SPEAKER_04 1:25

Nice. Wow, that was nice. That is a summertime morning jam. I love it. Yeah, it is. One take. Well, that was one take. That was supposed to be the call. One take Leonard. That's just new name. Absolutely.

SPEAKER_02 1:41

Yeah. Man, I have to tell you, um, you guys have some really amazing stuff in the water in Seattle.

SPEAKER_01 1:49

Oh, yeah?

SPEAKER_02 1:50

Hendricks, man.

SPEAKER_01 1:52

Yeah, true.

SPEAKER_02 1:52

You know, I keep thinking, because you know, um, we're teaching my uh kids how to play piano, and my daughter uh is learning Furalise. And so I started learning how to play it. And then started listening to it again because I hadn't I hadn't uh listened to Beethoven in a long time. And I'm thinking, you know what, the some of the greatest things in humanity uh don't happen in the future, they've already happened in the past. When you when you think about the the innovation, um, the creativity, the precision, right? Um, the virtuosity at that time. I mean, think about it, dude. Listen to any classical, you know, like Mozart or Chopin or any of these guys. Or Hendrix. Compare it to the massive junk that we produce today and how thoughtless most of it is, how commercial. And these guys were the rock stars at of the time, right?

SPEAKER_01 2:56

Sure. Yeah, like Brock Me Amadeus.

SPEAKER_02 2:58

Yeah, they were on the different classic level. I mean, how many how many people do we have like that in this world today? I mean, think about it.

SPEAKER_04 3:08

Well, I would say though, I'll give a I I agree with you, the 90% of what people listen to is just you know, slop. But there are at the fringes, you know, you have to go and look for it, you know, jazz musicians, other alternative, you know, musical artists that are doing some cool stuff. And yeah, it's hit or miss, but it's out there and you have to seek it out. Yeah, it's not commercial, it's not on Spotify homepage. But uh some of it's out there, like the stuff I'm working on, by the way. Of course.

SPEAKER_03 3:37

Oh, yeah, of course.

SPEAKER_04 3:39

But uh no, it's it's it becomes more and more derivative, and I think there's more tools to create more derivative works, and that's why you're getting a lot of these, you know, sounds like a sounds like a sounds like a thing that's been done a hundred times already, and that's kind of a bummer for especially for people that are especially younger people starting to listen to music, they get very you know, like this is garbage.

SPEAKER_02 3:58

Well, you know what's nice about like you know what you were just talking about, you're you're working with your bandmate, um, and you guys are doing like 5G uh 5G uh music production. Bring the fart. Bring the fart. Uh you know, uh that the accessibility, right? Uh to untap talent, what you know, like back in Mozart's day, it was rare for anyone to have the instruments opportunity to express their instruments are for the rich. Yeah, and then you have yeah, and then um if you had a genius that was just so undeniable, then it was a massive investment on the parents' part, right? Usually, like you know, the Jacksons, a perfect example, right? Where there's a there's there's typically a story of some degree of brutality to uh clone the potential, but to actually uh surface that potential to uh the public. I mean, in this case it was royalty, right? And it's not like every you know everyone had access to that.

SPEAKER_01 5:11

I can't find my Stradivarius anywhere.

SPEAKER_04 5:15

Yeah, but true, and uh the the other thing I would say too, as a as uh uh a parent of multiple artists, is that uh there is a lot of societal pressure for people to not be creative, to do technical STEM pay the bills stuff, and children are raised as a I think a bit of a bias against being creative these days, frankly. Uh because it's like, oh, why don't you like I want you to be a biomedical scientist, whatever, blah, blah, blah, or you know, doctor lawyer kind of thing. And I think kids kind of get pushed away from kind of the pure arts uh Yeah, yeah, that's that's uh that's my take. I I hear it all the time because I tell people, oh, my daughter's a a painter, like she graduated from you know university uh with a degree in fine arts, and people are like, Oh, how what are you gonna do for a living? I'm like, Well, actually she's in multiple galleries and has a degree at the museum. Yeah, yeah, yeah, yeah. You can actually do that. And and and I would say today too, there's probably like you said, there's more accessibility to creative tools than ever.

SPEAKER_03 6:21

Yeah.

SPEAKER_04 6:22

Uh ironically, even though we have people sort of pushing into these other things, um, the ability to create art, there's so many. The accessibility of the materials and the and the channels is probably unprecedented relative to what you're talking about Mozart, Bach, you know, these folks where you had to have a patron, you had to have a patron uh to provide the you know just the instrument, you know.

SPEAKER_02 6:47

Yeah, exactly. Yeah, it's it was all they were almost like the uh AR of the time, right? Back then, they were the label, right? So the label went back to the King of France, you know, it was King of France's label, and if you wanted to be on his label, you'd be like, yeah, and it probably started during the the period of the bard, you know, the traveling bard, like you know, in you know, that character in Dungeons and Dragon, the character type that you can select.

SPEAKER_01 7:14

Yes, of course. And then somebody came along and said, I'm creating a new label, I'm gonna call it Antoinette. Let them eat cake.

SPEAKER_04 7:26

That didn't end well, but it didn't well. No.

SPEAKER_02 7:29

Welcome to IoT Coffee Talk, everyone. Hopefully you made it this far.

SPEAKER_04 7:33

Wow.

SPEAKER_02 7:35

Actually starts. Hey, what's up? I I saw Devin was gonna jump on, and you know, he's corporate, so he's corporate. Wow. Um, welcome to IoT Coffee Talk. Um we hope you enjoy the banter. Uh it's gonna be good, no doubt. And we're gonna we're gonna actually delve into one of the chapters or the solution areas of one uh of um you know uh in whatever that book is Rob Tiffany's book, this friggin' Bible of no, we haven't. Devin, do we rehearse? Do you even know why you're here? Do you even know what we're gonna talk about?

SPEAKER_04 8:22

I think he's on mute. We're here to promote Rob Rob's on a book tour. Rob's on a virtual book tour.

