The 404 Media Podcast (Premium Feed)

Hello, and welcome to the 404 Media Podcast, where we bring you unparalleled access to hidden worlds, both online and IRL. 404 Media is a journalist founded company and needs your support. Do subscribe. Go to 404media.c0 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 at404media.c0. I am your host, Joseph. And with me are 404 Media co founders, Sam Cole

Hey.

And Jason Kebler.

Hello.

I will say we've uploaded the 1 year subscribers only podcast. So, there should have been a trailer in your feed. If you like the sound of that, please subscribe, and you'll get the full, episode on the main subscriber's feed. Alright. This story from Sam.

The community is in chaos. Wordpress.org now requires you to denounce affiliation with WP Engine to log in. I'm gonna be real. This is very confusing to me, editing it. I'm very much an outsider looking in, but I understand this is incredibly important to a lot of different people.

So before we get to the main story, Sam, I've got 3 definitions here that I'd really like you to spell out. First of all, what is wordpress.org?

I'm tired just, thinking about the the emails that I'm gonna get after this because every time I define any of these in any way, people email me to say I'm wrong, and they contradict each other. So no one really understands what the fuck

this

is. But so as I understand it, and I've checked this with many people, wordpress.org is part of the WordPress Foundation, which is a nonprofit organization, and it's a open source project. So it creates software, that is open source. There's this big community of people who collaborate and develop tools and parts and pieces of the WordPress open source project. And that's the dot org, which is very important.

Right. And I I I put it in the wrong I put it in the wrong order. Yes. I put it in the wrong order. But when you say sort of WordPress software, broadly what you're referring to is basically the publishing tools and the CMS, which many websites around the web use.

Okay. So they're connected to that. And then, I guess, just the other part that we need to define is, what is WP Engine?

WP Engine is a it's an independent company that, it's it hosts WordPress sites. So as a company, you can take, from the WordPress community project, open source project, and make your own kind of, like, what they've done, which is like a hosting service. You know, you can kinda build off of it on your own because it's open source, and lots and lots of, companies have done this. And you're you're kind of expected to like, as part of, using the WordPress, projects and taking part in that, you're expected to kinda give back and contribute back to the open source in in some way or another, and kinda uphold the same values. And then I guess we should also define WordPress, the dotcom.

Dotcom, the commercial entity, which is owned by Automattic, which is owned by Matt Mullenweg. It took me a long time to realize Automattic has two t's because it's automatic, because it's Matt, just foreshadowing for you. And, so automatic funds, and contributes to the code base of WordPress, the project. So they're separate things, like, in theory,

you know I feel like I'm just a wordpress.com. Yeah. WordPress.com, though, is as Sam said, a for profit entity, but it's also like a whole, like, hosted version of WordPress, where you can just like go on there, pay them a monthly fee and spin up a WordPress site. It is sort of like ghost pro, which is what we use, where we are using a version of Ghost, but it's all like one package where you can, like, set up a website on that entity. Yes.

Does that make sense?

I think most people are more familiar with dotcom. If you were gonna say, I'm gonna make a WordPress site, you're probably gonna go to WordPress.com, and you're immediately gonna have all the tools right there. It's very easy out of the box. You're probably not gonna, get super into, like, the coding of using the dot work, stuff, which you could. It's not that it's not that hard, but, you know, the the dotcom is kind of like if you're just gonna take home the full thing without having to fuck with it too much.

But then some people go to WP Engine as well almost as an alternative way of doing that. So then with all of those definitions in place, hey, if you didn't understand it, hit that rewind button. Just listen to that bit over and over again. Because you you I don't know. It it is very, very crazy confusing, but I think I think we established it.

So before we get into the news, really broadly, before we get into those specifics, what is the conflict happening here between these various part, parties? Because this has kind of been brewing for a little while. Right?

Yeah. Yeah. So in September, Matt Mollenweg, which, I just mentioned, owns or he founded he cofounded, WordPress dotorg. He owns and founded, automatic, which owns.com. He posted he wrote a blog post on the wordpress.org site saying that, WordPress engine basically, accusing WordPress engine of not contributing enough to open source and also, for confusing people by calling itself WP Engine, like confusing people into thinking that WP Engine was WordPress.

Like an official thing. Yeah.

Yeah. And have I maybe I've been saying WordPress engine. That's not really what that's not what it's called, by the way. See, people are already people are already mad at me. I can feel them I feel my ears tingling.

