not booing, but they were not like sufficiently enthusiastic. And I remember someone from our college got on stage and they're like, like, give them a warmer welcome than that.

Like, can do better than that. And that's the thing is I I don't remember what it was, but I feel like around like the Iraq war and other like controversial things, there was a time where college students were like booing or were just like not sufficiently deferential to commencement speakers. And college administrators got like really pissed about this. And they're like and then commencement speakers were like, well, I'm not gonna talk if you if they're not gonna cheer for me and that sort of thing. So I do wonder if there will be consequences for the person who who shouted AI sucks or for like people who boo because colleges are not.

They should be free speech zones. This is like a debate that's been happening like over and over and over Yeah. Like for many years at this point. But like college administrators really don't like embarrassing people that they've invited to speak.

Yeah. And paid all Exactly. Money to

And so I I do wonder if there if there's gonna be some sort of freak out here. Yeah. I added one more to the podcast document here where Magic Johnson, the famous basketball player who also is like involved with the Lakers still, he gave a commencement speech at Stillman College, which is a historically black college university. I'm not sure exactly where it is, but this was also last week. And he talked about AI.

And what he said was something to the effect of like, you know, it's gonna be, it's a scary time to be graduating to this like AI is, you you should like know how AI works because it is a threat and you need to know that it's a threat. And he didn't get booed. Like he gave like a nuanced take and it was like well received. And I think that's important to note just because it's like, well, if you are just a booster who is like here trying to replace workers or you're like a tech billionaire Schmidt vibes, like, you're gonna get booed. But if you are like acknowledging that it exists, that doesn't mean it's a third rail topic.

Yeah. That's totally fair. Sam, the last thing I was just gonna ask is that there was sort of this other viral video recently, I think during a graduation ceremony where the camera pivoted to the crowd. And have you seen this? There was a student and he had his laptop open Haven't seen that.

And he was scrolling through chat GPT. Jason, did you see that?

Oh my god. Are you serious?

Yeah. So No, I didn't see. So basic yeah. Yeah. So the camera pivots away from the stage to the audience.

A graduate presumably holds up his laptop and is just like smiling and grinning and like scrolling through all of his ChatGPT history, which I think the implication is that was using ChatGPT to graduate here in some capacity. And the only reason I bring it up is that while all of this booing is happening and these commencement speeches are happening, AI is across all of these campuses, like in some way, like you mentioned, Sam, the stuff you covered, how he was like regurgitating lectures and spitting it out that way. Students all over are using ChatGPT, and it just like sucks all around that they're going to go into a market where the job prospects are frankly lower, even if it's not because of AI, it's because companies, as you say, Jason, are not giving entry level jobs and that sort of thing. And then presumably because of the pressure that's getting to people to try to get these jobs, maybe they actually have to turn to these tools as well, or they feel like they have to turn to these tools as well. So it's just a fucking mess all around.

All right, We'll leave that there. When we come back, we're going to be talking about, I mean, literally the Internet of shit basically in a whole new way. We'll be right back after this. Alright, and we are back. This is one that Jason wrote.

The headline is Internet of shit. AI poop analysis app offered to sell me database of its users' poops. That's that's a whole new sentence.

Is something this is something that I found on my travels on Reddit, over the weekend a few weeks ago.

Uh-huh.

And it was on the subreddit called DHXchange, which is data hoarder exchange. And so it's a place where people who have large databases of things share them with each other. And the title of the post is I hoarded a large database of something valuable, just not what you expect. A 150,000 stool images. And basically, it is a post from a person who said that they launched an app a few years ago where they have been using machine learning and, like, user reported data from within the app to collect images of poop from smartphone.

And that he's been hoarding all of these images for years at this point. He has a 150,000 of them from, he says, roughly 25,000 different people. And then he says, feels like I'm sitting on a pile of coins, but can't find who wants them. Basically, he's like, I think that this should be worth money. I don't know what to do with them.

I am trying to sell it. And if you want to buy it, like, hit me up.

