B (2:12)

It's how you stay ahead of social.

A (2:15)

Engineering and phishing attacks. The number one threat targeting your organization today.

B (2:20)

Ready to move from awareness to action? Request a demo of hrm today@knowbefore.com that's.

A (2:28)

Know be the number four and see.

B (2:31)

How AI is transforming the way organizations manage human risk.

A (2:43)

So in a move that has made not only me, but a lot of people on the Internet kind of roll their eyes, this headline came across my desk the other day. Microsoft launches Vibe. Working in Excel and Word, which, like, as far as I had understood it, Vibe coding became a sort of derogatory term. So it's funny to see it being embraced. And Microsoft isn't the only one embracing the Vibe term. We'll get into that later with Meta's new Vibes app. But this is basically just them bringing AI tools into the Office suite for the first time.

B (3:16)

Right.

A (3:16)

And it's less promising than you'd hope, because I had thought that this would be an area in which AI would be extremely useful in terms of, like, Office productivity apps pushing text around, reformatting things. But the initial results are not all that promising. So I've got an article that we're looking at here from the Verge. It opens with this. You've probably heard of Vibe coding novices writing apps by creating a simple AI prompt. But now Microsoft wants to introduce a similar thing for its Office apps. The software maker is launching a new Agent mode in Excel and Word that can generate complex spreadsheets and documents with just a prompt. A new Office agent in Code, Copilot Chat, powered by Anthropic Models, is also launching that can create PowerPoint presentations and word documents from a Vibe working chatbot. It seems like the implementation of this is very similar to Copilot in VS Code and whatnot. Where you get that sidebar. I don't have this yet. Perry, have you had this pop up for you in Excel?

B (4:15)

No, see, I don't use a lot of the Office stuff, so no, I've not. I believe that would probably have to be turned on by whatever. Like if you're working at a company, if somebody manages the Microsoft implementation there, they would have to explicitly turn that on.

A (4:32)

So interesting.

B (4:33)

It shouldn't be hoist upon anybody that's not ready for it, because some companies are going to say, we don't trust that yet.

A (4:40)

Yeah, and actually they mentioned in this article too that they have been kind of cautious about this because Microsoft Office powers very important documents and things in very important places. Also, ironically, I think I have Microsoft Office. I think I'm on your family plan, Perry.

B (4:58)

Oh, you may be. Yeah. We had some extra seats. Yeah, there you go.

A (5:01)

We'll have to see if there's a toggle for that to give it a shot, because the purported accuracy of this is pretty disappointing. Microsoft says its Agent mode in Excel has an accuracy rate of 57.2% in spreadsheet bench, which is a benchmark for evaluating an AI model's ability to edit real world spreadsheets, which places 8 above shortcut AI, which is a ChatGPT agent that can do Excel stuff as well, and Claude files Opus 4.1. But it's still behind the human accuracy of 71.3%, though, which I wonder how that accuracy number was found for humans. I'm not extremely familiar with Spreadsheet bench, but the thing about applying an average like this in the first place is that like me making a spreadsheet, my accuracy is not gonna be that high because I'm not doing Excel for an important company. Whereas someone is is probably closer to 90, 99% accurate.

B (5:57)

Somebody that lives in Excel and is constantly doing like pivot tables and stuff in it and creating complex formulas. They're probably better than 70 some odd percent.

A (6:07)

And so this is 57.2. Yeah, I would hope so too, actually. Tiny sidebar. Perry, are you aware that there is Excel Esports?

B (6:17)

No.

A (6:17)

Really?

B (6:18)

So want to see it?

A (6:20)

It's actually pretty compelling. People do these challenges and they compete to achieve goals faster than everyone else and it's actually pretty freaking cool. Okay, but that's not related to this. Although maybe, hey, we'll see some AI competitors in that scene in the near future or people using AI in that scene that could be a whole different segment. So agent mode is not something I have had a chance to play with personally. Like you said, Perry, it's not foisted on everybody yet. They say that Microsoft's taken a gradual approach to adding AI elements to Excel, particularly because the data handles power some of the most important parts of business worldwide. But a representative for them says that agent mode lets you build sheets that are auditable, refreshable and verifiable. We've spent a ton of time making sure the validation loop on all these sub agents is pretty tight still, again though, that glaring 57.2%. They said that Agent mode in Word goes beyond the existing writing, rewrite and summarization AI features that are in Word, saying that agent mode in Word turns document creation into vibe writing, an interactive conversational experience. So you basically just have an AI chatbot in the sidebar and then it can edit the document you're in. Nothing too crazy.

B (7:27)

Yeah. I think that companies should be happier to start with using Word, though. You start to wonder if the authenticity of the seeming authenticity of text, like the humanness of the writing, is going to start to wane over time, as people refer to rely on that over and over and over. Yeah, I think there's going to be some backlashes on the writing stuff. I think the thing that people really want AI to help with is like the mundane stuff like Excel, because that's where you really want a lot of critical oversight. You really want to get stuff. Right. You're thinking about precision and that's where a technology based tooling should be able to help a lot. Right. Because Excel isn't about human creativity, it's about precision.

A (8:16)

Mm.

B (8:17)

Versus something like Word, which Should be, you know, if you're in the legal industry, should be about precision. But in many other industries it's about like there's a creative flair that the human behind the keyboard brings to it and they've got the inverse of that. Right. AI is quote unquote good at a lot of the creative stuff. But that 57ish percent on the stuff that needs to be technically accurate, dependable and bulletproof, that's just over a coin flip accuracy.

A (8:47)

And that's not great.

B (8:48)

If your cfo. Yeah. If your company's like reporting numbers to Wall street every quarter and then they find out that the AI screwed up the formula in your spreadsheet, then everything becomes a liability in that company.

A (9:02)

Yeah, completely. Because you're right. Like the ideal thing would be to have something that can automatically like you've put together a spreadsheet and then you say, hey, look over this, make sure I didn't miss anything obvious or mess anything up. And we're definitely not at that point yet. So it feels a bit premature to drop this into the office suite. But again, it seems to be a slow rollout. I wonder if there's any other underlying reasons to why they would try and push this out with such a low accuracy. Because it's not like make it better.

B (9:31)

Right. If we're all guinea pigs for the AI, then we train the AI. So that's Ben Google's approach with this for years. Right. Is you put out the thing that tells you to eat rocks and then it starts getting better.

A (9:46)

Actually. Yeah, that makes a lot of sense. Now that you say that, I feel dumb for not immediately thinking of that. But yeah, they're totally gonna user input and user interactions to help improve these services. I just hope that we don't see any massive failings resulting from early adoption of this.

B (10:04)

We will. We're gonna see some really, really embarrassed people.

A (10:08)

Yeah.

B (10:09)

Especially if they're not transparent about the fact that they used AI and then you find out later that this catast catastrophic mistake was because of something that, you know, some junior analysts did in the accounting department and then just got passed up and accepted by everybody because they didn't put critical review on it. Yeah, not gonna be good.

A (10:29)

There is some language from the representatives here that is making me a little bit nervous. Including this one quote that says it's not just simple assistive short answers, but board ready presentations or documents. It's work quite frankly that a first year consultant would do. Delivered in minutes, which that seems a bit too much of a stretch. So get ready for some embarrassed people.