Were WP Engine is the name of this the company.

Yes. And

he's like, well, it's confusing people because WP stand for?

Who could say? You know, we don't really know.

Reading it. I'm just reading it as it's written in the document. I'm I'm trying to be really careful right now.

Well yeah. And and so Matt writes his blog, and he's like, WP Engine confused my mom into thinking that they are part of my company, which by the way, why is he talking about WP Engine with his mom? Who knows? You know, we that may come to light someday. But yeah.

So he's saying that they are confusing people and that they need a they need to pay for a trademark license to use that name. He's pissed off. He's accusing their investors of not contributing enough to the open source project, and kinda throwing out all these things. And then, basically, WordPress WP Engine. WP Engine, right, does, like it's like sends him a cease and desist and is basically, like, stop trashing us in public and or sends him and automatic a cease and desist, and then automatic sends them back a cease and desist saying, you you know, what I just said Matt accused them of.

Yes.

And then WP Engine files a lawsuit, which we'll get into, but that's enough

for

right now.

Yeah. That that makes sense. So you have these accusations that, hey, you're taking this code base, but you're not contributing enough to it. It's a brand confusion thing as well, and then there's firing back, and then firing back to the firing back, etcetera. Last week, I think at this point, there was an update to a login page.

What exactly was that update, and what was it sort of asking users to do?

Mhmm. So the updates on this is on the WordPress dotorg login. It says so if you wanna log in to your account for WordPress dotorg, that gets you into, like, support forums and, to be able to contribute. So contributors to the WordPress dot org project need to log in using this. There's now suddenly, as of last week, a checkbox that you have to check before you log in or register a new account that says, I am not affiliated with WP Engine in any way, financially or otherwise.

You have to check that. You have to agree to that before you log in, which is complicated because a lot of people are affiliated with them because they're doing, like if they're part of the project, they're doing, like, consulting for them, or they have clients who have clients who are of theirs. Like, WP Engine has a lot of, they have a lot of stake in this, like, ecosystem as does automatic. They're kind of, I don't know when I say competitors, but they're kinda doing the same thing. So, it's a very tricky thing all of a sudden for community members and contributors to even just access.org because now they have to, like, they're like, is this legally binding?

This is dragging us into this kind of, like, for profit beef, and we were on the nonprofit side. So it all gets very hostile and complicated and drags a lot more people into it that, didn't really expect to be, I think.

Yeah. And so they've basically, like, put up a wall saying, if you're affiliated at all with WP Engine, we're, like, not gonna let you in, in effect. And you said I mean, it sounds like that's quite confusing, potentially, for, users. Because as you say, well, they might be affiliated. They might not.

They just want to log in and continue, to do their job or their projects. I mean, how did users that you spoke to react to this login box change?

I mean, they were freaked out. It's there's a Slack, that is for people who are contributors to the project, and, they were asking Matt's in the Slack, and they're asking Matt all these questions. And then Matt confirmed in the Slack that this box is part of him banning WP Engine from using any of WordPress .org's resources, which is a whole other part of this. People were asking, like, you know, like, what's the deal with this? Like, can you explain?

And then he's just kinda saying, like, well, it's your choice whether to check it or not. Lots of people lie all the time. And it's like, man, this is crazy. Like, what are you talking about? Like, this is it just feels like a like, very petty to them that this has even reached this level.

So, yeah, the folks I talked to were very upset. And then in the Slack, which is another kind of part of this, people were getting banned for even just, like, questioning it, it seemed like. There were a few contributors who had been contributors for a really long time.

I mean, you you spoke to someone who'd been doing it for 10 years. Right?

Yeah. Yeah. Like like a really like, this has been part of their life and part of something they've been doing for a really long time, and got kicked out of the Slack. So now he doesn't have access to this community resource that he was using to help keep the project going, and he was, like, a really important part of this, like, maintaining lots of different parts of the open source. So that's kind of the the drama, where we left it last week.

And, of course, much has happened since then. There's a there's a whole lot of drama with, like, a a plug in being altered and banned. Like

Well, give me I don't say I don't even know about the plug in thing. So give me, your version of of that.

Before you do that, and and this has come through, but I just wanna stress, like, this is so besides being complicated, it's so messy. It is like messy, messy drama drama to the like 15th power. Like people, like there's a lot of like vindictiveness here. There's a lot of people who have been contributing to these projects for a long time getting kicked out of communities that they, like, were in for a while. There's, like, a lot of Matt Mullenweg posting blog posts, then deleting the blog post, then posting altered versions of the blog post.