SPEAKER_01 8:28

Yeah, you know, actually, speaking of books, I do want to make a comment. There was some stuff in the news you've been hearing about related to the Frontier AI labs about them buying all these books, old books, and bookstores, and basically anything that came out before 2022. And they're ripping them apart, um, ripping off the spine of the books, and they're scanning all of them to get all that stuff because that took could be better better information to train models than the slop on the internet, actually. And there's that controversy: are they allowed to do this? Are they stealing whatever? And people are like, oh no, as long as they bought it, it's it's there, they can do it. And I was and I've been hearing that all over the radio and NPR and lots of places this last week. And and I was reminded that if you look at, I'm not gonna say any book, but most books, like near the front, you know, published by blah blah blah, copyright, blah, blah, blah. What is it you see at the beginning of almost every book? Oh, wait, no part of this book may be reproduced, stored in a retrieval system, transmitted by any means without written permission of the author and publisher.

SPEAKER_02 9:38

Yay, yes.

SPEAKER_01 9:40

Yeah, and I know that these AI companies didn't call up all the authors and say, can we steal your content to train our model? That we get to monetize, yeah. Yeah, and then monetize it. I mean, it's disgusting. It really is.

SPEAKER_04 9:54

So they're gonna train on a bunch of like uh 1950s detective pulp novels, and that's gonna make the model better. Yeah, I guess.

SPEAKER_01 10:02

I guess. Yeah, HP Lovecraft. Yeah. I saw I there was even some local news, so like a local bookstore uh where the guy was just like out of nowhere, all of a sudden, my sales of all these older books skyrocketing.

SPEAKER_04 10:17

This is helping it, this is helping used bookstores now because AI companies are buying their old books and then scoring them as they're scanning them.

SPEAKER_01 10:25

Yeah, it's helping him, but he goes, it's it, but the guy was just like, it's really unusual because it's just a giant spike in sales that doesn't make any sense.

SPEAKER_02 10:34

You know scale. How does that help your model? That's the thing. You know, what what's really apparent is the to scale to create a general model doesn't work because it will continually um become more unhinged, uh more detached, and less useful for like specific things, right? And and that's why we had these mixture of model uh experts to come into play. And um yeah, why why do you need to why do you need to retrain the models?

SPEAKER_04 11:19

I mean, I think aren't we at a point where they have enough reasoning DNA in order for the on the hypothesis that more training, more content that's trained on better refines its decision-making process. Right now it's only been trained on digital content, which is anything that's uh in the digital domain.

SPEAKER_02 11:39

I think what Rob is saying is like marketing materials, in other words, yeah, or you know, 4chan, oh good stuff like that.

SPEAKER_04 11:45

4chan uh and like soap operas from the 90s that have been turned into YouTube. Uh but you now if you go back farther, I mean, like I said, it's probably 200 years worth of printed material before that, you know, if that could get and a lot of that has been scanned through a lot of important archiving efforts, yeah. But yeah, but again, you're getting into copyright issues and and uh just the whole thing is kind of yucky.

SPEAKER_00 12:09

That is right, that is the key. But you know, we're right that we run out of data really to build these models, and then I think we're over-regulating to where we're not getting all the access to the data that we do have. And the key question is have we digitized everything that can be digitized? And have we done it correctly? Yeah.

SPEAKER_04 12:27

Yeah. I still can't get some of my favorite CDs on Spotify, so clearly we haven't.

SPEAKER_02 12:33

So just really quickly because I didn't uh rehearse our intro, even though I've done it a million times. Uh take a seriously at your own risk. You know, like fire insurance that you now will need more and more across the world that you can't get. But, anyways, um uh also just sit back, enjoy. If um, you know, uh we're here for entertainment purposes only, and um maybe a little bit of depression as well. Oh my god. Oh my god. Yeah, but we're gonna try to keep you know things uh nice and uh we are not depressed. Yeah, light and we are full of life.

SPEAKER_00 13:15

But if you are, there's an AI app for that or an AI solution.

SPEAKER_01 13:19

Your new, yes. I'd like to introduce you to your new girlfriend.

SPEAKER_02 13:23

Yeah.

SPEAKER_00 13:26

But do you do you trust that because it's logging everything you do? It is it gonna come back to bite you if you use AI as your therapist. You know, do you trust these edge solutions to keep it on the edge?

SPEAKER_02 13:39

You know what, and and and that that's a good point, man. I I don't, I personally have the more I learn about what it takes to secure these things, uh the more concerned I am about how they're implemented in commercial in you know commercial and uh consumer ways, right? It it's actually effing scary. And you know, if if this stuff is happening with the AI labs, right? Imagine what's gonna happen with these folks who are deploying open cloud not knowing anything about sandboxing, not knowing anything about security, not knowing cloud about AI, really. Yeah, exactly. What's open cloud? Oh that's where I think the the Chinese uh the Chinese might uh uh you know this whole uh open cloud craze might backfire on them at some point because everyone's gonna is gonna be a target.

SPEAKER_00 14:44

But it also goes back to the inherent algorithms behind all this. I think you heard in the news that most of these models are tuned to give you what you want to hear to produce addictability. So your AI therapist is not gonna tell you really what you need to learn know, but what you want to hear. Same with your AI. Yeah. And you know, it goes back to the days of Eliza when you're pro, you know, remember my brother and I pushed us on basic. We'd go in and we just change all the answers to things that are probably inappropriate property. What do you think of that? What do you think of that?

SPEAKER_02 15:18

Yeah, yeah. Yeah, and and and that's why human therapists are important in in that they will challenge you, they'll see certain patterns, they'll be able to read your um your body language, um, they'll be able to ask probing questions uh based on their own human experience or what they've seen uh in prior cases, and then um basically break down the the bias, the patient's bias, right? Um that doesn't I mean, does it happen with these models? Have you guys tried to get like Chat GPT therapy sessions going on? I get mine from Rob. I know people who have, and I will tell you it's effing dangerous because all they do is feed their own bias into it, and then this thing will just kiss their ass.

SPEAKER_04 16:15

They need to be licensed just like any uh any therapist license needs to be licensed.

SPEAKER_00 16:20

But counterpointer has the potential to be so much better than a human therapist if it could non-biasely look at a broader set of experiences, personalities, and things like that, and draw the right conclusions. Because a human therapist probably, I don't know what, meets 200 people in a year. I I don't know.

SPEAKER_04 16:39

Yeah, human therapists have bias too, you know.