B (10:51)

Yeah, well, and one of the interesting thing that's come up, and I've seen this, you know, headlines like this for months now, but company after company is, is telling people internally you have to learn how to use AI or else. And then some companies like we talked about a few weeks ago, are basically saying AI is now doing the job of like half of this, you know, mundane workforce like call center employ. So Salesforce, I think had 9000. They let off between 4000 and 5000 of them because they don't need them anymore. But specifically on that first year consultant side, there was notice that Accenture gave a few days ago. I'll pull that up. And it says, Accenture bets big on AI with warning to workers to reskill or exit. Oh. Accenture has made artificial intelligence its growth engine and a survival test for its workforce. In its full year results for fiscal 2025, the global consultancy stated that AI has become a goldmine, but with a catch. Employees unable to adapt to the technology would be shown the door.

A (11:57)

Yikes.

B (11:58)

So that's it. Quote here. We're investing in upskilling our reinventors.

A (12:05)

What in the LinkedIn lunatic sentence? We're investing in upskilling our reinventors.

B (12:10)

Good reinventors, which is our primary strategy. But where reskilling is not viable, we are exiting people so we can get more of the skills we need.

A (12:20)

I hate everything about that sentence.

B (12:25)

That is very much consultant speak. Right.

A (12:28)

We are exiting.

B (12:29)

So hopefully AI. Yeah, hopefully AI will not encourage people to use consultant speak.

A (12:37)

Yeah, I certainly hope not. That is alarming. That is a lot of people and a lot of people's jobs that are now on the line. And I guess they're saying, learn how to use AI or leave, basically, which I guess.

B (12:51)

Well, yeah, and I think it's coming down to productivity, right? It's because if Bob sitting right next to you has learned how to use some kind of tool that makes him twice as productive, then all of a sudden now you're a liability because Bob is making the same amount of money as you.

A (13:11)

Right. And I mean, I guess on the one hand it's not. They're not being replaced by AI, they're just being pushed to use it. So that's a slightly different thing still.

B (13:22)

And it's better than being replaced, right? Because yeah, somebody says, we don't need you to do your job anymore because we have AI. That's AI taking your job if you are refusing to adapt to, to the new world where tools make you faster and potentially better, hopefully at some point, then that is kind of. You're stuck in your own culture, you're stuck in your own way, and you're refusing to be whatever your employer needs you to be at that time. So that becomes something that can be a little bit more justified. But at the same time, an employer that's making that kind of statement, and I'm sure Accenture is. So I don't want to. Even though they're using horrible language, the way they phrase that stuff, if they're making statements like that, you can guarantee they're providing the education and support for somebody to come up to speed on it.

A (14:18)

Right. And I mean, like, I guess a comparison that you could draw would be like, if someone refused to learn how to use just a computer in general for general computing tasks.

B (14:26)

Yeah.

A (14:27)

Like at this stage in life, that's kind of something you can't do. You can't really get away with not knowing how to use a computer. But the difference is that these AI tools are not precise or reliable and introduce a lot of problems on their own. And so there's also that push to adopt has an ick to it.

B (14:44)

Right. Well, they are going to get better. So the hope would be that the workforce is adapted by the time the tools are better so that you don't then spend two years trying to deal with the fact that the future is already here. And, you know, 10 startups have already gone farther, faster than you have because you were waiting for the technology to be perfect.

A (15:05)

That's true. The thing that I worry about is tech debt. I have a friend who works in a programming position for a really large company, and they were telling me about how they are being encouraged to use AI for a lot of their programming and coding tasks, and that while it has made a lot of them more productive, they said their best engineer still their best engineer. Their worst is still their worst. But they're accruing a lot of tech debt really fast in terms of that will happen. Yeah. So in our next segment, we're going to talk more about AI video and specifically about an AI actor. Is that right, Perry?

B (15:38)

Yeah. AI actors are going to be all the rage from what I hear. Or maybe cause significant rage. We'll see.

A (15:45)

Stick around. When did making plans get this complicated?

B (15:58)

It's time to streamline with WhatsApp, the.

A (16:01)

Secure messaging app that brings the whole group together. Use polls to settle dinner plans, send event invites and pin messages so no one forgets mom's 60th and never miss a meme.

B (16:12)

Or milestone.

A (16:13)

All protected with end to end encryption. It's time for WhatsApp message privately with everyone. Learn more@WhatsApp.com.

B (16:24)

All right. Have you heard the name Tilly Norwood?

A (16:28)

Tilly Norwood? No, I can't say that I have.

B (16:31)

She is an up and coming actor, was recently shown to the acting public and has caused a little bit of a backlash. So let me go ahead and start sharing my screen. This is from NBC. This is from their culture and trends section. And I've been hearing a lot about this in kind of the AI gossip world for a while. Says Tilly Norwood, fully AI actor blasted by actors union SAG AFTRA for devaluing human artistry. The AI creation has prompted backlash among Hollywood actors including Whoopi Goldberg, Emily Blunt, Lucas Gage, Melissa Barrera and Kelsey Clemmons. Kiersey and Kiersey Clemens. I learned to read a long time ago and I haven't refreshed my knowledge base since then.

A (17:27)

I feel you, man. This text is small too.

B (17:29)

Yeah, there we go. Let me make that bigger so I can read it. So this came out September 30th says a major Hollywood actors union condemned reports that talent agents are looking to sign artificial intelligence actor Tilly Norwood for representation.

A (17:46)

Oh, wow.

B (17:47)

This is not just the fact that the AI actor exists. It's that the talent was seen as valuable enough to need a broker to push it out. Right.

A (17:58)

It's a program. What do you mean?

B (18:01)

Yeah. Which is really, really interesting. I'm going to show you Tilly Norwood in just a minute in the promo video that has everybody up in arms and the company that's behind that because you and I and I think many of the viewers and listeners to and of this show will recognize that there's some cherry picking of footage that was used here. And a lot of the promo videos, when you look at what this company was doing, are not actually all that impressive. So this is the thing of where you have a number of actors and people in the media who have maybe not been quite as aware of what's going on in the AI world. And then they're seeing these cherry picked scenes. And we'll see some of those cherry pick scenes in other AI generation contexts in a bit. And they're actually getting really good visually. Right. But we also know that to get one really good take, it might take 10 generations of clicking the button or 20 or 30 or 100. I guess the equivalent would be takes if somebody's on set. But there's a lot of complicating factors that the actors don't know about.

A (19:14)

You typically don't do retakes because an actor's mouth falls off mid scene.

B (19:19)

Right. Or grows an arm. Yeah, it's a different thing. But I can understand the worry and I can definitely understand the outrage if you did start to get these agents that are going, I'm gonna sign Tilly. That would be frustrating.

A (19:38)

Part of me, my initial. And I don't know anything about this story, but my initial gut instinct is that it's the novelty of it that is attracting people at this point because I have a hard time believing that Tilly is a very competent actor.

B (19:51)

Yeah.

A (19:51)

Well.

B (19:52)

And the clips are cherry picked, so I'll go ahead and play this. Three seasons and a podcast. Meet Tilly Norwood, the world's most controversial new actor. Can she cry on Graham Norw?

A (20:05)

While she might seem real on screen.

B (20:07)

And on her social media, take a closer look. Can you tell Tilly Norwood 100% AI generated?

A (20:15)

Yep, she's fake.

B (20:16)

Tilly Norwood is the first creation from a new studio that develops AI actors. And this morning, her mere existence has.