Like he's going on other podcasts and, you know, it's super messy. And I know, like, we're being very careful because it just is comp 1, it's complicated. 2, it's messy. Like, there's not an there's not a better word for it than, like, this is drama, drama, drama, drama, drama, and it's, like, such a distraction.

And it might be having a real impact. Well, it is having a real impact on people because now they can't get in and they can't, you know, contribute potentially. And, yeah, very very distracting. But as

well the worry is, like, it's gonna break because it's, like, the people working for automatic are they're and also, like, people contributing, in the community. Like, WordPress is running a lot of sites on the Internet at this point. I think it's, like, something close to, like the the number that gets thrown around is, like, 40%. I don't know if that's accurate, really, but it's a lot of sites out there that depend on WordPress to function. And if things are kind of imploding, which they seem to be within automatic and within the pro the larger project and the foundation, that puts a lot of the Internet at risk in the way it functions.

So just very quickly, the plug in thing happened, like, late last week where .org took over a super popular WP Engine plug in, and just took it just like just took it over and invoked some kind of, like, guideline, and it feels it feels that feels vindictive, but, you know, who knows really? And then, also, a couple weeks ago, Matt offered people working at automatic this big alignment offer, where if they didn't agree with him and his actions, they could take a it's basically a severance agreement where they got paid out, and we're told, you know, you're free to go, which a a lot of people took. It was, like, a 159 people took that offer. God knows how many stayed because they need health care and, like, and need a job.

Yeah. While your boss is doing these various things. You know, like, I just I wanna work on WordPress. Yeah.

Yeah. It's like, I'm here to do a job. I'm here to do my coding job. Yeah. Not, not get involved in all this drama.

So

I will say that, yes. You know, allegedly, 478,000,000 websites built on WordPress. That's according to a particular hosting provider, who would potentially have at least some insight into that. But so that's, that as yeah. And, of course, if it continues, maybe there could be a dramatic impact on at least some of those sites.

Right? You mentioned this lawsuit, and that's going a little bit back in the timeline, I think. But, like, I mean, we don't have to do a play by play of what it is exactly. But is it just like a more formal sort of manifestation of the beef here? Like, what is the lawsuit?

The lawsuit is from WP Engine against Automatic, and it accuses them of extortion and abuse of power.

Well

I would

just We'll leave that there.

Based on everything else we've just said, I will I will, I'll leave that there. I mean, it's, you know, it's we linked to it in the story, the lawsuit, but, basically, you know, WP Engine is accusing, Matt and Automatic of, you know, not letting them have the freedom to to run their their company the way that they want to even though that's kind of what was the agreed upon. Like, that's the premise of why they built their company on WordPress.

Yes. Be because there's an open source license, you're supposed to be able to go use it, and you can't or you shouldn't just add extra terms later, when you feel like it, that sort of thing. And I guess it just leads to our last thing, which is, sort of, what's the latest in that it seems that Muhlenberg is now making potential enemies with other high profile people in this space. There was sort of a blog post and then a response to the blog post. I mean, what's happening there?

Is it is it just more of the same? Or

I mean, it's more of the same. I actually have not had time to read way into what's going on, but, yeah. I don't know. Jason, did you read the DHH blog?

Yeah. I mean, the the short version of it is there's a beef between Matt Mullenweg and David Heinemeyer Hansen, better known as D HH, who is a race car driver slash, open source developer, better known in this world for his open source development. He created like Ruby on rails. He created campfire, which was like an open source version, like a predecessor to Slack and a couple other pretty big, tech like software things like very, successful, highly successful. He runs this software development firm called 37 signals and basically like D HH, criticized Matt Mullenweg, and then Matt Mullenweg wrote a blog post basically railing against DHH, and that claimed he was not as successful as he should have been because, like, companies like Slack took his general idea and turned and made them into, like, gigantic companies.

So he was pretty much to paraphrase saying like, you're not as successful as me more or less. And then he edited that post to be like, sorry. I was mad. And that all happened in, like, the last 24 hours as we're recording this. And

Yeah. It's not it's not the sort of communications you expect from a senior person or a leader of, a company, and not just a company in the organization, but one with that's very, very important to the Internet. And I know it sounds like it sounds like, oh, we're on one side, we're just talking about drama, that sort of thing. And that is true in some way because, well, they've got it. They've done that.