Yeah. So just two questions. First, and I mean, you you get the answer to this later, but I guess, hypothetically at this point, why does he think this might be worth something? Like to an AI company, like, oh, this is quite an unusual set of images. Like, what's the potential value?

Yeah. I mean, that that's the implication and that that then became like very clear once I started talking to him is that like, he thought that this could be trained, training data for an LLM essentially. And possibly not wrong. Like, I I believe there's probably not that many databases of this sort out there because it's just like not something that people are regularly uploading to the Internet, especially not in the way that that ultimately this was collected, which we can talk about in a second. Yeah.

But, you know, there there is tons and tons of discussion over the last few years and and a few promising studies and applications and then like some not so promising studies and applications of AI being used for health reasons. The ones that are like most often cited are like AI can be used to spot cancer or read like CAT scans or x rays more accurately than humans or like earlier than humans. And there's been a ton of hype around that. Recently, there's been some studies showing like that LLMs especially have the same issues with health data that they have with everything else where they can start hallucinating things. They can like have false positives, this sort of thing.

But, you know, one could imagine wanting to train on human poop and sort of use it for like a health related reason. Yeah. And that's sort of like the implication here.

Sure. So as far as you know, this post doesn't mention the app specifically. So how do you go about looking into this? Like, what's your next step?

Yeah. So I messaged him on Reddit and I was like, I'm interested in this database. Can you share any more information about what you're looking for? And then I said details about the app, where it was collected. And they responded very quickly and said the app's called poop check if you wanna see it.

Poop check. Then also yeah. Poop check. And also said, let's switch to email. And so, I mean, first I was like, okay, like it's a weird thing to make up and lie about, but people lie about stuff on the internet all the time.

And so how do I know that it comes from this app? Like, have no idea. This is just like a random Redditor. So I looked at the app, but then notably, it's like the email address that he gave me was associated with this company called Soft All Things, which is the name of the developer of this app. And they have like an extensive web presence for Poop Check.

It is the same email address that's associated with the App Store, on on Apple App Store. And so I was like, okay. Like, it is from this app. You know, I sort of did some some additional clicking around and it's like, okay, this all checks out.

Well, What does that app do? What is Poop Show?

Yeah. Great question. Great question. So it is a free app called Poop Check, and it basically purports to be an AI analysis tool of your poop. And so you start an account and you take photos of your shit and it analyzes them.

And you also create like a profile. And in this profile, say like your height, your weight, your age, your sex. You also tell it some things about your diet, whether you drink coffee, whether you drink alcohol, whether you are a drug user, like all these sorts of demographic information. And then the idea is that, like, you're supposed to take a picture of every bowel movement. And when you do that, it also asks you a bunch of questions about that and how that experience was for you on any given day.

And so it asks you questions like, how much water have you had today? Like, when was the last time you ate? What did you eat? It also asks like, how long was the duration of this session? It asks you if it was painful, things like this, things like this.

And then you fill all that out and then it uses image recognition to further analyze the picture that you upload. And then it it does do some of this, like to be clear, or

it

seems

Yeah. Like Yeah.

I mean, I think that it does because the things that it it it basically categorizes it further and the things that it categorizes it is like, does it detect mucus in the poop? Does it detect blood? And then it analyzes like the shape and consistency of it. And then it categorized it on this thing called the Bristol scale, which is a like health thing. Like it is, it's an actual medical diagnosis that you certainly don't need an app to do for you.

It's like, I think it's one out of six or one out of seven and there's different categories of the shapes of stool. And it basically like tracks over time your bowel movements. And then finally finally, there's a community aspect to all of this where you can opt in to its credit, it is opt in and they make it like pretty clear where basically there's a forum and you can like upload information about your bowel movements to this forum and it's goes there with a photo of it. And then the community can talk about your shit. And there's a poop leaderboard as well.