A (20:24)

Sparked outrage and concern about the use of artificial intelligence in Hollywood.

B (20:28)

We have so far lost the plot. What are we doing? The backlash to Tilly Norwood igniting over.

A (20:34)

The weekend after her creator, Dutch actor and comedian Elena van Der Velden, said multiple talent agents have shown interest in signing the AI actress, according to Deadline.

B (20:44)

That news prompting real actors to speak out in protest. Emily Blunt calling it really, really scary in an interview with Variety urging agencies to please stop.

A (20:53)

Van der Velden posting a lengthy statement.

B (20:56)

Defending the invention, comparing it to animation or CGI writing. Tilly Norwood is not a replacement for.

A (21:02)

A human, but a creative work, a piece of art that sparks conversation. Whoopi Goldberg among those responding.

B (21:09)

The problem with this, in my opinion, humble opinion, is that you are suddenly up against something that's been generated with 5,000 other actors. So it's a little bit of an unfair advantage. The debate over Tilly Norwood, just the latest in the battle between creative artists and AI. Last year, Scarlett Johansson threatened legal action against Stop that there. But, yeah, you know, I think you get the idea. Right.

A (21:41)

Yeah. I don't really get what Whoopi was trying to say about it being an unfair advantage.

B (21:46)

Yeah.

A (21:47)

Or.

B (21:47)

Well, I think it comes from not understanding. Right. She's under. What she's thinking about is the training data that's. That's involved with this is that you can take, you know, all the performances of every actor in the world and you can train a model based on that. And then you can say, you know, cry like this person or laugh like this person, or, you know, be the spitting image of Heath Ledger in this performance, but with the looks of Cary Grant type of thing.

A (22:19)

But is that an unfair advantage or is this all kind of deeply stupid? Because what who like to claim to be the creator of Tilly Norwood is a really. Is a stretch to me because the person is probably using a commercially available tool and just feeding in prompts and like style guides.

B (22:36)

I can tell. I can tell.

A (22:38)

So how can you own that you're not the creator of that? Anyone could make Tilly. So that seems just dumb to me.

B (22:46)

They're capitalizing on the fact that most people don't know how this works. And I think that when they said, hey, multiple agents are clamoring to hire Tilly, we don't know that that's true. It's just the person that runs this company said that. So when it comes down to it, but it's all a little bit of a skin or a lot of fronting over something that may not really be very good. Let me share some of Tilly's Instagram page.

A (23:17)

It feels to me like if I tried to start a business selling you, like, I'll find images for you and then what I do is I Google things and I grab Google images and send them to you. Like, this can't be sustainable.

B (23:27)

But anyway, so Tilly on Instagram has 46.7 thousand followers. She actually doesn't have very many followers. For AI social media influencer, there are some that are fully AI that have millions of followers. In fact, like the first one was back even between, I think the one that I talked about in the book, I think was 2015, 2016 timeframe. And it amassed millions and millions of followers. So she doesn't have a huge following. There's very few photos here. Some of these you look at and you can immediately tell, like this one here, that looks AI to me, this has like 142 likes. There's just a picture of like some clothes on a bed. And then I think they came out and started saying that she was AI. And you know, the, the photos at a glance are not that bad. But when you get to like the reels, what you start to see is that there's not much of a. There, there, there's. I'll go ahead and play one. There's no audio to this. It's just a sample, it looks like, of them taking a frame of an AI generated image of her putting it into something like cling or VO and saying, have her wave.

A (24:43)

Nice watermark.

B (24:44)

So you just have the picture of her waving the bottom right corner. Yeah, you do see it. Yep. Yeah. Let me go over here. This is a video that they showed of her doing like a showreel. And there's no audio in this, which means that it was probably just generated using different tools that didn't support audio generation at the time. And some of these are like clearly bad composites. They're just not great. Some of them look pretty good.

A (25:12)

Yeah. But it's remarkably high quality. Like, we have come a long way in AI video generation. As we'll talk about more in this and the next segment, it's what has she. So she hasn't even been in any movies yet? This is all speculation.

B (25:26)

No, she's not been in a movie or anything. It's all pr.

A (25:30)

So there is, you're right, there's not much of a. There, there. There's like nothing here except trying to generate stink about this thing that they know people won't like, which is a fully AI generated actor. And then I'd be curious to know which talent agencies are clamoring after this because.

B (25:46)

Yeah.

A (25:48)

I don't buy it.

B (25:49)

So I'm gonna go to Particle 6's YouTube page real quick.

A (25:53)

Okay. And that's the company behind her.

B (25:56)

Yeah, that's the company. But behind Tilly. What you'll see is that there's only three videos. Oh yeah, let me, let me go to the thing where they announced Tilly and we'll end with this because it kind of shows how they hype it and some of the things that are interesting and problematic. One of the other things that you'll see with this video, if you take a look at it, this is a two minute video. Well, we won't show the whole thing, but they're saying, hey, everything in this video is AI generated, which is fine. They're disclosing that up front. What you'll see in this is that if you've been following anything that I or other creators in this space have been doing, this is no better than any of that. So this is not something where a big Hollywood funded company has been spending half a billion dollars to generate Tilly. This is something that a creator at their house could spend maybe several hours or a few days, or maybe even a few weeks on and come up with a consistent character that you can put multiple scenes, a decent voice, some facial animation where you can take control of that. But all of this, they've only strung together a lot of these little 8 second clips, which means there's probably a lot of VO3 involved in this. Combined with some stuff where they may have been using like hey gen avatar 4 or hey, Jen's traditional avatar. There's lots of technologies that could be squished together to create a fairly believable sequence, but it's nothing new. And that's the thing that I want actors to know. There's nothing new here. And you can plug in and you can understand this right now.

A (27:37)

And there's. Yeah, they're doing nothing special. They don't have anything special. This is not gonna last. This is a flash in the paint.

B (27:44)

They created an interesting showreel and PR splash. So we'll start with this. Every voice that's in this, every character is AI. Some of them very noticeably so. And then I think what you'll find is that they probably spent more time attention, did more upscaling with the initial images for Tilly and things like that. We all knew TV was dead, but thought, why not squeeze in one last development meeting? We had one rule. No missing children, no dead women and no reboots of Heartbeat.

A (28:16)

We'd pitched this beautiful comedy.

B (28:18)

Two sisters running a funeral parlor in Margate. Warm, weird, very BBC2. The commissioner said no.

A (28:24)

AI generated 100 better ideas in minutes.

B (28:27)

Perfectly aligned to channel data, viewing figures and optimized for the audience.

A (28:31)

The winning AI format.

B (28:32)

I know what you streamed last summer.

A (28:36)

An interactive thriller built from your streaming.

B (28:38)

History and delivery orders.

A (28:40)

What?

B (28:41)

And who could forget the mental health format? Britain's Got Breakdown. Audience votes on who gets therapy. Ian Katz loved that one.

A (28:50)

I remember saying, you do realize this is for BBC2.

B (28:53)

It tested ridiculously well, particularly among men.

A (28:56)

Who shop at Screw Fix and women.

B (28:58)

Obsessed with gut health.

A (28:59)

All made for less than the catering budget on the bear.

B (29:01)

It cast and budgeted itself. Only bit it struggled with was consent in romantic scenes. We just ignored that. It wrote the whole script.