They've brought it upon themselves. But there is, like, significant impact here for people's jobs, for the tech industry, and potentially the web, you know, writ large, depending on what happens now. So I hope people don't listen to this and then just think, oh, you know, this is, like, silly gossip or something because, no, this could actually be very, very real, and it is very, very real for a lot of people as well.

Yeah. People are really scared for their jobs. I think people are it's I mean, we've all worked at a unstable workplace. Like, many people work at places where they don't have any control over the whims of the people, the higher ups, the executives, the owners, the CEOs, what the what craziness they get up to. Not saying mass doing craziness, but, like, you know, like, can relate to a workplace where things are, things are unstable and confusing, and people are leaving en masse and taking offers to, like, shut down whole departments.

And it's just it's not fun, and it's honestly very scary for a lot of people because they need they need the stable work. And it's just you know, it seems to be this drive from this one kind of beef that's occurring.

I think without knowing the intense specifics of open source licensing and these specific projects and their histories and all of that, there is, like, probably space to have some sort of nuanced conversation about, like, open source contributions and licensing and who can profit off of open source software and in what ways and how is that ethical and naming conventions, like, should something like WP Engine, like, should they should it be called WordPress engine? Is WP Engine okay? Like, are they contributing enough to the project and in what ways are they contributing to the project? Blah blah blah. Like, all of this sort of thing, these are conversations that we sort of talked about before, and I think that we we most talked about them when there was, like, a social engineering attack on this really important part of open source software.

And it's, like, in general, open source software, it can be the case where there's like a few people who contribute to a project and then a bigger company takes and spins that off and makes a lot of money on it and the people who contributed to that project feel taken advantage of in some way. I'm not saying that's what's happening here. I'm saying, like, that is a constant topic of conversation and this is, like, a very nuanced thing that people are gonna need to continue talking about. But, like, what is happening here is there is like a central figure, like in this case, Matt Mollenweg has like a very, he's very outspoken, he's very powerful, and rather than having, like, the nuanced conversation, it's, like, playing out in court, it's playing out in, like, on social media. It's playing out in people getting, feeling like they have to take a buyout to leave their jobs.

People feeling like they're being kicked out of a community that they're they work on, so on and so on and so forth. And it's, like, at the core here, there's a very important thing, but the way that it's playing out feels frankly to me, like, very immature, very, like I said, messy, and people don't like that.

Yeah. For sure. Alright. We'll leave that there. I'm sure we will have more updates in the future, but we'll be right back.

And then when we return, we're gonna talk about how Jason used AI to apply for a crap ton of jobs while he was having breakfast, I think. We'll be right back after this. Alright. And we are back. This is one that Jason wrote.

Quote, I applied to 2,843 roles, end quote. The rise of AI powered job application bots. This is crazy. First of all, walk us through the story you tell at the top of the article. Sort of where were you, what was happening, and then what were you doing?

Yeah. So I was at a coffee shop working, and I had found this thing called auto jobs applier AIHawk bot on GitHub, and it does what it says in the, in the name. It automatically applies to jobs for you, using AI. And specifically, it does it on LinkedIn. And it's a Python script that plugs into OpenAI more or less, or you can use it with Gemini and a couple other LLMs.

And I had seen it going, like, a little bit viral on Twitter earlier, that week. And so, basically, I wanted to see if it worked. So I installed Python. I, like, set up the script and I am in this coffee shop. I've ordered my food.

What did you what food? I'm trying to do narrative journalism.

It was scrambled eggs. There was a sausage involved. There was, a salad. I was in Taiwan, which is maybe important. I and will be the subject of future stories.

But anyways

That was the that was the important part. I don't know what you're saying anyways. That that that's like crucial context. So you have this food.

Yeah. I have breakfast food. Yeah. To be totally honest, it was the first coffee that I'd had in a month, which is good. It went well.

The coffee went well. Didn't get too jittery from it. So that was a win. But I typed, I clicked enter, like I, I had a terminal window up, which is really cool. Anytime you're in the, in the terminal, it's very cool.

And you're, you're hacking essentially.

It's true. Yeah. And then, and then you change the color scheme to be, like, green and black. Exactly. Hacking.

So what happens when you hit enter?

So I hit enter and the script starts running and, like, all of the so, like, lines of text, like, you know, start going and the bot is essentially explaining what it's doing. And so it opens up a Chrome window. It navigates to LinkedIn. I had already logged into LinkedIn, so it had access to my LinkedIn and it starts searching for jobs. And it starts searching for jobs according to parameters that I had given the bot.

