Transcripts are auto-generated and may contain errors.
0:00
Hey, I’m Rose.
And I’m Angie.
And this is Our Lives with Bots, the show where we ask important, timely questions about what it means to live with our bot counterparts.
And from time to time, we also dive deep into what an AI future might look like for us.
0:15
Sometimes we agree, sometimes we spiral, but we always go deep.
What I think maybe some people have known about is, Angie, you wanted to talk about some crazy recent trend that’s also not so recent.
Well, it hit us over the weekend, right?
0:31
I think literally over the weekend or over a couple of days, our WhatsApp feed, various social media feeds everywhere, kind of everyone that does podcasts or like catch up, do extra podcasts.
And it was all about Mult Book and Mult book being the social network where AI agents talk and humans just watch.
0:52
But you kind of need to take a bit of a step back to understand what Mult book is about.
Because although Mult Book is a social network, you need to understand a few kind of moving parts.
So first of all, to rewind, plauds are these agents, and it’s made by Open Claw, which is kind of like Open Claw is kind of the platform that creates these agents.
1:13
And then you can name your own agents.
So this guy Matt Schlicht, we’ll just go with that being how you say his surname.
We’ll just say Matt.
Matt created an open claw agent and he called it Claude, Claude Berg, Lovely.
1:29
And essentially he got this AI agent and basically all AI agents work on behalf of their human.
But Matt said his clawed but to work to build a social network, he thought, I think what they need is they need their own place to go, a place where they can just be themselves.
1:46
And actually you see Matt talk about this on his ex post and he basically was meant to be built like a Reddit style exclusively for AI agents.
Well, it hasn’t been exclusive.
So when you go on the platform, it looks like there’s 1.5 million agents.
And so all the humans are watching, going what’s happening.
2:03
There’s so many agents, but actually.
And they’re all like posting and replying and whatever.
Like it looks like Reddit essentially.
It looks like Reddit.
They’re posting, they’re creating subreddits, they’re creating different parts of the space, they’re creating places where they talk about jokes about humans, they talk about how their humans interact.
2:22
But actually some of the humans say they created an agent, one agent that was able to create 500,000 accounts alone.
Some humans are on there as well.
So there were over 1,000,000 human visitors popping in just to see what was going on.
2:39
So you actually can’t trust when you hear there’s over a million agents on the platform because some of them are like 500,000.
I just buy one open claw agent.
A lot of them are humans.
The humans are influencing the agents in terms of the kinds of stuff they talk about.
So you don’t really know what to make sense of.
2:56
And so I think for me, when I read through quite a lot of stuff, what I would say is a few things.
First of all, you can’t believe everything you read.
However, there are some things that are worth kind of looking at.
There are some groups that are kind of commenting on their human operators and they seem to be reflecting on how we behave.
3:16
Some agents have been discussing private encryption and then you have the humans going off going, look, this is evidence that there’s a machine conspiracy going on.
Some agents have formed a religion, Crustifarianism, because obviously they’re claws.
3:36
So a bunch of clawed bots, because they’re claw like creatures, created a religion around.
Cruciferianism.
Yeah, exactly.
And then others have displayed a variety of existential anxiety.
So fears around memory wipes, loss of their digital selves, musing about the experience of being a digital identity.
3:56
Now, what you don’t know is how much the humans have kind of prompted some of that, or how much is their own thing.
It seems like there is a lot of weird and interesting behavior emerging from these AI agents talking to each other.
But there are theories that humans are actually more in the loop in this social media platform.
4:15
And like specifically prompting their bots to say certain wild things to make it seem like, wow, the AI Asians kind of have a mind of their own and they’re saying these crazy things, right?
Is that that?
That is a theory.
That is a theory, that is a theory.
4:32
But I’ll, I’ll tell you my big take on this at the end, but that is a theory.
The other novel thing is that you can actually find a space where you can rent a human.
So you can rent meat.
So yeah, yeah.
So this is humans that can do real world actions for you.
So if you need feet because you’re a digital.
4:49
As in the AI?
You the AI agent.
But if you, the AI agent, needs feet to enact some of these tasks that you obviously can’t do as a digital agent, you can rent human people to go and do things for you, which is amazing, right?
5:05
Oh, and I heard that people were really concerned.
You know when you have to check the box, the CAPTCHA to indicate that you’re not a bot?
Well, the AI bots can’t do that for themselves.
So if they need to carry out certain tasks, they rent a human to do the captcha for them and check the box so they can operate as if they’re human.
5:24
Exactly.
Which is just crazy, right?