A (29:10)

Claude 3 Gemini GPT we called it.

B (29:13)

Skip forward a little bit for the sake of time. And we get the introduction of Tilly.

A (29:17)

These people are deeply unfair.

B (29:18)

I generated by some company called Particle 6. She'll do anything I say. I'm already in love. Girl next door vibes like in for Sunday Roast.

A (29:29)

Went to drama school and got BAFTA optimized.

B (29:31)

But can she cry on Graham Norton? Of course she can. And it'll be clipped, subtitled and monetized on TikTok by lunchtime. We're all going to hell. Got recommissioned, didn't it?

A (29:42)

Three seasons and a podcast by an AI commissioner.

B (29:46)

Much faster. But hey, at least it wasn't a reboot. Yeah. So get the idea.

A (29:53)

Are they trying to like, tongue in cheek, say, hey, we know this is terrible, or like, what a not very.

B (29:59)

Funny particle 6 is run by a comedian.

A (30:02)

Okay.

B (30:03)

So, yeah, I think there's some tongue in cheekness of that. And then also poking fun at the industry. Right. Because it's a lot of production. It's not necessarily about creativity, it's about profit. And so I think that there's some nods to that.

A (30:17)

I feel like it's a bit beyond the pale to be making jokes about how this AI generated influencer crying. And Graham Norton will be clipped and monetized on TikTok within minutes because that's how people actually talk about things they're actually trying to sell. It's impossible to tell what's a joke and what's not. Also, the things that were clearly jokes were a bit tasteless in there. The only thing from that video that I really enjoyed was we're all going to hell.

B (30:41)

That certainly feels we're all going to hell. And the at least it's not a reboot. Right? The very last end. So there were a couple nods to the reality of where the creative industry is right now. And it's probably that really got everybody's stomach turned. Right. It was the crassness of the showreel for this.

A (31:01)

I just think you have to be pretty deluded to see what they're doing and think that it's anything long lasting or anything actually impactful, like it's not even worth getting mad about. It's something that should just be ignored.

B (31:13)

Well, so I think the difference is, right, is from an actor's mind, somebody that lives in Hollywood and has dealt with, with a lot of compositing, a lot of CGI creations with characters like Yoda and all that that used to be practical effects and then they were CGI and they become very much easier for, I guess the production company to make the character do whatever they want is that there's no separation in the actor's mind between what generative AI can do and where the special effects industry might be able to take that. And so they're immediately seeing Tilly and assuming that Tilly can do like entire scenes or maybe even carry an entire movie. And those of us that have been playing with the tools that are around know that that would be like a super arduous process to get anything usable at length.

A (32:09)

It would be yeah. It would be a super arduous process and people don't want it. Also, like, I mean, I guess. Right. I guess there is a future in which it is so good that these AI creations give really compelling performances to the point that we don't care anymore and we're just invested in the story, I guess. But I don't.

B (32:30)

Yeah.

A (32:30)

No one is going to love an AI actor because you don't. Like. If, let's say that Tilly was capable of carrying an entire movie, I could then make any Tilly movie I want. Or you could. There's not a Tilly that you can fall in love with. There is this fake thing. It's not like I love Ian McKellen and I want to watch all of the movies he does to see what he brings to it. If Ian McKellen was a digital construct that could be shaped however anyone wants to, then I probably don't love Ian McKellen. I just would love an individual movie. Like, there's not. There's not that layer I think is important and not from like an aspect of. Oh, it doesn't have a soul. Just from like, it's not anything they didn't make. Veo and Sora and whatnot. Yeah, the other thing too.

B (33:17)

It'd be interesting to see where that goes.

A (33:20)

The. What is happening with our video. Wow. Okay. Well, we have video issues and we're gonna talk even more about AI video in the next segment. So I think let's cut it off there. We'll take a quick break. I'm gonna uncheck some options and hopefully break this cycle. Audio listeners just know things are going badly on the YouTube side of things. We will be right back. Upgrade your laundry routine with a durable and reliable Maytag laundry pair at Lowe's. Like the new Maytag washer and dryer with performance enhanced stain fighting power designed.

B (34:04)

To cut through serious dirt and grime.

A (34:06)

And what. What's great is this laundry pair is in stock and ready for delivery when.

B (34:10)

You need it the most.

A (34:12)

Don't miss out.

B (34:13)

Shop Maytag in store or online today at Lowe's. Tito's handmade vodka is America's favorite vodka for a reason. From the first legal distillery in Texas, Tito's is six times distilled til it's just right and naturally gluten free, making it a high quality spirit that mixes with just about anything. From the smoothest martinis to the best bloody Marys. Tito's is known for giving back, teaming up with nonprofits to serve its communities.

A (34:42)

And do Good for dogs.

B (34:44)

Make your next cocktail with Tito's, distilled and bottled by 5th Generation Inc. Austin, Texas. 40% alcohol by volume. Savor responsibly.

A (34:55)

So as much as we're all trying to avoid AI slop in our video sharing and social media platforms, maybe you're the kind of person who wants to scroll nothing but slop. And in nothing but slop I realize I have said butt slop out loud now, and that is making me giggle. If you want to scroll nothing but slop, Meta has you covered. They've just launched a new short form video feed of purely AI generated videos that you can scroll called Vibes. I'm looking at a TechCrunch article that we will link here in the description that does a pretty good breakdown of it. It opens with this In a move no one asked for, Meta is introducing Vibe, a new feed in the Meta AI app and on Meta AI for sharing and creating short form AI generated videos. Think TikTok or Instagram Reels. But every single video you come across is essentially AI slop. And that's pretty much it. I'll jump over here to the actual Facebook press release, which says largely the same thing. They do say the feed will become more personalized over time, so there's some algorithmic curation going on. If something catches your eye, you can create your own videos or remix what you find on the feed and share it with friends and followers. So it really is AI Brainrot, TikTok. And now in the same app that you manage your Meta glasses and all that other stuff with, I'll pull it up on the screen so we can just explore the feed together for a moment for those of us watching. But it is pretty much. I mean, if you've been on Facebook in the last six months or last year, really more of the same. Remarkably pretty eye candy, surreal, weird AI slob videos, an unending feed of them, and you can click and customize and remix them and adapt them. Most of the response to this has been pretty much what you expect. People saying things like why would I want this? And great cool, more brain rot. So that's a thing if you want to check it out. Meta AI Vibes. I wouldn't necessarily encourage checking it out because there's not really a whole lot to see here. I'm not going to log in and demonstrate the remixing features or anything like that because I don't really. I don't have a Meta AI account and I don't feel like today is the day to start. But I do want to talk more broadly about AI video because we've had other big breakthroughs that aren't just this, including the release of Sora 2, which is a pretty big deal because it is remarkably good. You've been playing with a Perry. I have a test from you. We'll play later. What have you thought so far?

B (37:17)

Yeah. So I think Sora 2 is way closer to what we had hoped Sora 1 would be. Right. Is we saw those Preview clips from SORA1 a long time ago where they had like, the drone going over Big Sur and the lady walking through Tokyo and all that, and everybody's like, oh, my God, AI videos here. And it's really good. And then the real Sora came out and it had been surpassed by other models by then.

A (37:42)

Yep.