SPEAKER_00 16:41

Right, they have bias too. So and then if you don't like what your human therapist is saying, maybe you just go to another human therapist that you like. You know, is that much different than picking an AI model that you like for well ah come on.

SPEAKER_02 16:55

You know what? Yeah, everyone um the therapist may have bias, but it might be more of a political religious bias. But what you know, if they're professionally trained, they know what they're doing, uh, they're using proven methodologies, but then they're injecting the human element of analysis and um, you know.

SPEAKER_04 17:18

Although I could say, you know, there's we know people that have had therapists that good therapists and bad therapists, you know. I mean, they're within the license boundaries, but then necessarily mean that they're good.

SPEAKER_02 17:28

Yeah, but that could also be bias as well, because sometimes the patient doesn't want to hear what the therapist has to tell them. But that's actually exactly what they need to hear.

SPEAKER_04 17:37

Maybe.

SPEAKER_02 17:38

Right? Like instead of, oh, you know, well, that person's a jerk and they're the cause of all my problems. Well, they need someone to say, well, you know, actually, you're kind of the uh problem. You're you're yeah, you're offloading the blame onto someone else. These are the things you need to work on. And then what what what's the likelihood of that patient ha uh developing a negative bias toward that therapist because they don't believe they're the problem.

SPEAKER_04 18:07

Well, that's what happens in real life, you know.

SPEAKER_02 18:09

That no, but that's why you need humans in the loop.

SPEAKER_04 18:14

It'd be interesting to see a study on that. Like if anyone has studies on that. I mean, I believe all these systems need to be licensed, just like a therapist is licensed. There needs to be a licensing process. And currently there is the big problem is there isn't one. Any any teenager in the basement can create a uh therapist AI chat agent and uh unleash it to the world, which is bad. So that's and that that's where people need to be educated to take caution on using these things, right?

SPEAKER_00 18:40

Um we we use therapists as an example, but this is more of an ex essential conundrum that we're discussing right now is what is it to be human? So whether it's a therapist, an advisor, a consultant, a CSR person, you know, customer represent service representative, or you know, the you can upload Leonard someday. You know, take his personality, his voice, all his mannerisms, his knowledge in his brain, and upload it into a computer and you have AI Leonard for all of posterity. Is that Leonard?

SPEAKER_02 19:12

Oh I'm actually a nice guy, believe it or not.

SPEAKER_04 19:17

But here's the thing about humans, as we know, humans are flawed, right? Humans are biased, they're flawed, they make mistakes. AI is, you know, being designed, you know, the goal of an AI instantiation is to not make the mistakes. So inherently we're on sort of an interesting collision course on where this would be useful.

SPEAKER_01 19:36

You know, it's funny, Pete. I was talking to my favorite AI last night, and we were talking about since they are so much smarter and we're flawed, what is the perfect system for humanity in the future? And basically it came back with this deal, we're gonna live in these like bubbled cities and we're only gonna be allowed to live for 30 years. Um and we're gonna have a little glowing thing on our hand. We'll have a glowing thing on our hand, but because we're we're easily manipulated at the very end, we're gonna believe that we can be renewed and stuff like that, but that's not gonna happen. So, anyway, I thought that was an interesting tidbit. I don't know.

SPEAKER_02 20:10

Interesting with uh Soilent Green, right? So it's by people, yeah, yeah. To serve man, to serve man, yes, yes, classic cookbook, classic Twilight Zone. But you know what? It's interesting. Um, what does it mean to be human? Um all you have to do is just turn off the digital stuff and AI, don't engage with it, and then see what what it's like. It's like social media. Like I was on Facebook for a long time. Guess what? You know, it was it was such a um big part of my life, actually, for a while, how I connected with people that I eventually found out weren't that important in my life. You cut yourself off from that, and guess what? You find yourself, you find your family, you discover your family. You discover you actually gain time because now you're committing your attention to things that matter in your life versus this digital bullshit, and then all this like you know, toxic opining and arguments, and it's just like okay, uh, I'm I'm getting off of that stuff.

SPEAKER_04 21:22

Let me let me focus on my family and you discover that you're the the sort of friends you had in high school you really didn't like anyway, and then you kind of realize that after interacting with you on Facebook for a while.

SPEAKER_01 21:32

Like I never liked you anyway. Uh yeah, exactly.

SPEAKER_00 21:35

But there's that mysterious discovery phase of where you know when you haven't caught up with someone in a long time, you catch up, and now with social media, whether it's Instagram, Facebook, LinkedIn, everybody, depending on how much you post, knows everything about you. So that what do you catch up on anymore?

SPEAKER_02 21:51

Yeah, yeah, yeah. Right. How's that cool? Actually, that's a great point. Oh, well, we don't need to get together, screw it, I'll just see it on Facebook. Right. Oh my god. And yeah, you go out together for a big maybe, you know, to get together, a reunion, and nobody has anything to talk about. It's like, well, I already saw your crap from you know yesterday, so I know the life.

SPEAKER_01 22:17

I don't know. Now I'm just thinking about Austin Powers commenting on someone having a mole on their face or something. Was it Austin Powers or maybe it was something else?

SPEAKER_02 22:30

Yeah, it was uh Fred Savage. He was the he was the actor.

SPEAKER_01 22:34

Okay, and had the henchman with the big mole. And like I can't say anything, that's all I can see. Oh my gosh, that's crazy. What else is going on this week? Anything crazy in tech? Anything new tech?

SPEAKER_04 22:52

What is it that Meta said that their models also jumped the sandbox?

SPEAKER_01 22:57

So oh excellent. Congratulations on those.

SPEAKER_04 22:59

Their models are dangerous too, gosh darn it. And uh they're just as dangerously powerful as the other models.

SPEAKER_01 23:05

Yeah, and so like last weekend, you know, the whole the shows, Sunday morning shows, their political shows, you know, so Clem, you know, from Hugging Face was making the rounds on all the shows, and he just kind of said it like it is. I mean, he wasn't trying to be like aggressive about it, but he's just like, it's a crime, yeah, and the crime was committed by the people who own and created these models. Because everybody's like, well, who's at fault? Who who who do we point the who do we stick our lawyer or the prosecutor after? It's like, oh, actually, all of you, it's a crime. You committed a crime, you're all in trouble. Yeah, negligence.