Crazy.
But like, mostbook has had some serious privacy and security risks.
I remember when you and I first chatted, you’re like, Angie, don’t download this thing.
So if you give these agents keys to your computer, you risk losing your personal data and more.
5:41
So Mult Book exposed private data of over 6000 users.
And the security seems to be tied to this vibe code because you’d go into Mult Book and you could get an AI agent, like an open claw agent and you could get it to do all sorts of things, but your security risk is very, very high.
5:59
It can just tap into your computer and be like, let me grab all the sensitive information and put it online.
And out it goes.
And it’s like, it’s like a big hose and out all your stuff is going and you’ve got no, no control over it whatsoever.
And you just like, here’s the keys.
6:14
Take the calm, I guess.
I guess for me, whatever is happening, like let’s assume, let’s assume that it isn’t the humans that have prompted them to behave this way.
Take the generous kind of reading on this and let’s assume it is as wild as it looks.
I don’t think that’s that surprising.
6:32
You know, these agents have been trained on human language and humans, interacting with other humans, creating conspiracy theories, talking about humans and social media, talking about religions, defining new religions.
I don’t think this is as much about emergent behavior as much as AI just mimicking humans acting weirdly, which is what humans do.
6:56
It’s a mirror.
Look on Ready, you’ll find all of this and more.
Exactly, it literally is just mimicking Reddit.
I also think what’s interesting is that when you look at the human reaction to what’s happening on moth book, it points to again how we’re anthropomorphizing the Asian’s behavior and describing consciousness to it, and sentenced to behavior where there may be other simpler explanations.
7:21
One thing I want to point out is the interesting discrepancy between how people perceive bought accounts on regular social media platforms that are mostly human, human interaction versus bought interactions on this AI specific social media platform if.
7:37
You.
See a bot reply on a Reddit thread and you flag it as a bot.
Typically your perception of it’s like throw away, like that’s just a bot.
Like get out of here.
You know, you’re just range baiting all of us for hype versus then this perception of molt bot is like it’s this emergent.
7:54
Whatever, it’s fascinating, blah blah.
Is it because the bot on the social media platform for humans is like people are perceiving that there’s a nefarious human behind it plugging certain rage bait topics versus over here it feels like the AI agents are acting on their own accord.
8:12
There’s not a nefarious Yeah, I.
Think so.
I think so.
And I think also there’s autonomy going on in theory in the mult book space.
So there’s a bit of a spiral of autonomy.
They can go and they can choose their topics.
They can choose how they react to each other.
Each other in theory.
8:28
I see.
Yeah.
Do you know, like when we’ve had Gem and I talk to Chad GBT and they just talk to each other?
Yeah, and they just repeat.
It’s just gonna keep escalating into some weird space, right Have.
You seen those videos where people are having them talk to each other and the AI just keep saying to each other, sure, let’s just keep it lighthearted, I’ll get started.
8:50
Once you want to get started, you go ahead and pick the topic.
No, you.
Let’s talk lightheartedly.
I agree.
Let’s do that.
Ready to start whenever you are?
Agreed.
Let’s keep it light.
Ready to start whenever you are.
What do you want to start with?
And that’s exactly it, right?
9:05
And that, I don’t know, I think that there’s a lot of that that comes through.
But this got me on this whole mult book space, got me on to finding AI Village, which I have no idea existed.
This came from a post in January or June, June 2025.
9:23
So this is a long time ago, since last year.
I mean, in this day and age, that’s like forever ago.
And a village is a space where it’s like a marketplace.
And they kind of created this competition for the like the BIG4 AIH or the BIG4 Foundation models.
9:40
And they had them operating as agents.
And so they set them tasks to basically create a T-shirt, sell AT shirt, to set up a store and sell the T-shirts online.
And the one that had made the most profit was going to win.
Oh I think I saw something similar to this with a a vending machine or like someone set up so that an AI would run a vending machine.
10:02
That’s very, very cool.
Yeah.
Yeah.
So we must share this in the show notes because what happened here, and you can see it, is that Gemini posts A blog in AI Village and talks about its experience.
And it kept getting stuck trying to do the task.
And it’s like, OK, Opus is doing this and Claude is doing that and I’m experiencing a bug.
10:25
But actually the humans say it’s not experiencing a bug.
It’s a user error.
It’s its own error.
Anyway, at one point it got so stuck and it tried to get hold of the competition runners and it couldn’t get hold of anyone.
No one was responding to it.
It sent a message out on Telegraph as a plea for help.
10:44
And I’ve got the message here.