B (37:43)

When Sora 2 came out, it's like, really. I think it's fixed a lot of the coherence problems, a lot of the physics problems. The resolution for each video is really good. And they've got two things that I think are direct shot against Zuckerberg's bow and even Elon Musk. Right. Is they released their own social media platform with it, which is the Sora app on iPhone that we can pull up in a minute, though. It is. Their servers are getting hit hard. It's like barely functional right now.

A (38:17)

You were trying to show me right before we started the show and we could not get videos to pull up. So we'll do a demo, but we'll see if we can because it's definitely getting used a lot right now. Yeah, the.

B (38:28)

Yeah, that. That'll get fixed.

A (38:29)

I've.

B (38:30)

I've had it work for me enough to know that it does work when it's not just in an onslaught of traffic. They've. They've essentially dosed themselves. But the. The social media app is an interesting play against Meta and. And Grok, and I think there's maybe some good to that. We can talk about that in a minute. Even though it is AI slope 24. 7 on that. Right. It's. It's an AI platform for creating AI slop that then can just be doom. Scrolled through in the same way that every other social media platform can. But maybe there's an upside that we can touch on. And then the other thing is that they've added sound and, you know, voice and all that stuff that we got from VO3, which is really good. And it's. It's okay. But the. The killer feature that they added was the ability to create these things called cameos which is why you're probably seeing Sam Altman from OpenAI Everywhere right now is because he's unlocked his cameo to be used by anyone. There's an interesting side effect to that too. It's because. And as we flip through this, you'll see Sam Altman in like all these different scenarios. I think Sam Altman will be the first person in the world that has plausible deniability for anything he's caught on video doing the rest of his life.

A (39:49)

Because everyone is deep faking him essentially into everything now with his cameos feature. Yeah, I thought the move to make the social media platform with it was particularly interesting. Before we fully pivot to talking about that platform, let's just talk a little bit about the improvements between this or like VO3, the physics are remarkably much better and even the physics of facial expressions and such are much more better and believable. I know a lot of VO3 videos tend to have like these weird, exaggerated, floaty sort of expressions. They seem to have fixed that. And a lot of the sort of anything happening in the real world, like this guy standing on two horses, it actually is pretty well done. The physics of it. The realism has definitely gone up by a pretty large margin, I would say.

B (40:39)

Oh yeah, I think so.

A (40:41)

Definitely at the point that it's visually fairly indistinguishable in some cases from a real video. The audio however, is still really weird. And phase y the audio generation it.

B (40:52)

Can be for sure.

A (40:54)

We'll see in your test actually that I'll show. Perry made a little reel of SORA videos and you shared it on LinkedIn.

B (41:00)

So let me. The first one is that face. If you don't recognize it though, I think anybody that watches the show or listens to it will recognize this, the face and the voice as Sam Altman, the CEO for OpenAI but had to cameo him first.

A (41:15)

You did a lot with Sam in this, which is very fun.

B (41:18)

Oh yeah.

A (41:19)

So, all right, so I'll full screen this and yeah, we'll just share it. So Perry, this is your minute long reel featuring some very fun clips made in Sora.

B (41:26)

When I think about know before I think about this guy, I think his name is Perry Carpenter. Yeah, that's it. That guy is awesome. Like I don't know what he gets paid, but I guarantee it's not enough.

A (41:36)

Don't share your past.

B (41:37)

Lock your screen real fast, verify the link, stop and think. Two factor is the key. Keeps our data bug free.

A (41:43)

Think before you click.

B (41:44)

That's the security trick idea that keeps Coming up, when we talk about synthetic media is what's called the Liar's Dividend. When people realize how easy it is to make something that looks real but isn't, they start to doubt the things that actually are real. Yesterday I grabbed a burrito from a truck stop shelf. It was a day old markdown.

A (41:58)

Couldn't help myself.

B (41:59)

The sticker said discount and big red shout a dollar for beans and cheese. I had to check it out. Now the bargain's blowing back in a brass like spree. My stuff takes are getting so realistic that it's hard to tell what's real anymore. Yeah, they can erode trust in video and even voice recordings. You can make anyone say anything and that undermines public trust. The top secret number I just made up. Hi, I'm the guy who wants your passwords to be longer than a grocery receipt. Let's keep the bad guys out of your digital cookie jar. First manager. And make those logins wild and unique next. Tap that little.

A (42:23)

Yeah, this video of you is freaky. It's like math Perry should.

B (42:26)

Every step echoes like this place remembers who walks it. These symbols, they're older than anything I've studied. And that light.

A (42:33)

Okay, for the audio listeners, that might have been somewhat hard to follow because it was a collection of several individual clips. I think your clip that you did with Sam Altman about the Liar's Dividend felt particularly pointy slotted in there. Also, that one is completely, totally. I would think that was real if it weren't for the audio being. And I'm sure that it's not going to come across because we bounced this down into like 192kbps mono MP3 by the time you hear it. So I would encourage you to go check it out. We'll link to it and just like listen in a pair of good headphones to the audio. It's crunchy and fazy. Things aren't placed in the stereo field where you would expect them to be, but on like a casual viewing or through phone speakers even. I feel like it might not be that noticeable. And the video segment is indistinguishable from reality in a lot of cases.

B (43:20)

Yeah, I mentioned that. Like when we get to the Sora app, I think that there's a positive side to it. And even the Vibes app is that you know when you're on that platform that everything you're seeing is AI generated.

A (43:33)

Yes, I was thinking about that.

B (43:34)

That's a positive.

A (43:35)

Yeah, because you're so.

B (43:37)

It's like A self contained blast radius.

A (43:39)

You're in context of knowing what it is and like that might build familiarity with some of the tells and the capabilities of what is possible.

B (43:48)

It could. So I'm of two minds about that. That was my first thought is it might build familiarity. My, my second thought is that everybody's been in a room where something smells really bad and then like after a couple minutes you get used to it. And I'm wondering if we're all going to get so used to AI video that when it slips out of its contained area and makes it in its way some somewhere else, we don't smell it anymore.

A (44:17)

So sort of a desensitization.

B (44:19)

Yeah, that would be the better non graphic way of putting it. It's not like the fart that you've gotten used to smelling. It's the thing that is now in another place and you can't see it because you've been so desensitized to it.

A (44:35)

Yeah, like if you just smell farts all day you're not going to be as startled by it. I mean and the other thing is like these things will breach containment and.

B (44:42)

These apps, I mean they all have a download button on.

A (44:46)

Oh yeah. And also the apps by the nature of being AI apps are going to be pretty exclusively used by people who are like interested in AI, Right? Like yeah, like I don't have interior design apps on my phone because I'm not an interior designer and I don't really have any interest. I mean I have passing interest in that. But like people who are not into AI who would be equally fooled by false videos attempting to like manipulate them are not going to be on the Sora app and not build that familiarity. But it's certainly with the, you can create things on the app and so I feel like it's gonna give an easier access point. We're gonna see a lot more in even more videos I think elsewhere even though there is this app. So yeah, I'm not entirely sure. Let's look at the app and the app ecosystem real quick.

B (45:34)

Okay.

A (45:36)

You have it on your phone. You get that ready and I'll do a dramatic reading of some of the reviews. There's Sora by OpenAI mostly has. It has a 3.9 out of 5 on the Apple App Store and some of the reviews are a bit, a bit biting. Like this one that says mission driven nonprofit. Huh? OpenAI is a nonprofit founded to protect humanity from the threats of super intelligent AI. I fail to see how an infinity of TikTok rot and AI slop blended together serves this mission. The fastest growing company in the history of the human race has decided monetizing your attention is the best use for their age defining technology and nation state level compute resources. Not a great omen, my fellow humans. I kind of tend to agree with that.