So if you like participate in this community, you get points for uploading, get points for commenting, so on and so forth. And a lot of that community is full of people who think that they're sick. Basically, they're like, oh, this is not good. And then asking people to like comment like what disease do you think I have? And so a lot of them are like, well, you probably have like a tapeworm or you should probably take ivermectin, which made a return here.

Ivermectin for those who don't remember is a horse dewormer that I guess sometimes humans take it under medical supervision. It's like an anti parasite, but not for COVID. And it became like a popular thing for COVID. And it's like, well, you probably like need ivermectin. And it's like, cool.

Cool. Cool. Cool. Cool. So that's that's poop check.

That's the app.

Yes. So they're doing all of that. And, I'm really, really resisting the urge to make poop puns. I'm not gonna do it. I've already

I'm trying to stay very serious. This is the, you

know Yes.

Okay. Is an article about privacy and consent and health, etcetera, and we're being very serious.

Okay. We are. We are. So I'll go to the next question, which is that the app is doing all of this and like it is already doing this AI or machine learning, sort of powered analysis and that sort of thing. But then the app, judging by this post and your conversations turning round and is saying, there's some sort of secondary value to all of this data that we've collected or has been submitted by users.

And now we're trying to find a way to monetize it. So what was next? You you're talking to this developer, like, do they describe what the data is? Do they send samples? Like what happens next?

Yeah. So I emailed back and forth with him a handful of times. I I used my real name and all of this, but when I first emailed him, I wasn't like, hey, I'm a journalist doing a story. I was like, hey, my name is Jason. I'm interested in this dataset.

Can I buy this dataset from you?

Cause you wanted to see if you can buy it basically.

Yeah. I wanted to see like, is he just gonna send it to anyone? Like, is this is there any sort of guardrail here? And basically, I ended up telling him that I I wanted it for like, I was like for AI and machine learning. And also I'm doing like an information project, which is an article in the end.

And he was like, yeah, like I'll send you a sample. Here's my pricing options. So the pricing options was for like 10,000 images, unreviewed images, he said, was $3,000, 5,000 fully human reviewed and annotated on top of AI validation is $4,000. So basically, like, it sounds like he had the AI analyze this. And then for this subset of the images, he either went through or had someone else go through and like assess whether the AI did a good job.

And that was $4,000 and then it was for 5,000 reviewed and 5,000 unreviewed, it was $5,000. And he and then he sent me a sample. And in the sample, there was data from four users and it was like five stool images each. So it was like 20 images, definitely of shit. And then the associated JSON files.

So JSON is like a database file, basically, like a text database. And in each of it, it had like information about the user, like their age, their weight. It had all the answers to their questions about like whether they had lactose intolerance, whether any given bowel movement was like painful, etcetera. And it had the AI analysis of all of that, which was just like mucus yesno, color, it like guessed the color, whether it was floating, things like this. And then it had like a confidence score that seemed to be out of one.

So it was like zero to one. So like one of the examples was like floating and it said the answer was sink. So I guess it sunk in the toilet. And then the confidence score was like point five five. So it's like 55% sure is my I think that's what it means.

It's not that sure.

It's not that sure. Like good. No. Whether it's sunk or whether it's floating. Okay.

So you do that, and then eventually you do say, look. I'm a journalist. I'm gonna write about it. As far as

I know, he, he didn't get back. Right?

He didn't respond to my questions once I said I was a journalist. I'm writing an article. No response.

Which is part of the reason why sometimes, like, again, we did the buying the Chinese deep deep fake software. So you you definitely don't do this all the time, but sometimes you have to because that's the way you're gonna be able to cover this story and actually see if a purchase is possible.

Yeah. Yeah. And I think it's like a sliding scale and it's something that we talk about. It's like how are we gonna represent ourselves? Are we going to like, how far are we gonna go with this?

Like, I could have bought this, but like, what were we gonna do with it? Like, we we determined what it was from and that this was for sale and all that. And also, didn't wanna spend $3,000 on shit images.

No. I was already feeling guilty about spending 2,000 on Chinese deep face software. We can't spend 3 k on pictures of shit. Yeah. Yeah.