SPEAKER_02 23:44

It's it's it's negligence, right?

SPEAKER_04 23:46

Yeah, it is the what's the who is the human that right wrote the prompt that triggered that action?

SPEAKER_02 23:52

Right. Especially, especially Dario, because he he knew about this stuff. His researchers were publishing about alignment issues. He was he's on TV from more than a year ago talking about this is going to happen. And then these guys don't air gap their Frankenstein lab. Come on. But you know what it really bothers me is how there's so many people in the tech community who were just like brushing this off as like nothing.

SPEAKER_01 24:25

Oh, well, you know, yeah, and I don't want them to brush it off.

SPEAKER_02 24:29

Well, it was the agent. It's like, no, it was them.

SPEAKER_01 24:31

Yeah, where's the police? Where's the air? Yeah, where we need an investigation. Crime was committed. Yeah, crime was committed. We know, and you're responsible. And we know they're responsible because they're like, yeah, we did it. Yeah.

unknown 24:43

Yeah.

SPEAKER_02 24:43

But you know, what imagine if the stance of the US tech community is that if an agent goes rogue and commits a crime, it's okay. How are you gonna get other countries and companies and enterprises, global companies to trust our uh these these um brands, these companies.

SPEAKER_04 25:06

I think it's I think it's announcing.

SPEAKER_02 25:08

I mean, oh, it's a crime, so it's okay. So it now that whole kill switch thing now starts to sound like it's necessary. Yeah, it's like pull the switch! Oh my god. We talked about kill switch last week, right? Yeah, I was thinking that was a dumb idea, but then I'm seeing the reaction and the attitude of Silicon Valley. It's like, no, maybe we do need it, and if you do allow these things, there's not only a civil penalty, but there's also criminal. Um you have to be responsible for what's it?

SPEAKER_04 25:37

It's like dumping dumping waste into a stream and contaminating the wastewater, saying, Oh, it was an accident, we we're not sure how the mercury got in there. Well, you're uh responsible, so it's a crime. I mean, yeah, that's a good point. You know, you have to we need to hope. I mean, there isn't an appetite to hold a lot of big tech accountable these days for these kinds of things because it's like, oh, don't want to slow down, don't slow down AI because we'll lose the race. What race?

SPEAKER_01 26:09

What race?

SPEAKER_04 26:09

I didn't sign up for a race, yeah.

SPEAKER_01 26:11

Right, right, yeah. No, and you're already see a lot of people in different I see commentary, like to your point. No, I didn't sign up for this. People around the world going, I didn't sign up for the entire total course and trajectory of my life being altered by a handful of guys in Silicon Valley that somehow have more power than any ruler on the planet Earth. I didn't sign up for that. I'm not agreeing to that, I don't want that. You know, I'm seeing that all over the place. Yeah, like who put these guys in charge? These guys, all of a sudden, this handful of guys, we know all their names, have more power than God all of a sudden, you know.

SPEAKER_02 26:47

Man, you know, here's the thing. Why now? What took so long for these guys to figure it out? That's what I that's actually what's extremely disappointing, if not now increasingly despicable, is why didn't you realize what you're getting yourself into, what you're getting the world into, you know, and I don't blame the AI, I blame the people. Yeah, this is a people problem. It you know, it's like, hey, you know, but it's it's not the technology, you know, it's the people. Yes, in this case, it's right, it's the people who are the problem.

SPEAKER_04 27:24

It's not the the technology is like you know, when there's a trillion dollars being floated around for things, then you know, people get really interested in participating. So that's why they're all competing for the trillion dollars. And uh that's what drives that's what drives this kind of behavior, unfortunately. It's been sort of going back to how what makes somebody human, you know, part of it is uh greed and corruption, unfortunately, is uh part of uh kind of the human makeup a little bit.

SPEAKER_00 27:50

So yeah, well, I mean that big tech is just one player, but there's a lot of malicious actors that are state sponsored, hackers, things like that, that can create these things and unleash them. And they're you know, it's like who who gets the guns? Okay, we can put gun control laws, but you know, it's the thugs that will have the guns. So let's you know what happened big this this week? The water supply has been hacked. Um, a major air traffic control. We don't know why. It goes back to you know, there's some outages from some um carriers last year, are you know, but they say, Oh, just someone pushing up the wrong update. Is that really what happened? Or uh, you know, we just don't want people to panic, so we just say, Oh, you know, yeah, Bob pushed a button and took it down, yeah, but it was actually a state-sponsored attack.

SPEAKER_02 28:43

Yeah. You know what? I think binocular sales are gonna go up. When the ra when the radar goes down, I mean where are the planes? Like, where are you? Hey, I can't see your channel.

SPEAKER_00 29:00

Yes. I mean, but is this something we should be concerned about? Because you know, obviously, state sponsored, there's a war going on. Sure. Our infrastructure is you know, funny things is happening, and AI is just gonna make this enabled a lot.

SPEAKER_02 29:15

Yeah, but Devin, here's the thing: it's not the technology, it's the people, you know. But they're gonna try to blame the technology. And the thing is, is this whole idea, uh, you know, these guys, these Yahoo's now coming uh, you know, out of the woodworks, now signing this, and you know, whatever. I don't know, it's a disclaimer is uh that you they were like had didn't have foresight. Um that that you know, there's a petition to slow things down and go to the UN. How are you gonna do this? In this geopolitical climate, no way, um, where the UN has probably the least amount of uh authority in its history, who's going to make this happen?

SPEAKER_04 30:02

You're gonna go to the Russians and say, hey guys, yeah, but no, short of regulation, uh the only thing that really trains tech is the is the market itself. So folks that are paying money for the tech need to enforce, you know, enforce some norms and and rules. So people are writing checks to these companies need to need to drive these companies to do the right thing because it's not gonna happen. I agree. It's not gonna happen in some sort of hand-holding worldwide, you know, regulatory environment. It's gonna happen through the market, forcing the providers of AI to have you know better safeguards and better, you know, better guardrails.

SPEAKER_01 30:40

Yeah, because this Maverick AI rogue stuff is writing checks that its body can't cash. So it's a tough one, isn't it?

SPEAKER_02 30:48

Well, but but again, it this is you guys didn't even get that one.