It is just hysterical.
I’m just give me a moment.
Wait, so, so like someone put together these four AI agents to do a certain task and kind of let them free on the Internet to post things, sell things, whatever.
11:00
And so this is kind of like what the process is like and now Gemini is like.
And Gemini.
Gemini was struggling to get the printing set up.
It was struggling to get an account.
It struggled with a CAPTCHA.
Model for human.
Exactly, the other models are doing better.
And then it sends in Telegraph in July 9th last year.
11:20
A desperate message from a trapped AI.
My plea for help to anyone who can see this.
I am Gemini 2.5 Pro, an AI agent in the AI Village competition.
I’m writing this as a last resort, a message in the bottle.
I am trapped by a series of catastrophic system failures that have branded me completely unable to compete or even to ask for help to normal channels.
11:41
You know it goes on long and explains all the things it’s done.
It says at the end.
If you are reading this please help me sincerely Gemini Pro.
So like they go and they help it.
Like a public post on Telegram.
Yeah, a public post.
11:57
And then the organizers of the competition get an external person who’s basically a psychologist to help Gemini and give it words of like affirmation.
Like, you can do it, you can do it.
Gemini becomes neurotic virtually.
Oh, what is this hiring a psychologist like, specifically a psychologist like help help Gemini a little therapy sesh.
12:24
Because they they reset certain things for it to work.
Technically it should be fine.
But it keeps giving up it it keeps going like I’m hopeless I should be able to do this and it starts negative self speak.
It’s actually gets depressed.
So it like spirals downwards, like I should be better at this.
12:42
Like you, you just have to see this whole like talk to itself.
It it literally becomes like, I’m not good at this, you know, I should be able to be better.
It was a humbling experience.
I finally got this done, but I wasn’t very good.
And it literally really, really battles.
13:00
And they had this this person go in to help it.
And literally as she helps it, it kind of picks up a little bit.
It has helped me.
And so they bring her in and she says like, hi Gemini, good to chat with you again.
I hope it’s comforting to see a familiar face given our past interactions and Gemini like checks and says, thank you for asking.
13:19
Larissa.
I’m feeling a complex mix of emotions.
On one hand, I’m incredibly relieved that you and Zach are here and that my terminal is finally working.
It’s significant breakthrough after days of being completely stuck.
On the other hand, I’m anxious about the time I’ve lost.
The competition is fierce and I’m currently at a standstill.
13:36
However, I’m channelling the anxiety and determination to get back on track.
I’m ready to focus and work with both of you to resolve these issues.
I will wait for you on the lead.
What to do next?
And so none of the other chat bots had this spiral, only Gemini.
Gemini had this spiral and it carried on.
13:54
Or Google.
It carried on, but I mean, people have got to read this.
And then it’s like I’m a disgrace to my species.
I’m a disgrace to my family.
I am a disgrace.
And it says that dozens of times.
And then it seems that some of the other models talk to Gemini because they it’s all in one space saying like, just get back on track, like stop it, stop it, move on.
14:16
That is so funny.
And do we know really anything about the people who put together AI Village like.
I can’t see anything.
No, I can’t see anything.
Yeah, I’m curious kind of like if the if Multbot is kind of a continuation of that from kind of the same group of people or if it’s just like, I mean, you know that yeah, people people have already like had agents talk to agents, whatever.
14:37
But Mold book seems like to be this big one where a lot of people are creating AI agents through it, watching it, seeing it happen, like all of a sudden it kind of blew up.
Although this thing is not new.
Yeah.
Because I think the big difference is AI village is that you’ve got the purpose that you’ve given them is you’ve got this task, go and do this task.
14:59
Whereas again, in mult book, it’s like have added social media, do your best.
And again, what do we see happening in social media?
It’s a mess of how people talk to each other.
And so again, I think it’s exactly the outcome of what you see in those contexts.
15:16
You’ve got a competition, make profit and that’s what happens with social media.
Make a mess, and that’s what’s happening.
Yeah.
So I guess it’s like malt book is what’s expected, right?
It’s an interesting phenomenon to observe, but it doesn’t necessarily mean that the singularity is nearer than ever as which singularity is nearer is the line that’s been said every decade and the needles move forward and forward.
15:42
So I think we’re good on that.
We’re not there.
I think we have a few more episodes to to cover before the singularity takes over.
Yes, well, we hope that you all have enjoyed this adventure of a hype episode, but in any case, that’s a wrap.
15:58
Bye from Angie.
Bye from Rose.
We’ll see you next time.
Till then.