So I do wanna talk about like what this says about privacy Yeah. And just like the sale of data and and all of that. Because the messaging to users on this is really weird. Like when you download an app from the App Store, they have a privacy policy in the App Store. And then the App Store also has this box that's like, what information does this app collect?

Yeah. And on the App Store, it says it collects no data, which is like obviously not true.

And I think it's not self reported. I think it's self reported.

Yeah. But it's like if you were a privacy conscious person, like, maybe you would look at that and then you click the associate privacy policy. And their privacy policy is basically like, we care very much about your private data. All your poop images are encrypted. And it says like, you can delete your account at any time.

And then if you go to their website, it says basically the same thing. But then when you sign up, like when you download the app and you open the app, you have to make an account. And you don't have to click this link, but there is a link that says service agreement. And if you click the service agreement, it's totally different. And it says, by uploading stool images or any health related data to the app, you grant Soft All Things LLC a worldwide irrevocable perpetual unconditional royalty free fully paid transferable sublicense or license to use, reproduce, modify, adapt, distribute, sell license and create derivative works from such content for any lawful purpose.

And then it goes on and on and it says, you acknowledge that your images and data may be used to create, train, improve, and commercialize AI technologies and machine learning models, and that such models and any outputs derived from your data may be licensed or sold to third parties, including medical organizations, research institutions, and commercial partners. Finally, it says you can ask to have it deleted, your data deleted, but that your data may be irreversibly incorporated into AI models and aggregated data sets, and that deleting your account just deletes your account on the app, but does not remove your poop images from the AI model. So it's basically like if you use this in any way, you're giving us full access to this. You're giving us a license. You're giving us, like, we can do anything we want with it.

And in this case, it's images of your shit, but like this is how it works. This is how like all of these apps work. This is how social media works. Like this is this is the whole thing. Like this is the problem that in the fine print of some terms of service somewhere, if it says, you know, you're giving us this data and we can do whatever we want with it, then that's what happens with it.

Yeah. Yeah, totally. And then sort of the other thing it highlights, which sort of goes back to the beginning of this subreddit where people are buying and selling different datasets is this sort of shady underground economy of all of these left of field weird datasets that people may buy and they definitely try to sell. What do you think this shows us about that? They're like, I don't know, people are just willing to sell anything or there's probably a bunch of selling going on that we don't know.

Like, what's the deal there?

I think there's a bunch of selling we don't know. I think there is probably a fair bit of like underground, market for this sort of data. I don't think like a lot of it's going through like random Reddit posts and things like that. But like a lot of academic data and we've seen like papers about different like weird things. And a lot of that is coming through like somewhat surreptitiously collected data and things like that.

So I I think that there's a lot of this happening. We've seen, I don't know, like data brokers, like the information can be shared like multiple times across different companies and things like that. So I think it is just like, it's a I mean, it's somewhat serious. Like I think that it's not cool that this is happening, But it's like a weird and kind of gross example of a practice that happens like all the time is my sort of perspective on it.

Yeah. Totally. That makes sense. Alright. We'll leave that there.

If you're listening to the free version of the podcast, I'll now play us out. But if you are a paying four zero four Media subscriber, we're going to talk about how researchers wanted to put cameras onto preschool teachers to train AI. That's one I wrote. It's really crazy. You can subscribe and gain access to that content at 404media.co.

We'll be right back after this.

And we are back. I'm gonna take over this section because we're gonna talk about a story that Joseph did called researchers wanted preschool teachers to wear cameras to train AI. Lovely chubbly. This one was like very crazy, very crazy experiment that was proposed. So basically, like you got a document about a planned research study.

What was the document and what did it say?

Yeah. So this was a document that was given, I believe, by the preschool out to parents, you know, getting their consent to participate in this and sort of explaining this study. And in some ways, it was very open about what was happening and in some ways it actually very opaque as well. But when it comes to the openness of it, I'll just read a quote from the document which says that, your permission, your child's lead teacher may wear a small teacher worn camera that captures the teacher's approximate first person perspective and or we may place a fixed video camera inside the classroom. These videos simply capture the normal interactions between teachers and children during regular classroom activities.