Like, I'd preprogrammed it into the bot, you know, in according to the read me file that I had. It wasn't, like, that complicated to do.

That was probably, like, a configuration file or something.

It was a configuration file. And, basically, I I in that configuration file, I gave my name, my address, my, like, work history, stuff like that. Like, basically a, base very basic resume of my life. And then I also gave it information about the types of jobs that I wanted. So in this case, I had like software engineering, marketing, I had, like communications.

I, I originally had editor and writer in there because that's the types of jobs I had had in the past and it starts searching through LinkedIn for any jobs that, like, meet this description. And when it finds it, the bot clicks on this button called easy apply, which is, like, you can apply directly on LinkedIn.

And And that's a that's a native LinkedIn feature.

It's a native LinkedIn feature, where you can, yeah, you can just go through the entire application process on LinkedIn. It doesn't send you to a second website. So you're applying on LinkedIn itself. And so then the bot, like, in the terminal says, like, generating resume, generating cover letter. I can see that it reading the, the job, like, description more or less.

Like, it it will intake it and show, like, hey, this is a job at Brilliant Earth, which is a, jewelry company. And it was like a social media job at a jewelry company. So it, like, read that and it, it said like, oh, blah, blah, blah. Like, we have precious gems and we're looking for someone who's passionate about jewelry. And I could see that it was reading that or like the LLM was like, intaking the information and then it generated a cover letter.

It generated a resume and it uploaded them into LinkedIn. And it also answered a bunch of questions like it, it would input my name and address. A lot of job applications ask if you have like military experience because there's like all of these, not because they're gonna ask you to join their militia, but because there's like, job incentives for hiring veterans and things like that. Sometimes there's a question like, what makes you qualified to work here? And it would generate an answer to that question.

And then I would see a click through all of these, you know, these menus and click apply. And, like, I set this up, you know, right after I ordered my food at the cafe. And by the time I had finished eating the food, I had applied to 17 jobs.

That's crazy.

So it was like, it was maybe like 45 minutes and it had applied to 17 jobs. There was one moment where I had to pee and I, like, went to the bathroom for, like, 1.5 minutes. And I came back and it had applied to 2 jobs while I was gone. And I don't even know what jobs they were because I was just like, oh, I wasn't looking at it. So it was pretty wild.

Like, it was a pretty wild software. And as someone who has applied to jobs in the past, it's like, I've spent days on a single job application before and it this did it very, very quickly. Yeah. Whether it did it well is something we can talk about, but it it generated, like, a plausible looking, job application in a matter of, you know, seconds

Yeah.

For some of them.

So, basically, it's not only finding the job, it's also applying to it's like doing the whole end to end process of applying for a job, basically. Finding them and doing the process as well. Obviously, you tried it and it was, I mean, successful in the sense that the script worked. Yes. As you say, I guess, we'll see if you get any jobs, I guess.

I've been rejected from 10 of them so far, which is devastating. Hey. But there's still 7 more that I'm I that I'm up for. So could be could be having some personal soon.

Yeah. If if you could let us know as your cofounders, if you do get like, I I don't know. I just I'd appreciate the heads up. You know? Exactly.

I think that would be nice. But many, many more people are using it. Tell us about this Telegram group where it seems these users, you know, gather, and how many are in there?

So before I wrote this, there was 4,700 people in it. There's now 5,000 in it. So it's been a few days. There's 5,000 people in it. And it it's basically like a Telegram group where people ask, like, how do I set this up if they get stuck?

Because it is it took me, I wanna say, like, a couple hours to get it to work, but that was primarily because I didn't have Python installed on the device I was using, and there was, like, some troubleshooting involved there.

Installing Python is the worst thing you have to do when you get a new MacBook because it no longer ships by default, I seem to remember, and it's a pain in the ass.

It's a pain in the ass, and then I installed it, and then they're, like, oh, actually, we installed an old version. Yep. And then it's, like, oh, now I installed the new version, and then there's like, well, the actually, this one isn't compatible with the new version. You need to go back. Yeah.

So it was like a lot of that.

And then you have to make sure the script is using the correct version of Python. Sorry. I'm just really glad that you're feeling these pains

as well now. Crazy because it worked. It worked. I did it one day. I did it the next day.

