IoT Coffee Talk
Welcome to IoT Coffee Talk where we talk tech and all things related to the Internet of Things. This program was started by a group of friends who just happen to be thought leaders, experienced practitioners and business leaders in the technology and communications services industries and more. This pod will feature weekly recordings of our casual conversations on the tech topics of the day with the hopes of delivering engaging and thought-provoking insights to our audience. We hope you enjoy program.
IoT Coffee Talk
#308 The FAFO Pandemic
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
There you go. A worthy intro.
SPEAKER_03Yes. A worthy intro. Transition. How's it going, man?
SPEAKER_02Good, good. In transit once again. You know? I thought maybe we'd all dial in from our various airports today, but I think this is just me right now.
SPEAKER_03Yeah, I know. Rob's on a flight right now, and no one else seems to be able to make it. So this could be the shortest IoT coffee talk ever. And um, oh well, you know.
SPEAKER_02Oh well, we know. Um plenty of things going on in the world. And they'll probably get like the most most views because well, you know, they say shorts are the new the new thing. So we can see the new law. But uh was it that open AI bought the uh bought that um podcast company? Oh, yeah, yeah. I I yeah apparently they have shorts, they embed advertising like AI-driven ads into their shorts at something. And that's kind of like an interesting use of AI, but still seems like it's like uh like a hundred million bucks. They should have bought IoT Coffee Talk, you know, for half the price.
SPEAKER_03Yeah, exactly. Exactly. Yeah, yeah. Um who knows? That might have made their their uh uh made-up valuation tank. Um anyway.
SPEAKER_02Uh the whole IPO thing, yeah, that's gonna be crazy. There's also like some interesting stories that whole Ronan Farrow came out this week with a whole expose on Sam Alton, and it's it's all getting buried.
SPEAKER_03Yeah, yeah. Look, I mean, we're going on year four here, and there's really nothing to show for it. Uh let's be honest. I mean, I can appreciate it anthropic. Yes, cogeneration, big big deal until people have to deal with all the slop. Sure. Um, you know, but anyways, before we get started, okay. Welcome to IoT Coffee Talk. This is gonna be the shortest episode ever, but remember to take us seriously at your own peril. Uh, and take out insurance if you do. We're here just to have fun, uh, for entertainment, mostly entertainment, slightly informational purposes only. And we're slightly, yeah, just here to just talk crap quite frankly. Yeah, and it's the weekend. Uh, it's been a long one. Um, my whole house has been under construction, it's been annoying as hell. Um set up with it, but I finally got my office done. And uh anyway, um yeah, we hope you enjoyed this. Uh what's probably gonna be a very short episode, but yes, I need to board my flight in about 15 minutes. So yeah, okay.
SPEAKER_02Maybe someone else will come up and maybe somebody else will jump in. Yeah. Uh yeah, yeah, another week in uh in the AI world. And um yeah, like you said, there's uh there's this a lot of smoke and noise. I think a lot of this now there's a lot of, you know, as you mentioned before, like because there's so much big money involved. Yeah, you know, these you know, these kind of crazy non-public market valuations, and um and now all this kind of you know dirt being slung around. And then there's like talking about SpaceX going public, you know, which would make Elon a trillionaire. Um so you know, no shortage of noise and smoke. Meanwhile, people are actually trying to make tech work and uh actually solve problems with the Garcuman.
SPEAKER_03Well, actually, meanwhile, people are trying to figure out how um how they can pay for their gas.
SPEAKER_01Yeah, tell me about it.
SPEAKER_03So with all this, I I posted recently, you know what? Um the last month and a half has proven that AI is non-essential. It's oil. It boils down to oil and energy. AI is not essential. And I think that is the question that a lot of the a lot of these people who've been investing a lot of money into the infrastructure uh don't want answer because it's an inconvenience. Not a very nice answer for these folks, right? To say you just spent uh and we have to remember where we came uh where we started with all this. It was FOMO. None of these guys, okay, three four years ago knew where this was going. They still don't know where it's going. All we know is ASI and AGI are nonsense. What's ASI? I don't know that one. Uh artificial super intelligence. Oh, super intelligence, right? Not general intelligence. Yeah, God forbid. I mean, I think it's more like artificial, stupid intelligence.
SPEAKER_02But I thought you were going to say ASDI, artificial super duper intelligence, which is like oh, what's the next thing?