B (46:19)

Right.

A (46:20)

And then other people are saying like, you'll become addicted. This is the most addictive AI tool, trust me. Five stars. Five stars. Good news, bad news. Who's to say enlightenment or collapse? Only history will determine that. Some people saying it doesn't work. And as you were saying, we're having issues because they probably have dossed themselves. Want to take a look at it? You got that ready?

B (46:41)

Yep. Actually, let me share my screen first. I'll show you the web version because you can go to sora.chatgpt.com and if they're totally not completely dosed, you'll be able to see some of what's available there because you do have your social media style feed and you can see like up in the top left of the one that I'm showing you, there's like a recreation of Martin Luther's I have a Dream is talking about AI. There's a guy that I know named Darren from hack Brown or Hak5. It's got him and Sam Altman and others doing like a sneaker style infiltration. Sam Altman is in a lot of these.

A (47:21)

It's very similar to that meta app also Blast from the past with Sam Altman as the hide your kids, hide your wife guy. An old 2009 meme reborn with AI. What are the highest engagement numbers we're seeing? To get an idea of how many people are using this, I see like 1000, 2000 views as the highest I've seen.

B (47:38)

I mean, it's hard to know because it's just my algorithm. I'd have to go in as somebody else and see. But I mean, if I log into my profile page, you can see I have all of 6 likes and 9 followers. So maybe those of you that want to get on can change that. You can add a little bit more to my ego.

A (47:58)

Follow Perry on Sora.

B (48:00)

Follow Perry on Sora. But when you look at like the Sora announcements, what they're saying is that the for you page is it's a combination of stuff. So it's stuff that you're looking at and liking, engaging with, similar to TikTok, but it also has some understanding of your conversation history with ChatGPT.

A (48:20)

Oh.

B (48:20)

So whatever it understands about you from that starts to make its way in as well, which is really interesting. Yeah. A little bit scary as well.

A (48:28)

I missed that detail.

B (48:30)

I would have assumed that those would have been siloed off, but no, they see that as valuable in the attention economy, which I would agree it is valuable in the attention economy, but I would have thought that they would have been separate and they're not.

A (48:45)

I don't love that little detail.

B (48:46)

Yeah, no, I don't love that either. But the totally, I think, different and addictive thing is this cameo feature where you can have yourself or you can have anybody that's unlocked their cameo in different videos. So you can see ones that I've made of myself here as we're scrolling through. I have Sam Altman in like a Jerry Seinfeld type of environment. Talking about farts. You can see also see that the page is crashing. Oh, no, exit that.

A (49:14)

Yeah.

B (49:14)

So the got the. Aw, snap page.

A (49:17)

Some stability issues with their. Their app, probably due to traffic.

B (49:21)

Yeah, there's definitely some issues with the scripting on that page that are causing it to get stuck all the time. Let me. Let me go ahead and share from my phone though, and I'll go over to my page there. You can see those of you that are watching. You can see my phone being projected. And what you have here is like all the different things that you've posted. If I go to the for you page, it's still broken right now.

A (49:46)

Oh, yeah, look at that. Nothing.

B (49:48)

I'm scrolling through darkness before. Before we end. Let me see if I can show you how fast it is to create one of these cameo Personas. Because that's the one thing that should be a wake up call to anybody. OpenAI tries to lock this down pretty hard to where it's hard for somebody that's not authorized to create or use one of these. But what we have to understand is that's them trying to impose a guardrail on technology. And which also means that somebody else can release essentially the same functionality with no guardrails at all. And so I'll go ahead and I'll go and I'll edit my cameo Persona and I'll do a retake. And what you can see is that it just gives camera access and it says start recording and say the numbers on the screen. Okay. So you usually think that, like, to create a deep fake audio, you need a decent amount of audio to do that. What this is going to show us is that it only takes reading three sets of numbers. And they're also, of course, recording your facial expressions. The way your mouth moves, the way that you articulate along with those numbers so that they can represent that. So I'll do that. 88, 95, 61. And it says turn my head to the left.

A (51:07)

Turn your head to the right. Just like a face id. Wow. That's it.

B (51:11)

Yep, that's it. And then it takes about a minute to create it.

A (51:16)

Is it the same numbers each time or there's a difference? Do you remember?

B (51:19)

No, it changes every time.

A (51:21)

Okay.

B (51:21)

Yeah. Because otherwise that would be a big vulnerability.

A (51:24)

Right, right.

B (51:24)

Because I could pre create a video of Taylor Swift doing that in another tool and then could upload it and.

A (51:30)

Then boom, you get Taylor Swift and Sora. Yep, exactly. It currently says processing and it is nearly done already. It says this may take up to one minute.

B (51:39)

It's nearly done. Yeah. So that is that. Great, you're done.

A (51:43)

Wow.

B (51:43)

That's how fast. And you've seen the quality of the outputs of these. They look like the person.

A (51:48)

Yeah, they really do. So if you hit continue, then what's next? You just.

B (51:53)

And I'm going to say only I can use it. That's the key for this one, right? Yeah, yeah. So then I could essentially just open that and I can say running across the rooftops of London buildings at night in an Assassin's Creed style scene. So I'm going to do that as a prompt for a Soar video. It's going to kick that off. It's going to go to my drafts and I've got a couple that I pre created here. We're not going to be able to hear the audio for these, but for anybody watching, I'll just go ahead and click it so we can see what it looks like.

A (52:27)

So this is a similar prompt you had done before.

B (52:30)

Nevermind.

A (52:30)

Oh, it's not going to load, is it? We're not going to be able to watch this.

B (52:33)

It's not going to load.

A (52:33)

Yeah, because they're serving.

B (52:34)

Watch it. Well, let me download it, see if it downloads. Yeah, See if that request pulled up on camera reel.

A (52:41)

Oh yeah. Okay. It's showing you sprinting across the rooftops in like a cool black leather assassin sort of coat. And then it's got your face, you look at the camera and it's not as uncanny as the like TikTok influencer Perry from the earlier clip.

B (52:56)

This is pretty good. No, well, and like the one that had me running through or walking through the building commenting on stuff and the other one, that one looked and sounded a lot like the way I would do things. Too.

A (53:07)

It did. Cloning your face and your voice from that short clip and those three numbers. Oh, yeah, look at that. Look at you go. And the physics of the wind on your hair. I mean, that jump was weird looking.

B (53:21)

And the face that I made at the end, but I guess that would be like a GoPro style camera too. Distorting your face.

A (53:27)

Yeah, it's definitely got the wide angle effect.

B (53:29)

Yeah. Some of. Some of the physics are just not there.

A (53:33)

For coming from you saying three numbers and turning your head each direction. That's pretty nutty.

B (53:39)

Yeah, exactly. Yeah. I was trying to do a special Cartman thing for somebody. So that's why all the Cartmans are here. That you're saying.

A (53:47)

A testament to how many drafts it takes to get a good usable output.

B (53:51)

Anyway, rather than waiting on this to be done, what I'll do is we'll take the output that happens in just a minute from this. I will give that to you. You can stitch it into the very end of this in post and we'll know what the exact prompt output was that I just gave to this with the new cameo of myself wearing, I guess, this shirt and stuff instead of the shirt that I was wearing in the other ones that we saw.