And then on the 3rd day, I tried again, and now it's like, Python's not installed. And I'm like, what are you talking about?

Yeah. And it's

like I love it. It's not in the correct directory or something. Like, I have no idea what's going on. It's driving me crazy because I tried to to do it again to test something else, and I don't know where it went. But, anyways,

these people are in there and they probably have the same sort of issues because I think the kind of the point, at least for some of them, these people aren't gonna be software developers necessarily. They just want to apply to a bunch of jobs. Right?

They just wanna apply to a bunch of jobs, and now there's, like, a couple of startups that are, like, we are taking this tool and we're putting in a web interface so that, you don't need to install Python. But, basically, people are, like, asking for tech support, but then there's also a, like, success room where people are talking about how many jobs they've applied to, whether they've gotten any, any offers or interviews. And, like, here's some of the the quotes, that I pulled from there. Ran it overnight, applied to around a 150 jobs, got an HR call this morning, and I'm just getting started. LOL.

I'm gonna have it run 247. I've been using this platform for a little over 3 months now during which I applied to 2,843 roles. In that time, I've had 4 interviews, received one offer for a senior data engineer role at £85,000. Others are like, I've, just started already 20 applications sent in 20 minutes. The bot will be running all night, LOL.

So it's like a bunch of people saying this and a bunch of people using this. And, you know, it's the AI of an entirely new thing. But what I think is very interesting here is that companies and HR software developers and recruiters and stuff like that have been using AI to review job applications for a really long time and that has been really bad. There's been like a few academic studies about how, you know, AI review is discriminating against people in an all new way. There's all of these people who just like apply for jobs and their, applications are never ever reviewed by a human.

Joseph, you wrote about a company where the person, the quote unquote person interviewing someone for a job was an AI, so it's being integrated even further. And as someone who like, if you are someone who's applying for jobs, I think that can be extremely demoralizing. There's tons of people who have applied for, you know, dozens and dozens and dozens of jobs manually and got no response. And that's really frustrating. So I think that that this is a response to that state of affairs.

And the other like specific thing is, I'm looking for what it's called right now, but there's a specific type of, like, there's specific language for the type of AI software that reviews, resumes. So there's a specific type of AI software called applicant tracking system, and it is this HR automation tool that automatically, reviews resumes and cover letters for specific keywords that are, like, in the job description itself. And we are already training humans how to write cover letters so that that type of AI is properly reviewing their resume and cover letter. Like, it's basically SEO, but for your cover letter. And it's like, make sure that you have these keywords and you'll get past the first, like, automated review process.

And what this tool is doing is it is taking ATS, this applicant tracking system lingo, and it is automatically writing a cover letter that applies to the best standards for that. And, you know, it's AI so, like, Lord knows, it's probably making some mistakes, but basically you now have an AI large language model writing cover letters with the explicit purpose of having it reviewed by other AI. And I think that is super bleak on one hand. And on the other hand, I'm like, I don't know if I was someone who applied to a 1,000 jobs manually and didn't get them, like, maybe I would do this because, like, at that point, who gives a shit?

Yeah. Exactly. It's almost like this arms race between the companies trying to use AI to streamline it for their own purposes. And then, I mean, people, quite rightly, I think, being like, well, fuck you. I wanna use AI as well.

You know? Alright. Let's leave that there. If you are listening to the free version of the podcast, I will now play us out. But if you are a paying 4 0 4 media subscriber, we're gonna talk about how the National Archives in the US are pushing AI onto employees as well.

We got chat logs, a presentation. We're gonna go into all of those details in the next section. You can subscribe and gain access to that content at 404media.c0. We'll be right back after this. Alright.

And we are back in the subscribers only section. Jason, you wrote this one. AI amazing tech venture. Such that's the title of the presentation. And you gotta admit, that's pretty good.

National Archives pushes Google Gemini AI on employees. I guess, just first of all, for those who don't know, what is the National Archives and Records Administration, Jason? Like, what is it, and what do they do?

I mean, they have the constitution, for example, like, in the National Archives building in on the mall in DC. It's like they have I believe they have a copy of the Magna Carta, the declaration of independence, you know, the You

know, those are, like, PDFs online. I could have a copy of those.

Exactly. Yeah. So so they are responsible for, like, preserving our nation's history, you know, like, broadly speaking, archiving and preserving our nation's history and records. You know, this includes our very important documents, but then also things like soldier records, like, you know, govern like, government bills, things that the government does over the course of the history of the United States?