SPEAKER_03No, it's artificial ridiculous IRI. ARI. Artificial ridiculous intelligence.
SPEAKER_02Well, it's funny, you know, the biggest use case. I mean, we talked about some use cases where AI is actually being used. Obviously, code generation is huge, and that's you know, that's a that's a seismic uh impact to all of our software development friends out there. But um you know, there's also um there's you know other things that are going on with it, like um I had a brain fart after I said software development. Uh oh yeah, so Google, look at Google Google's results. You know, it's search. It's search. It's AI mode and search is actually one of the biggest uses right now, consumers of tokens is uh AI what? Is is Google search, is basically if you look at Google's results. Oh yeah. It's it's it's used it's being used to enhance search, which is good. You know, again, not even close to the promise uh that we were promised.
SPEAKER_03You have to be still have to be really careful with their search, you know, like uh Google, whatever the uh overviews or whatever. AI mode. Yeah, the results uh they haven't improved. I mean, that's the thing. They have not improved. Uh and the thing is is we've become less uh sensitive and discerning mistakes. We've been more uh accommodating, which that shouldn't be the case. I mean, we have to remember when they uh introduced BART, uh or they uh you know they were fast following um uh copilot, or what was it? Yeah, co-pilot Siri, I think. Yeah, and then you know everyone was saying the the leader in AI, Gen AI, and then next thing you know, uh Gemini tries to uh you know uh up game them by demoing Barton. It got one you know one thing wrong, and people freaked out, right? That's true.
SPEAKER_02Now everyone just accepts all the errors without I think it made the uh Revolutionary War soldiers like people of color or something like that, so that it was it more aligned with like uh diversity, equity, and inclusion guidelines of the whatever. So that that was one of the big big uh flaws of BARD at the time. But that's uh, you know, a lot of the stuff's been fixed, but you're right, it's still you know data center investments way ahead of the curve.
SPEAKER_03Yeah, no, I don't think they've been fixed. Have they been fixed? I mean they still have better, the quality's improving. Yeah, that well uh it's uh it's in it's improved definitely. Um you know, but you're also scraping a lot more stuff, and the rag architectures have become much more um yeah, yeah. I mean, the rag architectures have evolved and improved over time, but you still have those foundational issues. Sure.
SPEAKER_02Yeah, and yeah, I was gonna say the uh we published this past week the Durga Malati keynote from our San Diego conference.
SPEAKER_01Oh, yeah, yeah.
SPEAKER_02He had some great data in there on kind of like it's almost like quality per billion parameters. Like basically, you can show, you can kind of graph it out, it's almost linear of the improvement in model quality as the size is going down. So the efficiency is going up, the quality is going up. So you can see like you know, the 10 billion parameter model today, which is pretty small, is equivalent to a hundred billion parameter model in terms of quality from a couple of years ago. So that's an interesting trend line to look at.
SPEAKER_03Yeah, but I guess the question is is it for general applications? And is the bet you know the benchmarking done for general? Because when you get smaller and smaller and you look at edge cases, you know, like edge AI scenarios and use cases. Yeah, they're more full. Yeah, it's more domain-specific. And so, yeah, it can improve. But you know what? There's nothing wrong with that. That's actually what we've been talking about on Copy Talk for years now. Years, not just, oh yeah, we figured this out last week. We've go back and listen to you know prior episodes on the AI. We've been saying this the whole time. You know, these large models are not going to be able to produce the level of quality and reliability needed for industrial applications. Yeah, especially for mission critical things. Yeah, I totally agree.
SPEAKER_02Hey Debbie, what's up?
SPEAKER_00Hey, what's up?
SPEAKER_02How are you? Debbie. Are you at it?
SPEAKER_00Are you Chicago?
SPEAKER_02Are you guys? You know what? I actually need to board, so Debbie, good timing. So I'm gonna hand it off to you.
SPEAKER_03Yeah, dude.
SPEAKER_00Have a good flight. Thank you.
SPEAKER_03You're gonna leave me hanging here, no picture, no nothing.
SPEAKER_00Oh, yeah, yeah. I'm gonna face for radio today.
SPEAKER_03Oh, really? You're giving it.
SPEAKER_00I'm not like you, Leonard. You just wake up, you're just naturally beautiful. Oh.