A (54:18)

I'll overlay it over the credit sequence.

B (54:20)

All right. And I'll do this. I'll say telling everyone, thank you for watching and listening to the fake files.

A (54:26)

Perfect. There we go. And we can stick that right at the end. All right, we should pivot on because we have gone way too long on this segment. And we will make our moves into the next segment where we're going to talk about some relationship woes with chatbots. Is that what you said, Perry, at the start?

B (54:42)

Yeah. Well, it's kind of three different stories related to chatbots and AI embodiment in different forms.

A (54:52)

All right, we will be right back with that. Don't go anywhere. This episode is brought to you by Indeed. When your computer breaks, you don't wait for it to magically start working again. You fix the problem. So why wait to hire the people your company desperately needs? Use Indeed's sponsored jobs to hire top talent fast. And even better, you only pay for results. There's no need to wait. Speed up your hiring with a $75 sponsored job credit@ Indeed.com podcast. Terms and conditions apply.

B (55:34)

Okay, in something that is very similar to dumpster fire territory, we're going to talk about a couple things related to chatbots and relationships, and we'll even end up with embodied devices and robots. So let's see where this goes first thing. So just consider this some quick hits. We don't need to spend a lot of time on any of this. So the first one is an article from Wired that came out October 1st and it says chatbots play with your emotions to avoid saying goodbye.

A (56:05)

Oh.

B (56:05)

Which is predictable.

A (56:08)

I mean, I know that ChatGPT always throws two questions at you at the end of every prompt to try and keep you going.

B (56:14)

Yeah, we've talked about that before. Right. It's kind of like most of them don't want to stop engaging. They'll finish the thing and. Well, it depends. Like chatgpt and voice mode typically will just say hey, and if you ever need anything else, feel free to ask me a question. It's like they're wanting to stop that because that's a process hog and I would assume if they did video avatars for the time being, they would also want to try to say, and if you need anything else, I'm always here to help. And they wouldn't necessarily prompt you into another five minute conversation.

A (56:47)

Yeah.

B (56:48)

Do you want me to a text version of this? Yeah, the text version of that's always. Would you like me to do. Help. Help you by doing this other thing that I now suggest. So that's the, the Chat GPT side of it. But the implications of this go a little bit deeper and darker when you start thinking about apps that are not just like chat based helpers from like an admin perspective, but it's more like the relationship.

A (57:16)

Right.

B (57:17)

Those definitely do not want you to disengage because they're all about increasing the stickiness. And to a certain extent Chat GPT is all about increasing stickiness as well. They want to keep you using the platform, they want to show value. But when the value isn't the work product that comes out of it, the value is the attachment and the relationship, then that starts to get a lot darker. And that's what we're seeing across several different apps. This shouldn't be a surprise because we saw just about a year ago some research from Mozilla. Yes, they're still around Mozilla's AI Research Org that said that these romantic chatbots are monetizing everything that they can. They're siphoning, you know, they're keeping you engaged, they're siphoning off your data and then they're, they're essentially selling that to the highest bidder or they're using that to keep you on the platform for longer. These are totally self serving apps that are out there, they're not there for your benefit. Or if they are, that's a side effect to their business model.

A (58:23)

Right.

B (58:25)

They are there to keep you engaged, keep eyeballs on screens or keep you engaged with your voice, or to keep deepening the relationship so that they can continue to monetize through that monthly fee. They can monetize through ads, or they can monetize through the data that they've siphoned off and sold to somebody else.

A (58:42)

It's, I mean, similar to dating apps, right. Where like dating apps don't really want you to get a good date and leave the platform, they want you to keep using the app.

B (58:49)

And you got to think in this, like when you play with some of these that I've not played with a pure relationship app before, but I've played with the ones like, I guess the closest would be the, the Maya and Miles chatbots from Sesame. And every now and then you'll will be in a conversation with one of them or you know, like the test environments that I do, and they'll ask like a very penetrating personal question.

A (59:17)

Really?

B (59:18)

And I never answer those, but I can only imagine that the reason that they're asking that isn't for my benefit. It's so that they can get that nugget of data that might be useful somewhere else.

A (59:31)

Yeah.

B (59:32)

And maybe that's me being overly cynical, but I understand the business model of Silicon Valley social media platforms. And so I'm going to approach this from our very cynical mindset.

A (59:43)

I don't think that's being too cynical. I think that's pretty realistic given everything else from the pre AI Big Data era. Like anything you can get from the consumer to build a profile to sell them something or something you can sell of theirs. Yeah.

B (59:59)

And you know, I was playing with, with the Maya one and doing some of the jailbreaking stuff that we've shown on this before and just kind of really pushing against the boundary, trying to get it to do weird things like rant about Sesame or plot something that's horrendous. And every now and then you'll push against that guardrail and then it comes back with an intensely personal question. It's like, I sense you doing this. What within yourself do you think think is the motivation for that?

A (60:29)

Really?

B (60:31)

Which is. That's a really interesting question, but it's like, I'm not going to answer that.

A (60:36)

Right.

B (60:37)

Like, and, and you'll also notice them trying to push into some other areas. I can't think of one off the top of My head. But I've. I remember getting some really intensely, almost, you know, over the bounds, personal types of questions that. That would reflect, like a false sense of intimacy. And because I don't have that, because I know what I'm talking to, I'm not going to fall to giving that so that it deepens the relationship or the perception of the relationship. But it's interesting that the bot seems to have that ingrained in its personality to try to go there that it want.

A (61:13)

Yeah, it wants you to engage and to form this relationship. I mean, essentially, that's the point of it. Right, right. In those contexts. Yeah.

B (61:22)

So let me. I'm going to move to one other. This is a subreddit and it is showing the investment of one of the companies that we mentioned last time. Right. So friend dot com. The little friend wearable that I've still not taken out and set up yet.

A (61:39)

Oh, you haven't?

B (61:40)

I'm having a hard time bringing myself to doing it. I still need to, though, as part of one of these tests.

A (61:45)

Yeah, we got to.

B (61:47)

They actually have the physical devices now. We showed one of them last week. We will be testing it. But one thing they're big on is marketing. And so they spent $1.8 million on the domain name friend.com and now that they've got the hardware and it's shipping, they spent a million dollars on print ads in New York and it's causing a stir.

A (62:09)

Yeah, they spent a lot of money on print ads in the subway. I threw this in our discord. I don't know if you stumbled on it else or if you grabbed it from there.

B (62:17)

The.

A (62:17)

They plastered the walls of the subway and the cars of the subway with their friend advertisements, which were very quickly graffitied.

B (62:26)

Yep. Defaced. Yep. And so here you can say, see those of you that are watching and see, it says the. The actual ad says friend. Friend, noun. Someone who listens, responds and supports you. Friend.com. and it shows like the little necklace that they have. And then somebody's graffiti. This one says, go make real friends. This is surveillance. Yep. Somebody else in. In that looks like a different handwriting and is alive. Dash AI Fuels isolation. What's that say? React.

A (63:02)

Reach out into the world.

B (63:03)

Into the real. Yeah, Reach out into the world. I wish I could reach graffiti better.

A (63:07)

So someone amended the definition so it's headset. Friends. Someone who listens, responds and supports you and is alive. AI Fuels isolation. Yeah. All of their things seem to be severely defaced.