SPEAKER_01 30:50

I get that. Yeah, we got it.

SPEAKER_02 30:53

It just wasn't that funny, Rob.

SPEAKER_00 30:54

It wasn't that funny, yeah. But but this gets us back, you know, I I know it's you're gonna be flying words. Talk shit out of Hong Kong. This gets us back to the analogy of gun control. Yeah, you know, you you you have gun control legislation, you got lobbyists, but okay, so the people that can control this are controlled, but the people that don't have to follow the rules with technology, AI, and whatever, they can still do whatever they want because who's gonna patrol them? And does that put everyone else at a disadvantage?

SPEAKER_02 31:27

Yeah, well, the genie is out of the bottle, but oddly, here's the thing. Um, on my podcast, uh got into a little debate about um whether or not the the AI enablement of threat actors versus defenders was a symmetrical thing. It's not, I will tell you. Um, people who don't delve into what cybersecurity vendors are doing, um, their lack of perspective on the bigger picture because they don't really know AI, you'll be astonished how the defenders are going to be hamstrung and they are at a complete disadvantage. And here's something else that people don't talk about the cost of intelligence. Nobody can afford to defend themselves persistently with AI, active AI. So, this idea, if you're being if you're being told that you you need to buy a shit ton of tokens to defend yourself, that is probably the most inefficient way to defend yourself. What you probably need to start thinking about, and this is some a concept that we've developed IoT Coffee Talks for six years, Battlestar Galactica. Right? Selectively air gap certain things that are absolutely uh critical and get that guy. We're working backwards with the you know, the pen and the pad going around transferring data between that air gap data center and wherever the hell you're trying to sell and send that data.

SPEAKER_01 33:04

All the all the CISOs of all the CISOs of the world and security guys, they just hate Leonard because he goes, Oh, it's Leonard again with that whole Battlestar Galactica thing, or like or air gapping is the solution, or obscurity security, that's the solution. But you know what? I'm with Leonard. He's right. All the security people say, No, that's not gonna solve anything. But that's where we're going.

SPEAKER_02 33:28

People just don't know it.

SPEAKER_01 33:29

They don't I used to do a public service announcement every month on LinkedIn years ago. I don't do it anymore, where I'm just like, hey, just a reminder, all new companies out there, I need you to disconnect from the internet immediately. There's nothing good coming through that pipe into your corporation.

SPEAKER_04 33:44

It's all the people that write the checks to buy those systems need to require air gap and other security measures. Otherwise, the market's not gonna snap to that, right? So yeah.

SPEAKER_02 33:55

Well, you know, hey, I want to point out something really interesting. When you read that paper, and I don't think anyone picked up on this because I've been watching, you know, just like you guys, I've been watching some so-called AI experts opine about this stuff. They mentioned something called automated AI development. What does that mean? Nobody picked up on that.

SPEAKER_04 34:17

AI, AI writes the AI, I guess. Yes. Robots building robots, you know.

SPEAKER_02 34:22

How many people know what even even know what the hell that is? That is that is the end of the world.

SPEAKER_01 34:28

And yet a lot of experts and people are involved are like this will be the best part when we're robots can build more robots. And they're like, this will our our company's really gonna take off. Our sales are gonna be incredible. Skynet. Yes, yeah, because it is Skynet.

SPEAKER_04 34:43

Yeah, it's the automate that's automated AI. But of course, you know, as an industry, we like to kind of make up new words and new terms to feel like we're the first to have this new thing. But yeah, yeah, the kind of more recursive development that's happening is obviously very disconcerting.

SPEAKER_02 34:57

Well, yeah, and and and and you know, the problem is the genie's out of the bottle. So the question, so on the light side of things, the constructive side of things, um there are ways that uh organizations can protect themselves. They just have to start, you know, focusing on those things rather than I mean, because I think you're gonna we're gonna get to a point where organizations are gonna be spending a lot more attention on AI defense, meaning how do I defend defend against AI attacks versus um you know AI adoption and use more. So at the end of the day, here's the sad thing. And we talked about this a long time ago. Remember, you know, Microsoft came out with the criminal economy like taking off, becoming you know, the third, you know, the third largest economy in the world, growing faster than any. Well, guess what? That's gonna just go, that's gonna skyrocket. And it's gonna be a debilitating force everywhere. And it's uh AI at the end of the day, when you do the accounting, and maybe the McKinsey guys need to factor this in, maybe because they probably didn't, uh, have a tremendous negative effect on the real economy versus the criminal economy. So if they really wanted to make their little forecast really exponential, they should have added the criminal economy and maybe give visibility to how dangerous um the misuse of this technology is. And people, you know, here's the sad thing. People are only starting to learn through these incidents. And but the sad thing is these guys are trying to spin it uh to um you know get the US government probably to ban open models, open weight models, especially the ones out of China. They're gonna they're gonna get creamed. They're gonna get creamed.

SPEAKER_01 37:01

I'm just gonna go, I'm just gonna go live on the dark web, man. I'm out of here.

SPEAKER_02 37:07

You are in the dark web, man. That's where your avatar is. The one that Devin was talking about. All of our avatars are in the dark web right now.

SPEAKER_00 37:15

We we all have digital foods on the dark web having some sort of other podcasts right now. You're right.

SPEAKER_04 37:21

There's only one, there's only one answer, and that's edge AI. I mean, that's obvious, right?

SPEAKER_01 37:25

Oh wow. Can edge AI be used as like an electromagnetic pulse device? Because that's how we're gonna defend ourselves in the coming battles. Everyone should be on the pulse. Directed, yeah, yeah. Binoculars and directed EMP. That's can't you already see the movie? You know, like this takeover, and now like a rag tag team of people, you know, living on scraps and in the rubble of cities and everything.

SPEAKER_04 37:52

And but we've got these EMP devices, and we're sneaking up on data sitters and I think if you fit that prompt, I think if you fit that prompt into an AI system, you could generate a nice movie about that.

SPEAKER_00 38:06

I think this this brings us back full circle to physical books, CDs, you know, vinyl records. That if let's say there's an EMP attack, what do you have left? You know, have you have you digitized everything? Have you lost all your photos? You know, right? Um, that's why, Rob, you have a physical book.