SPEAKER_03Don't even go there. I need cream, I need a recommendation. I'm my skin is like very non-hydrated. And look at this, I got like a turkey neck thing going on now. Uh I um lost weight. Yeah. And I got like 10 years older. It's terrible, but um, I don't know. I'm almost thinking I'll gain another 20 pounds so that um, you know, I fill in a little bit.
SPEAKER_00Oh my gosh. Oh my gosh.
SPEAKER_03My wife says, You need a facelift. How do I look?
SPEAKER_00Does that look yeah, that works. That can work. That can work.
SPEAKER_03Well, you know, it's always an option, right?
SPEAKER_00So totally, totally. Yeah, yeah. I think, you know, so when you were talking about the quality of the models, you know, they're never gonna be perfect because their their nature is not perfect. Um, but that's that's not to say that we don't want to strive for it to be better.
SPEAKER_03I think exactly the problem that I have is not that they're not perfect, it's that that people think that they are, and that's the it it goes back to the problem from that we talked about Debbie U and I, over trust and over-reliance, right? You know, you look at the use of AI, there needs to be discernment and you know that degree of skepticism that you always need to approach any technology with, actually, right? Uh and that hasn't happened. And that's the the problem. And you know, the thing is, is most people have no idea what AI really is, right? Um, they they just casually uh use it, use the tool. Um they in a misguiding way personify it and humanize it when you really shouldn't. So you're doing all these things that are um sort of no-nos, right? And we've saw we saw like patterns of this like two years ago where people were falling in love with their chatbot, you know, or whatever, their AI girlfriend. I don't think any of that is diminished at all. I think uh it's probably getting worse, uh and undoubtedly getting worse. And it's um I don't know how that is a sustainable thing.
SPEAKER_00Well, I guess it's a sustainable thing if people continue to purchase your product, right? And so it's not sustainable because of the types of mistakes that can happen. That's the problem. So I think you know, like if you're searching for for shoes on the internet and it doesn't get quite right, you may be pissed off, but that you're not harmed by that. But then you have someone using the same technology for medical stuff and it misses something, or it put gets you uh points the doctor in the wrong direction, then that you are harmed by that. So for me, yeah, those the areas where accuracy is is critical, yeah. Like I I would think twice before I like put all my eggs in that basket.
SPEAKER_03Yeah, and like one of the things that really bothers me is you hear a lot of um people, uh like especially uh people who are in analyst relations, who really should be, you know, looking at um leveraging analysts to get an objective and grounded view on technologies. Well, you know, the purpose that we serve is we serve as a sounding board for um what is a uh a better view of the ground truth, right? And increasingly all the marketing noise, the PR, the the vicious cycle of uh you know, the competitive nonsense spinning, uh catches up to you, right? At some point, you need the sounding board and then some reference to uh equalize and rationalize the conversation, right? So that buyers can make a better decision and they don't succumb themselves to um you know investing in nonsense, right? Which people have with AI. So I mean look at the software companies. Uh just a year ago, they're like, we're we're about the agentic enterprise, blah blah blah blah. Guess what? These guys they cannot they cannot sell this story anymore. They're collapsing uh under this, you know, completely contrived and misguided uh narrative about how SaaS is dead. Most people uh in and this is coming from my own experience coaching a lot of companies on cloud computing, most people don't really even understand what SaaS is. Totally. What cloud computing models are and what neo cloud is. They really don't. And um, I say that because you know, you have some journalists out there, they might interview a hundred people, whatever, for their book. In a year, I interview and talk to probably a thousand people, right? Yeah, and I can tell you there's consistent gaps in understanding, and you don't need to uh you don't need to talk to all the CEOs to find out what's going on and how people are thinking about stuff, right? Um, but you really don't. You have to have a diversity of uh exposure and who you talk to. And so I always love talking to engineers, product management teams to figure out okay, how well diffused is the better um perspective on emerging technologies or what have you uh across uh the industry, right? And you have to you can't just listen to what the PR teams are pushing through um you know these executive presentations and um podcasts and blah blah blah and their media appearances. You have to really go down a few levels deeper. And yeah, it people people just don't have a very mature understanding of this stuff.
SPEAKER_00No, and then it I guess it doesn't help that the the way the media is, they're always trying to tell a story that gets an eyeball.
SPEAKER_01Yeah.