Recordings occur during morning program hours up to one hundred and fifty minutes, up to four visits in one month. Your child will not be asked to do anything new or different. Their daily routine will stay exactly the same. Then it goes on to explain, it's for AI. Basically, this is the teacher, they're going to wear a camera and it's going to be to collect data to help train AI models.

I mean, there's a little bit of nuance about sort of what the researcher's history is and that sort of thing, but that was the the main thrust of this document and of the study. Yeah.

Yeah. Yeah. So do we have any idea like what the data was gonna be used for? Like what the study actually was going to be for?

Yeah. So it was to to give some context, the researcher was, doctor Gail Joseph, and they were from the Cultivate Learning team at the University of Washington. That's where this study came out of in Seattle. And they said the goal was to better understand sort of children's interactions with teachers. And of course, you would be able to find out a lot about that if you were just filming what the teacher was seeing every day, not just what the teacher is doing, but of course, what the children are doing as well and how they're interacting with them as well.

But once they've collected the data, the idea was that this could be used to develop AI tools that could then help assess classroom interaction quality. So they don't super drill down into this, but judging from that, it sounds like, well, we can use this data to develop an AI tool, which will then say, I don't know, this teacher is doing a good job, this teacher is doing a bad job. I mean, kind of frankly similar to the app you were talking about, about the poop check one. It's just in this case Just

like that.

Just like it. It's exactly the same thing. It's a perfect comparison. But basically assessing the teachers and there's somewhere else in the document where it says the teachers will get like feedback after their session as well. And actually, this is definitely similar to the poop check.

There's stuff about how the footage will be, I think they use the word coded, right? As in that it's going to be annotated by a human. It says the videos will be coded in that way and sort of that information will be useful as well. So it's basically a data gathering practice inside a preschool to develop and to train AI tools, even though we don't know what really those tools might exactly be in the future. But I mean, this was wild to me because I had never seen data collection like this.

Like, we're just

the camera is very insane. It's very insane. Because one, you are, yes, recording children and, like, preschoolers are often like three, four years old. So we're talking about very young children who themselves can't opt in. It's like their parents have to make this decision.

And in this case, they had to opt out, which is like not a situation that you would want for kids. But presumably, and I mean, I guess we can't say for sure what this was doing, but we've seen other we've done a lot of education reporting on AI, and it's like a lot of companies are trying to replace teachers with like AI tools. Like Emmanuel wrote that huge article about Alpha School, which is an AI school, and that starts at a really young age as well. So one could imagine this like study being to like train like robot teachers or something to that effect. That's reductive, but, like, basically, sounds like that's kind of what the goal was here was to, yeah, like, see how a real teacher does it, and then presumably use it for like research purposes.

Yeah. That's that I mean, that's kind of what I mean about how the document was simultaneously clear, but then also really opaque. Because then like it straight up says we're attaching a camera to a teacher. Okay. That's pretty clear.

And then all of those questions that we're just bringing up is like, we don't know the answer to those. And I don't think really anybody know, maybe even the researchers don't know. Like, they may not know what the ultimate use cases of this is going to be further down the road.

In that case, it is similar to the poop one because they were just like, we're going to collect this and then like, whatever it's useful for, like, we like, we'll have it then. We have it then. Right. Okay. So I don't know if I was a parent, I'd be kind of pissed about this, quite pissed, I think.

What was the reaction of a parent that you spoke to and and sort of like maybe parents more broadly if if we know?

Yeah. So I get sent this document from someone in the community where this preschool was. We agreed not to name the preschool or the parent for that matter as well. Then I get in touch with this parent who is particularly concerned. But my read is that there were probably I think it was more than one parent just being mad about this because very quickly, and we'll talk about the university's reaction in a minute, but like very quickly, the University of Washington told me that there was some feedback from parents and we're incorporating that.