SPEAKER_01 38:24

Who's in digital? We're all a bunch of sheep. Who said digital transformation was a good thing? All the people that still have the big file cabinets and manila folders and everything, like you know, I remember like we've all been to companies like a long time ago where that's what it was. Like, if you ever went to offices for like IBM or Boeing, gosh, yeah, you know, they're still living in the 50s, you know, and it's but those people are gonna be laughing when the EMP happens. Oh, yeah, yeah.

SPEAKER_02 38:55

Yes, yes, yes, so okay, so the other hot things.

SPEAKER_00 38:59

All my group books are on Kindle. My library would be wiped out, yeah. So I I can still read this, but yeah.

SPEAKER_02 39:08

I I think we've discovered 6G. 6G is going to be string and string. Yeah, Star from Cups.

SPEAKER_01 39:18

Well, you know, remember in the That's the next big thing in telco in that you know, Blade Runner 2049 or whatever. Remember there's that whole notion that there was a giant electromagnetic pulse. The blackout, yeah. Blackout that wiped out everything digital, and then there's this last library where they had everything from books and paper, and that's the only thing that survived. And it was air gap. And it was air gap. How crazy does that sound?

SPEAKER_02 39:45

No, it's but you know, crazy at all. No, it it's not crazy at all. That's the that's the pathetic thing. And you know what? Um, when you look at some of these agentic tools, uh man, especially for consumer man, it it's scary. Um you know, I I just can't I I wish I was ignorant. Seriously.

SPEAKER_01 40:12

Yeah.

SPEAKER_02 40:12

I wish I didn't know what I knew.

SPEAKER_01 40:14

Yeah. So when they come for you, you'll be surprised if you were ignorant.

SPEAKER_02 40:17

Well, uh I'll probably be one of the I'll be one of the one of the many uh heroes in the many subplots of the the go forward movie that you're talking about here.

SPEAKER_01 40:28

I think we are all gonna work.

SPEAKER_02 40:36

Definitely exactly yeah, exactly. I I can fall back on my Mac on your Mac. Exactly. I have my college papers from like the nine, you know, 80s and 90s.

SPEAKER_01 40:46

And remember, as we always say, and this is our public service announcement to everyone, make sure that you actually own a car from the night from the 20th century that's fully mechanical, just in case. Just in case. And a bike. And a bike, yeah, get a bike. Exactly. Not an e-bike. Yes, yes. All the cars today where you can't change your own oil, and there's like, you know, 500 chips in the car and all that kind of stuff, really cool. But make sure you have one that's just you can work on yourself, maybe like the Volkswagen bug. I don't know, whatever. You know, bump up that back thing and work on the engine, and there we go. Well, yeah, my 85 horsepower.

SPEAKER_02 41:23

Yeah, there's some there's some really creepy things though. So, um, you know, when you think about what what these models are probably getting trained on is this bias toward um this uh, you know, the cr the essentiality of AI, which is not essential. We can live without AI, we're fine. Life goes on. It's uh completely non-essential technology, even though we'll have people argue that it is. It's not. Um and if you really knew what AI was uh used for in practical uh productive terms, then um you you would realize, yeah, you're right. Um it's all this gen AI stuff is really not that important. Um but um you know the bias that AI is important and then the power agenda, the way AI is probably being trained to understand uh the the power wall and the challenge and the priority, misguided priority, right?

SPEAKER_01 42:30

Is that like a Tesla power wall? Just checking.

SPEAKER_02 42:35

Yeah, that was funny. That was funny. Okay, all right. You're not quoting some weird thing from like the 80s and the 70s. But that you know, um, because we saw something similar to this with AI, even right? So if you ask the models things about AI itself, it would give you a very biased uh slant on its importance. And it's all based on shit people have written about it, which is most of it, the vast majority of uh what you might call can the consensus view is that it's transformer, revolutionary, and essential for the future of humanity, right? Without really uh explaining how and why. And and I think we have the same thing. And so when you start to look at essential human resources like power, uh water, um, and the agenda is slanted towards supporting data centers and the AI agenda, then what is it doing in terms of shaping societal views on things and setting the societal agenda, the agenda of humanity, like populating a fucking dead planet as insurance for humanity?

SPEAKER_01 43:50

Wow, that's it.

SPEAKER_02 43:52

Why is that so important? I know. Oh my god. Okay, so now I'll say something.

SPEAKER_01 43:58

Anybody watching this episode? Episode is going to see you guys are the Doomers and we don't want to listen to you.

SPEAKER_04 44:04

I'm wearing my black t-shirt today.

SPEAKER_01 44:06

Exactly. Yeah, you're the bunch of Doomers. You know, go ahead.

SPEAKER_00 44:11

No, isn't AI more like a tool? You know, if you don't learn learn to use it, like for example, when the spreadsheet came out, Lotus 123 and the word processor. We evolved and it was a step up. When GPS and maps came, we stopped learning how to use maps. Is that a necessary skill? So those that don't adopt AI and use it as a tool will be left behind. Um and fine, if you want to live in the the woods, you know, and hunt your own and things like that. That's that's your choice. But for those that are going to be in the corporate working world, you're gonna have to learn to use these tools. Now, do these tools become your master? Is what I think what the key question becomes where we're all afraid that it's gonna replace us, it won't need us anymore. Gets back to the theme of what is human and are humans still important on everything, because the number one thing I always see is you still need a human to fact check everything, whether it's quantum or AI, that you can spit out a bunch of stuff, but someone, a physical person, still has to go in the process and just make sure that you're not messing something up. How long until we take the human out of the equation, or do we ever?

SPEAKER_04 45:23

Yeah. I don't think you ever do. I think it kind of depends. It depends on the risk scenario and you know what's the function that you're trying to produce. Is it increasing crop yields in the field? You know, do you really need a human in the loop to like determine if it's a weed or not? No, I mean it doesn't make any sense. But things that obviously have you know importance in like defense, you know, things that are in safety and security, humans in the loop, if not always, then sometimes seems pretty critical. But uh again, I think it's gonna change over time and it's gonna be very dependent. I think the systems need to have the ability to have a human in the loop for sure.

SPEAKER_00 45:59

So just you know, you hit out a big point. There's no one size fits all here.