SPEAKER_00So they kind of go in one one direction or the other that really you know skews or doesn't really give the full picture of everything. And so you have people people making bad people in in high positions making bad decisions based on hype, which is not truth. And not not listening to the data people who actually deal with the stuff every day.
SPEAKER_03Yeah, yeah. And then you know the funny thing is you you gotta think, uh, and so a few of the AR folks told me that they're they really are starting to rely more on um, you know, what these chatbots or these AI um models um uh, you know, you know, instead of like SEO, I mean it's literally AIEO, right? Right. Um and I think that's terrible. It what are you doing? It's so what you're doing is you're justifying and you're sub you're substantiating and promoting um the average. Like you know, uh it's just uh the the generic perspective that these models have just put out because statistically that's I mean Yeah, right, right, right.
SPEAKER_00Well, and then then right, and it's reducing a lot of traffic to the real places that people can go to to find a more nuanced perspective that isn't so sane, you know.
SPEAKER_03Yeah, yeah, and then it it all becomes a joke because once you use that as the benchmark for who has the the you know the most influential voice or the and influence should be based off of merit, right? It shouldn't be based off of how much money you pay to platform, right? You know, gin up uh you know metrics on your engagement and your blah blah blah, right? Yeah. Um and you know, actually some of the you know, folks that I consider the best analysts out there, they don't show up on any of these rankings.
SPEAKER_01No, right, not at all.
SPEAKER_03You know, people who really know their shit that I know um people listen to them, they do not show up on any of these rankings because they don't give a crap. And also, yeah, quality perspective is rare. It's not it's not the mean. It's no yeah, so it's true, it's true.
SPEAKER_00Well, I so I have something I want your thoughts on.
SPEAKER_03Yeah, I don't know anything, Debbie.
unknownCome on.
SPEAKER_03I just I just jump on your call and cause a lot of problems.
SPEAKER_00I always love to see you on my call. You're so you're the rabble rattler. I like it though. But uh, you know, because we need all perspectives, right? So um if we when we're developing stuff, we need to be looking at it from all different perspectives, and I I I like that in the sausage making thing so that we're not so tunnel vision, you know.
SPEAKER_03By the way, you are so good at this stuff. I have to tell you. Um you know a lot of the reasons why I jump on the the the call is I just like watching you work. It's like really cool. You're wonderful. Yeah you're you're awesome. So anyways.
SPEAKER_00So I posted something about deception and the reason why I posted it because that word bothers me a lot when people talk about it in an AI sense. And they're like oh this AI told it to do this and it did this other thing and it deceived me. Yeah. And I'm like but it's not a person.
SPEAKER_01Yeah.
SPEAKER_00It's like so why are you using like human language? Like so you felt deceived but the AI doesn't have a concept of deception in my view. It just is it has a concept of goals and what it's going to do and what you give it latitude to do. So it's going to do what it can do within whatever parameter that you give and if you don't have any you know if you don't have the if you don't have the the the thing that bounds the boundaries then it's going to do it does what it does period.
SPEAKER_03As you say that it scares the living crap out of me because the average person that drops like open cloth onto their device they have no it they it's literally I wish Bill was on an FAFO situation. Do you know what I'm saying? With potentially catastrophic consequences like this thing can like basically rip you off and then send uh you know send uh all your funds uh to some entity or its own bank account that it creates for itself and um and it transfers all the or a wallet you know a crypto wallet and then transfers all the money into its own crypto wallet.
SPEAKER_00Yeah there's nothing you can do there's a you're screwed you know you can't go back and say well I didn't tell it to do that.
SPEAKER_03That's the one it's like yeah you know that's why you don't FA and FO with this stuff you know um and so you know I had a really great conversation with the the folks at Lenovo um you know they have this agentic it's sort of a they call it their own AI but there's an agentic element to it but you know uh like you and I have talked about for years now Debbie the they're getting to this safe gen AI getting to safe agentic AI which we've been talking about for a long time as well is really tough.