Communications between agencies, presidential records, all of that sort of thing. Yeah.

Yeah. It's very, very important. The main building is is on the National Mall. Famously, Nicholas Cage in National Treasure, like, breaks into it and steals the declaration of independence.

I feel like I haven't seen that film in so long. I totally forgot about that. I just do I always know I make reference to Nick Cage being a national treasure. Obviously, that's the joke with the title of the film. I should watch that again.

I totally forgot that was the whole

Here's, here's an Easter egg for our subscribers. Uh-huh. I was in National Treasure 2. What? I'm an extra in the background.

They came to the university of Maryland when I was there and they paid me $200, to walk in the background. It was really cool. Very fun experience.

And you can see can you see yourself?

A clip. I'll send you the clip. It's, you can see me. I went to see the movie, and it's me. It's definitely me.

I'm wearing a gray hoodie. I'm walking in the background. You don't

need to send me a clip. I'm gonna watch the whole film in all of you. And I'll check the credits, make sure, you know, properly attributed, everything like that. Okay. So very, very important government agency for preserving the historical record.

They had a presentation recently, again, called AI amazing tech venture. What was this presentation about a bit more specifically? And and this was internally to employees of the agency. Right?

Yeah. So very briefly, like, earlier this year, the Biden administration put out like a policy directive to, federal agencies, which includes like NASA, NOAA, the National Archives, the State Department, that they all need to explore the use of AI because, you know, the government loves to pretend that it's doing innovation and it doesn't wanna be left behind on anything. And so it directed each agency to look into how it would implement AI in some way, shape, or form. And the National Archives, like, started kind of doing this. And earlier this year, I obtained a an internal email saying that they had started trying to use chat gpt and they were like, uh-oh, because it was hallucinating things.

It was, well, it was making up things. It was also, they were really worried about data protection more or less. And so the National Archives banned the use of ChatGPT

and they They wrote the original article. Yeah. You wrote that at the time. Right?

And in the aftermath of this other National Archives employees reached out to me and said they banned chat gbt, but they, like, are really, they really like Google Gemini, which is Google's version of AI, like an AI language model. And so I got a tip that they had this meeting, this, like, I don't think it was an all hands, but many, many National Archives employees were, invited to this AI tech venture, presentation where they showed how they might be able to implement, Google's Gemini into the National Archives.

And this this wasn't Google giving the presentation. It was somebody from, excuse me, almost like the IT department or

something similar. Like, yeah, the IT department of the National Archives is like, we have we created a demo for you and we're gonna show you how you could use this. And in that demo, the the slides of which we published this week, they basically are like, you can train Gemini on National Archives documents, and then you can ask Gemini questions about it. And if you do that, productivity will go through the roof. It could be could be useful for you.

Okay. I mean, we'll get to that bit, but you mentioned the presentation slides that we got and published. How did you get that, and, what else did you get as well?

Yeah. So I filed a FOIA for it, so a public records request, and I just I got the slides, and I also got the chat records because it was a Zoom call. So all of the employee chats, while this was going on and the chats make very clear that the national archives employees were kinda like, what the fuck? There there's like a lot of, employees saying, like, they had taken issue with the fact that the archives had programmed Google Gemini to call itself an expert archivist because they were like, we're the expert archivists here. Like, why why you turn this robot into 1?

Yeah.

You just bought this tool, and you're, like, literally giving it our job title now. Okay. Thanks.

Right. And there's people talking about how they're concerned with the environmental impacts. There's people talking about like very, very concerned about the fact that, you know, Google's AI in particular has, famously given a lot of incorrect information on the search, like the search AI overviews thing. There's people talking about how, you know, how can we trust this technology when we are literally in charge of preserving our nation's history when, when it makes things up? And then there's also a couple of people who are like worried about the copyright implications of this, so on and so forth.

And and also worried about, like, how are we gonna disclose to the public if we're using AI? And unfortunately, the, archives refused to release the video of this. Do

do they have it?

They have it, but they said that for privacy reasons, they wouldn't release the video. So I have appealed that, but I don't know if I'm gonna get it or not. And, so it's not clear sort of like I have the slides, I have the chat, so you can kind of follow along with like what's happening, but I don't have, I don't know if the archives answered these concerns on that call, essentially.

You have, like, one half of the conversation, which is the employees reacting to what's being said, but you don't know specifically what was said. It's just you have a pretty good idea from from the slides. So Yeah. So they wanna use this internally, first of all, and they think it could increase productivity. Was there any elaboration in slides or anything about what they mean by that exactly?