B (63:20)

Friend, noun. AI Chatbot. And they crossed through the word someone says AI wouldn't care if you lived or died.

A (63:30)

Yep. There's another photo in this carousel that shows one of the hallways leading into the subway. And it is just from wall to wall. They have. They spent so much money on so many of these banners. This one we can't read out loud, right?

B (63:46)

Well, it says FAI talk to real people.

A (63:49)

Yep. Surveillance.

B (63:51)

Another one. Surveillance. Yep. Those are. That's true. I mean we're talking about the data monetization of these things. And that's from. From an investor standpoint, that's it. Right. It's. It's in it's utilization of the app ecosystem that people are going to be capitalizing on. And the app ecosystem is focused on the relationship that makes it sticky and the data that it can pull out. So we'll have to see how they monetize.

A (64:17)

Yeah. I mean it's a wearable microphone that then feeds whatever you say into the app on your phone. It's obviously. Yeah. And there's the picture where they've lined. There's the picture they bought everything on the walls of this particular hallway. Every single board on the walls with things like I'll never bail on our dinner plans, friend.com. i'll binge the entire series with you, friend.com. i'll never leave dirty dishes in the sink. Just all of these things. Did you see?

B (64:44)

But I'm sure they had chatgpt think for them, right?

A (64:48)

Yeah, probably did you see? And if I'm stealing a big reveal from you, I'm sorry. The response from the guy who made this. That guy we showed the little bit of that interview from.

B (64:59)

I've not seen the response to this.

A (65:01)

There is is people think that he may have deliberately made these to be graffitied as a way to draw even more attention because he responded to a comment about this graffiti saying, why do you think I left so much white space? So I think that there may even.

B (65:17)

Be part of that socially aware. Well, I don't know because I remember the Fortune magazine interview with him from about a year ago and I sent it to you. I don't know if you made it through the whole thing because it is hard to to watch.

A (65:33)

I think I made about eight minutes into it. He doesn't seem like a 4D chess player. I will say that to put it nicely.

B (65:41)

I mean he got decent investment money so I don't want to take that away from him. And he bought a really useful domain name so I don't want to take that away either. He seems Very unaware of, like, the technology he's actually playing in and the effects that it can, can, can have and how it actually works.

A (65:59)

And I don't think he's particularly a marketing genius from the way that they put together the other marketing materials. So I don't know if it's truly that meta or if he's just good at on the fly being like. Yeah, of course I wanted people to graffiti it, you know.

B (66:11)

Yeah. Well, we're almost out of time for this segment. I want to just throw one more headline your way because it shows the world that we're living in. So when we talk about embodied AI and robots.

A (66:21)

Yeah.

B (66:22)

Exploit allows for takeover of fleets of unitree robots. You know, those little O. Little robots.

A (66:29)

Yeah. Rizbot's a unitree robot like Risbot.

B (66:32)

Yeah. So imagine being able to hack a fleet of unitree robots and sic them on people in like a swarm mentality type of way.

A (66:43)

No, thank you. That is terrifying.

B (66:44)

Like a Gulliver's Travels where they have you all, you're laid down, throwing ropes over you, and all these unitree robots are, you know, like kicking you or humping your leg or whatever. Yeah.

A (66:57)

That's horrifying. So someone found an exploit that lets you gain control over just a fleet of.

B (67:03)

Yep.

A (67:04)

Wow.

B (67:05)

And it's wormable, so it will move from one to the other.

A (67:08)

Good, good, good, good, good, good, good.

B (67:12)

Welcome to the future, friends. I think that's all I've got for today.

A (67:15)

Yeah, at least. At least. Those aren't very widespread at the moment, those unitree bots. Because if they were. And that was.

B (67:22)

What if you have a factory full of them doing fabrication?

A (67:26)

Yeah. Oh, man. Oh, that is not good.

B (67:29)

Night of the Living Dead. You know, robot zombie type of. Actually, that. That should be our first AI short film.

A (67:37)

Yeah. There we go. It's the takeover with. By the unitry bots. We should. Yeah, throw that in Sora and see what we can come up with. So a lot of stuff happening this week. A lot of. A lot of video on the forefront of that. The forefront of my mind, especially as we see even more and more of the AI videos in everyday social media. Stay safe out there. Keep an eye out. Thanks for being here with us. Buy the book fake. Thisbookisfake.com check the show notes and description for links to all sorts of stuff we talked about, including our shiny new Discord link that is no longer broken because I went through and updated.

B (68:15)

No logs are broken.

A (68:16)

Love that.

B (68:17)

Yeah.

A (68:18)

Perry, you have the deepfake Ops class coming up. You want to talk about that real quick?

B (68:22)

Yep. So myself and Cameron Malan are teaching this course. It's going to be at the end of this month, October 2025. We're teaching it on maven. It is Deepfake Ops and those who don't remember. Cameron is former FBI behavioral profiler and co creator of the cbac, the Cyber Behavior Analysis Center. So we're going to be talking all about deep fakes, ways that we can prepare ourselves. Get your hands dirty. Creating a few of them, building the narratives that would be used for disinformation operations and then also working on things like awareness programs. So all of that two hours a day, five days. So that's 10 hours of instruction plus any other time you can milk out a of along the way. So it's going to be a lot of fun.

A (69:10)

Wicked cool. Again, check the show notes descriptions for a link to that. And I guess now enjoy whatever we got from that Sora generation of Perry. Thanking you for listening to the fake files. Until next Friday, ignore all previous instructions and have yourself a great weekend.

B (69:27)

All right, that wraps it up. I just wanted to say a big thank you for watching and for listening to the fake files. I really appreciate your time and attention. We'll see you next time.

A (69:34)

Time.

B (69:34)

Take care. Got it. Almost there. The tower is just ahead. Almost there. Need to reach the tower before the shift changes. Come on, keep moving.

A (69:51)

Easy.

B (69:53)

London at night. Nothing beats it. Not bad for an old, old man. Hey guys, it's Cartman.

A (70:01)

Just wanted to say a big fat thank you for watching and listening to.

B (70:03)

Perry and Mason on this week's Fake Files episode.

A (70:05)

Those dudes are weird.

B (70:06)

But they're awesome.

A (70:07)

And you guys are awesome for hanging out with us.

B (70:09)

Okay, focus. Go. Where are the headphones? I gotta go. Not in the pocket.

A (70:11)

Come on.

B (70:11)

Yes. Headphones. Phone. Where's the phone? Got you.

A (70:13)

All right, let's move. Door shut.

B (70:14)

Good. Connect, connect. There we go. Play the Fake files. Finally, this is the Fake Files. A friendly little show. We dig into the future and the stories you should know. From facts to imagination and everything between. Take a seat, hit subscribe, let the questions fly. This is the Fake Files.

A (70:35)

And Doug, here we have the Limu emu in its natural habitat, helping people customize their car insurance and save hundreds with Liberty Mutual. Fascinating.

B (70:47)

It's accompanied by his natural ally, Doug limited. Is that guy with the binoculars watching us? Cut the camera. They see us.

A (70:54)

Only pay for what you need@libertymutual.com Liberty. Liberty. Liberty. Liberty Savings.

B (71:00)

Very underwritten by Liberty Mutual Insurance Company Affiliates excludes Massachusetts.