So I spoke to one and they said when they got this document, they were taken aback after reading it. And that was for a few different reasons. The first being that family's abilities to actually give informed consent and how the opt out mechanism would work, and we'll talk about that in a second as well. There was a slight nuance here in that the document, and we published this and people can go read it in the article, but that's all in English. This parent said a lot of families and presumably children at this school, they are migrants or they're non native English speakers.

So are you trying to get a family or a kid or a parent to fully understand what is going on with this AI data collection if the documentation is only available in English, which apparently it was. So I mean, they were concerned for all of those sorts of reasons, but especially the opt out stuff as well. Because yes, this wasn't like you go in and like, yes, I'm gonna opt into the collection. You go into a special room. That's not at all is.

You have to opt out.

Yeah. And I think also, like, if you opt out, like, what happens to your kid? Like, are they are they gonna be put a separate room? Like, are they, like, are they gonna get the teaching that they should be getting? Like, or are they gonna be what are they gonna be recorded and then blurred and removed later?

Like, what what are the specifics of what opting out looks like?

There's a lot of confusion around that. And I think there still is. The parent who I spoke to, they said that let me just double check what they said, give me one second. So the parents said that it was only after questioning teachers and the school administrators that they learned that the researchers will put stickers on children who adopted out, which then sort of goes to your point, Jason, it's like, okay, so then you're going to blur all the faces of the ones who have stickers on or whatever. But then when I actually spoke to the University of Washington, like their press office, however you want to describe it, they said, no, if one child or parent opts out from that class, the entire class is taken out of this study.

So which one of those is actually the case? I don't know. I mean, that's what the university is saying and that's what the parents are saying. But overall, it's a fucking mess. And clearly, people are very concerned about, well, so how do I get my kid out of this?

And I don't know. It was a mess probably at the start and in the execution and, like, definitely the fallout as well when parents did start to push back.

Yeah. Yeah. So, I mean, ultimately, how did the University of Washington react after this parent backlash?

Yeah. So they told me, as I alluded to, I'll just read out a little bit of their quote. Our initial outreach was intended to help us better understand how families would feel about a project that uses artificial intelligence to support teachers. They then said, given the early responses from parents, we have terminated the study and are no longer seeking participation at any sites. They then say it's not unusual for these things to be delayed or stopped in the future.

But yeah, they canned study essentially. Yeah. And So it was actually

a study about how parents would feel about it. It's not no.

That yeah. That like, it it it's a weird statement where they, like, simultaneously say the initial outreach was just about gauging how they would react, but then also we killed this study that we were planning. So it's not I mean, it was an actual study and they just kind of fucked it up basically is what it seems like.

Yeah. Yeah. Yeah. I mean, anything else on this? I feel like there's been just a series of really kind of dystopian AI and education stories of late.

And, you know, we we keep doing we will keep doing them. So if you're aware of any other like weird AI and education stories, like, please send our way.

Yeah. I think we've been joking that we're turning into an an ed tech blog where Sam did all the lecture stuff, Emmanuel did the Alpha School stuff. Jason, what did you do? You did a teacher thing.

I talked to a bunch of teachers about how they felt about AI in schools.

Yeah. Yeah. Yeah. Yeah.

And now I've done this one. So now we just we're all complete, and now we are fully transformed into the EdTech website. Alright. With that, I said, we'll definitely keep an eye on it. If you have any tips, please reach us out.

But apart from that, I will play us out. As a reminder, four zero four Media is journalist founded and supported by subscribers. If you do wish to subscribe to four zero four Media and directly support our work,

please go to 404media.co.

You'll get unlimited access to our articles and an ad free version of this podcast. You'll also get to listen to the subscribers only section where we talk about a bonus story each week. This podcast is produced by Alyssa Mid Cav. Another way to support us

is by leaving a five star rating and review for the podcast.

That stuff really does help us out. I'll go back in and, collect some of them soon to weed out. This has been four zero four Media. We'll see you again next week.