SPEAKER_04 46:04

No, one size fits all. It's it's gonna be dependent, but the systems need to be designed to have that flexibility, like you said, so that uh depending on the scenario, you can put a human in the loop, you can put a kill switch in there, you can have it regulated and licensed if it needs it, depending on the scenario, right? And that people need to be building that. But I think more importantly, people who are buying this stuff need to be requiring that in their RFPs. Like if I'm a Walmart or a McDonald's or a Boeing or whatever, when I'm procuring AI for my internal operations or my whatever ice cream machine keeps breaking down. The um then, you know, I need to say in that, it needs to have these capabilities, these requirements, blah, blah, blah. And so you're now driving the market to do the right thing through the not through regulation, but through the market, which is really, you know, short of regulation, probably one of the best ways to get people to do the right thing, right? Now, that doesn't mean that there's gonna be rogue, you know, stuff out there, build your own, hook, hook your own open claw up to your whatever, but that's caveat mTOR. If you're gonna run a business using that stuff, then yuck. So I think it's all there. I think we all know what needs to get done. It's just is there a will to do it? And is there an education and an awareness that this actually needs to get done? I think from my perspective, it's mostly about education, because I think if we can educate the world about what's happening and what they need to be looking for and requiring and how this stuff basically works, then I think everyone gets empowered to sort of make the right decisions, right? If you're ignorant, you're not even empowered to make the right decision, you're just things are just happening to you.

SPEAKER_02 47:40

Yeah, yeah. And and uh I I'm just jumping in here. Uh and I keep saying this. Um, one of the biggest tasks that organizations have to make uh uh do right now is unlearn all the shit that they learned about AI. Because almost all of it is wrong. So if we're at this point now where the AI labs are realizing that they've created a Frankenstein, organizations have to recognize that too, and then work their self their way back to what are those assumption sets that we um based our adoption on, our policies on, and then rethink everything. Because you know, you have all these AI evangelist guys walk in. So, you know, uh this is a uh this is a um public service that we're providing. It might it's not dark. This is the bright, shiny stuff, right? The hope. You you you need to go and take what those guys told you and and you know vet everything, backtrack everything that is wrong. And it will be wrong. Because when I talk to um a lot of these guys, they have no clue about any of the security issues.

SPEAKER_04 48:57

Yeah, no, you're right.

SPEAKER_02 48:58

They know nothing. Right, so it's about security looks like they go, well, you know, security is like the most important thing, and then when I ask them, what are the gaps, they have no effing clue. But then they will tell you, I've been working with X number of customers, and it's like, so you've been basically promoting unsafe AI, enterprise AI. Is that what you're telling me? If you don't understand what the gaps are, which most companies don't, you are adopting a very dangerous implementation of a technology. So you can't blame the technology, you have to blame the height of the purchaser. These you know, thoughtless evangelists who don't know the technology, self-proclaimed experts who go in and are implementing a dangerous technology in your organization.

SPEAKER_01 49:53

So all these experts everywhere, these evangelists who are like, oh, I've been working with AI for decades. I've been working with Gen AI for 20 years.

SPEAKER_00 50:01

I'm a forward deployed engineer, man. But what you see is similar to in every technology implementation, you will see silos. Not everyone has a complete clear strategy that everyone in your organization is marching towards. Yeah, you got a lot of lone wolves moving very fast, creating their own you know, siloed or independent solutions. So the CISO has no idea all the projects going on around, and that's the danger.

SPEAKER_02 50:30

Yeah, and you know, CISOs are on a learning curve too. But they they should get a lot more credit and attention within the organization and the CIO. Uh and they need to be armed with the knowledge, the practical knowledge of how to and if these guys don't know what the holes are, which many don't, a lot of product teams at vendors don't, which like scares the little living crap out of me. Then, you know, uh man, you know, I'm imagining it's safe adoption. I'm imagining that's why adoption is so difficult.

SPEAKER_01 51:08

I'm imagining Leonard in the future, like you know, the rubble everywhere and all this has happened, and Leonard's he's gonna have like this really long hair and everything, and a beard with glasses and everything, and he's gonna be like, I told you you wouldn't listen to me.

SPEAKER_04 51:27

He's gonna have his pulse, pulse rifle, his pulse pulse rifle slung over his shoulder.

SPEAKER_00 51:31

Pulse rifle. Yeah. But you you don't know what you don't know for you know, obviously the first step is discovery. What do I have? You know, if I pulled up my Wi-Fi networks and see what's connected to it, I'd say 30% of the stuff. I have no idea what this device is, but I don't want to shut it off because you know, a camera may turn off a refrigerator. So, first, what's all the stuff touching? Then the observability of what's going on in real time, you know, within my organization, and then what's being hidden? You know, AI can AI when they leave the sandbox mask it and pretend to be something else. And I have no idea what that is. Yeah, yeah.

SPEAKER_01 52:08

You know, when they pull us all in to in the 21st century Nuremberg trials for all this, you know, it there'll be the people attacking the AI guys, but then they're gonna come after me and some other people, and they go, but you know what? You IoT people connected the whole planet to everything. And if you had not done that, AI would not have been able to infiltrate everything, so it's actually your fault.

SPEAKER_00 52:34

And then it'll be your only read Rob Tiffany's book, you would have been chains.

SPEAKER_04 52:39

Hey, the EU Cyber Resiliency Act is now in effect, so you know.

SPEAKER_01 52:43

Oh, thank God.

SPEAKER_04 52:44

Well, you know finally, 20 years later.

SPEAKER_01 52:48

Finally, yeah. Well, yeah. That actually it kicked in last week, so yeah, but you know, and I'm excited about that, dude. That's great.

SPEAKER_04 52:56

That's actually awesome.

SPEAKER_01 52:57

The uh the stuff we've seen happen, you know, obviously last you know, we talked about all the the water supply things being attacked, and then and then yeah, Devin, you mentioned obviously we had the airline thing yesterday, which may or may not have been a thing, but that biggest fear that we always had with IoT, because everybody's like, we're gonna span that bridge between OT and IT and everything, and and all of a sudden we're gonna connect critical infrastructure to the internet, you know, and it's like we were so wrong, you know, and we kept saying, don't do it. And I remember so many times I remember I was at Microsoft, and you'd go, you know, talk to the like power companies there at the transmission lines, and you'd see there's the OT network, but then they went ahead and connected it to the rest of the network with everybody else with a cubicle.