Like, how would that improve productivity?

So interestingly, they were talking about using it to do kind of like, for archivists to ask questions of this AI and use it to help them in their own reports. But while I was doing this article, another archives employee gave a speech at an archiving convention, like a big archiving convention, and they talked about, using AI to digitize revolutionary war pension files, which are 2,500,000 pages of handwritten notes, from the revolutionary war to basically like determine who fought in it more or less. And they basically were like, at the rate that we were doing this, it was gonna take 25 years for us to digitize these records. AI did it in couple days or whatever, and it got it 90%. Correct.

And it's like, okay. If 90% correct is good enough for you, I don't know. Not my,

I gotta say beyond obviously, I don't I don't know if it's good enough for the National Archives. Almost out of in its just an exercise by itself. That's pretty cool. You know? I like to just say, like,

they don't really this is my beef with it is they don't really discern between OCR, like optical character recognition and, like, generative AI. Like I feel pretty confident in OCR. It's a technology that's been used for a long time. It gets things wrong. When it gets things wrong, it's like a one letter is off very often, or like one word is off here or there.

And I think that if you have some sort of disclaimer or if you use it in some way, shape or form, and then have a human review it, like that, I can see that working, but they're also in these same presentations talking about generative AI as well. And so they mentioned this thing called Archie AI, which is an AI chatbot, like so many other AI chatbots. And I asked the National Archives about this because they didn't give very much, like, specifics in this presentation. And they're like, oh, yeah. We're just gonna, like, launch this to the public, let the public ask anything they want of this chatbot.

The chatbot is gonna, like, give it answers, give the public answers about the nation's history. Like, it's just gonna generate stuff and send it to people. And they're like, well, we're gonna have like a disclaimer saying this might not be correct.

Which is a wild thing for the National Archives to say. Like, if it's

That's really wild. And they also they they say, quote, as you can tell, Archie AI is a cornerstone of our AI learning. Like, they are all in on this on Archie, the AI chatbot, which it sounds like Clippy, but for civil war records and the constitution and the JFK assassination and all this stuff. And it sounds nuts to me. And you know, based on the response to our article, people seem to think this is bad, which I I don't know.

I was kind of like I was curious if people would just collectively shrug at this because it's just like another AI thing, but it seems like national archives puts AI into everything is like a bridge too far for some people, which is heartening to see for me.

Yeah. It's just every time I get the Google AI summaries, I've said this a 1000000 times, but I just scroll over them because I simply do not trust them. I don't wanna introduce more work where I'm like, oh, I'm gonna read that, and I have to go verify it. I just wanna go straight to the original source of the thing. I'm trying to Google or whatever.

So I just, like I have a problem with trying to get knowledge from AI chatbots. And I think it's really easy to imagine that they release Archie to the public. People ask you questions. Either it makes mistakes and then people tweet those and they go viral for whatever reason. Or someone tries to fuck with it, like, a bit more maliciously, and then it outputs something really, really bad, and then they run with that.

And maybe the National Archives thinks it's okay by saying, hey, look, you know, it's prone to errors, that sort of thing. It's just that any government agency that should not compromise on truth or accuracy, you would think it would be the National Archives. Right?

Yeah. I mean, that and also just to get very, like, high and mighty for a moment, it's like trust in the government is so low right now that I just don't see this ending well. It just doesn't seem like the sort of thing where I don't know the archives is, I don't know if there's any part of the government that's depoliticized at this moment, but like the archives is not super controversial as far as I know. And I think that if you have this AI chatbot, you know, talking about JFK assassination conspiracy theories or something and getting stuff wrong, that's not, it's not gonna be good. I don't think it's gonna work out very well.

Well, we will definitely retrying Archie, when it comes out. I definitely recommend people go look at the chat logs and all that sort of thing. Just scroll through it. And it's it's interesting because chat logs are not something you often see from a foyer. So, you know, it's nice to get that as well.

Alright. I will play us out. As a reminder, 404 Media is journalist founded and supported by subscribers. If you wish to subscribe to 404 Media and directly support our work, please go to 404media.c0. 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'll talk about a bonus story each week. This podcast is made in partnership with Kaleidoscope. Another way to support us is by leaving a 5 star rating and review for the podcast. That stuff really helps us out. This has been 404 Media.

We will see you again next week.