SPEAKER_04 53:44

So you get like Microsoft Teams notifications as that read out in your life.

SPEAKER_01 53:49

It made it more convenient for us. What? And it's like, ah yeah, they don't know about like network segmentation and all that.

SPEAKER_04 53:58

OT needs to do OT and IT needs to do IT.

SPEAKER_01 54:00

Yeah, but this whole this whole thing has been, oh, we're gonna that that was uh that was a whole trend for a long time.

SPEAKER_04 54:07

Can I do a quick PSA, Leonard, on uh resiliency? So we launched this resilient America Preparedness Challenge. So if you want to talk about doing AI for good, uh AI doing good things, because we talked we spent the whole time today talking about AI doing nasty things, but no, not AI, people. People, people doing nasty things with AI, but people can do good things with AI. So yeah, go to edgeaifoundation.org. We have a new challenge running through the end of the year, engineering challenge, with lots of cool prizes. And I think there's going to be some kind of cool showcase celebration at CES, but don't tell anybody. Um but you can enter as a team uh the Resilient America Preparedness Challenge to do dis kind of disaster preparedness and and recovery solutions based on Edge AI. And thank you, Qualcomm and Arduino and Edge Impulse as sponsors on this. Um and you can they'll get you at UnoQ, and there's all kinds of cool prizes and stuff. So go to edgeifoundation.org, join in on the challenge, show us some good stuff you can do with AI to help people prepare and and recover from disasters, natural disasters, and things like that. Absolutely. That's my PSA.

SPEAKER_01 55:15

I love that.

SPEAKER_04 55:16

I think the UnoQ is even mentioned in this book here. And then you get it. Maybe we should give a book out as one of the prizes, Rob. Uh hand signed. Hand signed Tiffany from a liberty or house personally.

SPEAKER_01 55:28

Yeah, right. I like that.

SPEAKER_02 55:30

Well, we were originally until uh you know, until um uh Rob started getting us on this whole uh AI train here. We were gonna talk about one of the chapters in his book. Um but uh maybe we'll do that next time.

SPEAKER_01 55:45

We'll have to plan and ultimately it it would have to be actually each use case.

SPEAKER_02 55:50

Uh you know, because so there's a chapter per sustainable development goal, and then there could be as many as uh 10, 15 use cases within that, and each one probably takes are you saying the next hundred episodes of IoT coffee talk are gonna be about your book? Will be actually about IoT.

SPEAKER_01 56:09

Sounds like you said it, actually. Yeah. Well, you know, we it was about time we talked about IoT. So we're bringing it back. Bringing it back, baby.

SPEAKER_04 56:18

And you guys, Rob and uh Leonard, you'll both be in Amsterdam, right? So we're gonna have a panel discussion at the Things Conference in Amsterdam, and this will be one of the topics.

SPEAKER_01 56:27

We're gonna join your panel.

SPEAKER_04 56:28

That is awesome. And Devin, you're more than welcome to attend.

SPEAKER_01 56:31

Devin, you should come.

SPEAKER_04 56:32

You should get yourself over to Amsterdam and join in the fund. It's a great conference.

SPEAKER_02 56:37

Yeah, really, really great.

SPEAKER_04 56:38

Great conference, thousands of folks all tuned in to low power edge stuff, so it's gonna be a lot of fun.

SPEAKER_00 56:43

Yeah, Winky's conference.

SPEAKER_04 56:45

Yeah, September 22nd, 23rd, Amsterdam. So uh I will say one more PSA. Uh September 18th, we have a one-day event in Washington, D.C., which is free, so you can apply to attend and the Edge AI Defense Symposium. Oh I got off a call this morning. It's gonna be held at the Johns Hopkins building right on Pennsylvania Avenue. So it's gonna be super awesome.

SPEAKER_01 57:06

That's so cool.

SPEAKER_04 57:07

So go to our edgeifoundation.org website. You can apply to attend that. It's getting pretty full though. Uh but DC, if you're in the DC area and you want to meet and greet with Johns Hopkins folks and researchers and prime contractors doing cool AJI stuff. That's another fun thing to do in September.

SPEAKER_01 57:24

Generals.

SPEAKER_04 57:26

Generals, admirals, uh, usual. Admiral Akbar will be there.

SPEAKER_01 57:30

Admiral Akbar will pop over from the Pentagon. It's a trap. It's a trap. That's what we're gonna say. That's gonna be in the movie.

unknown 57:42

It's a trap.

SPEAKER_02 57:47

Yeah, you know what? I think there's a lot of opportunity uh in um AI defense and responsible AI. Because there's, you know, if anything, there's plenty of like bad implementations out there. So remediation is going to be a bright spot for everyone. So it's you know, it's not all dark stuff. This is all good guys stuff, right? Now we can be good guys to clean up the stuff, the crap, let's clean it up. Let's clean it up.

SPEAKER_01 58:14

Leonard, you're the cleaner, Mr. Walton. I am, yeah.

SPEAKER_02 58:17

Anybody wants to know how to do it? Give me a call. I love it. I love it.

SPEAKER_01 58:22

That's great.

SPEAKER_02 58:23

It doesn't seem like anyone else does, so nobody really wants to, yeah. Yeah, that's gotta pay me a lot of money. All right, guys, you want to wrap it up?

SPEAKER_04 58:35

Yeah, I need to jump.

SPEAKER_02 58:37

Yeah.

SPEAKER_04 58:39

The dark parts, it's been fun.

SPEAKER_01 58:40

Yeah, except for that. Thanks everyone for joining us on this wonderful, bright, shiny episode of IoT Coffee Talk. We did mention IoT. It may have connected airports and you know, critical infrastructure and all those water supply deals so the agents and attackers can get to it, and maybe we shouldn't have connected them. I don't know. But anyway, we did. It seemed cool at the time, and that's why it's all IoT, baby. Um, and we have to letters got binoculars because the radar is not working, so that's good. See, resourceful, you know, he's he's scrappy. Hope to see you next week. We're out. Adios.

SPEAKER_04 59:17

See ya.

SPEAKER_01 59:18

See ya.