SPEAKER_00This is why Apple and others like Lenovo and all these other guys have found this journey to get to safe agentic AI and generative AI in in principle really tough right really tough and um yeah the I I I you know I think it's important for consumers in particular to understand that these guys have skin in the game they from an ecosystem and uh you know the perspective of uh being responsible to their customers uh there's a lot at stake right but they are going to be more prone to pushing out safer um frameworks and tooling for consumers versus like people who are just embracing open claw and just going bonkers in fa and f oing uh to their ultimate detriment right right so and exactly yeah um exactly and then yeah it's it's nuts to me AI the way that it will eventually settle in my view and that's that's why I think some of these other these bigger companies like Apple and Lenovo are being a little bit more cautious is because I feel like if we if it's used right it's helping in the background it's not like the marquee thing so maybe you don't see it but maybe it helps you maybe it helps them do something faster or different you know what I mean or maybe it helps them streamline a workflow or maybe it helps them help do a product or something in a different way but that may not be the the marquee uh thing that you get at the end is not AI. It's it is it's something else.
SPEAKER_03I no I agree with you. So like you know when it comes to things like um network management for a telco right there's a lot of folks who are uh you know like TM Forum they have this whole uh maturity curve uh or you know like different levels of autonomy for the network right and everyone's trying to get to level four and stuff like that but the point of this might not be that you make the network operations like um autonomous you use AI to help scale out deterministic um configurations and uh process flows like what you're saying in the network and allows people to now um manage more complexity um in order to uh take their operations to the next level but most of this is going to be handling um a level of complexity that's being introduced organically by how uh things are evolving whether it's the network whether it's your security you know your network security um all these things are getting so complex and there's already a underserved problem right just like call centers you know these these agents I don't I barely um encounter agents actually it's weird you don't really encounter them what you do is you end up talking to people a hell of a lot more than an agent still after all of this hype and hoopla about how agents are gonna chill off like um call centers you end up talking to people I don't know about you people don't like AI so they you know like they send them the voicemail and they do all types of stuff like I like to pick up I'm in the grocery store I just pick up the phone just let it just don't say anything you can just say hang up right so but but I mean that this people didn't like that before the before people got crazy about AI people did not like that yeah I'm glad that some companies are going back to more human so yeah I mean you know first off a lot of these agents are just like wordy as hell so you're spending a hell of a lot of time just reading a ton of stuff you know what I'm saying it's like look I just want a yes or no answer. Don't give me like a whole friggin' lecture on a simple question. And so it it actually becomes a very cumbersome interaction I know a lot of people out there will say hey you know it's because you didn't configure XYZ blah blah blah well well who's gonna who's going to continue to maintain these you think the AI is going to do it itself? No, you need to have well trained professional um call center folks who know how to handle all the different situations who go through exercises uh uh uh about uh to basically improve the overall customer experience now the system can give you metrics and some data to help you but it's not always the case that data just because you have data means that you're going to make good decisions a lot of times the data is misguiding.
SPEAKER_00Yeah totally yeah like just like the the AI uh it's so funny because like if you have to call a company or whatever and they have like some automated message and then they say okay pick these three options and like three options are never what you want so you're like okay I'll be Gerald so the three straight to the person Nobody chose the three options like a hundred percent of the people chose the other option so I've got to go back to the drawing board like yeah no the worst is like when you you do engage with a chat bot and it gives you a long ass answer you have to read the whole friggin' thing you get to the bottom and then maybe you have a follow-on question to to uh trim the response or get the agent to uh get closer to what it is that your problem is or the reason for your call and ultimately it says well I can't help you and we need to connect you with a a person right uh a live agent and it's like okay so what was the purpose of all of this?
SPEAKER_03But I I mean I do see a lot of um utility and you know you and I have talked about this before a little bit more of a conversational interactive FAQ.
SPEAKER_00If gonna help with that totally that's cool but it should reference and bring up real content so that the user knows that they um you know the customer knows that this is not just an interpretation you know uh and yeah the AI is getting really complicated because now they're trying to uh they're trying to extend the memory of these applications right I'm not gonna call them models it's literally the application to you know with the KV cache so that there's consistency in how they respond it's still not perfect the whole reason for it is more uh for reducing um uh the number of uh times that you actually have to uniquely process a similar or the same um you know prompt right uh the answers can still be wrong inconsistent uh yeah and you know uh it it it's it you know seriously this is like over engineering right yeah taking a determin uh probabilistic technology and then trying to make it deterministic or to have deterministic expectations of what that technology can do and these are the two fundamental misconceptions that is creating a trillion dollar mess right well well yes it let's create a mess on two parts so one is yes it's creating a mess because people are creating systems that aren't actually working the way that they think it should be working or the way whatever but then the part that offends me is like basically have these companies have hoisted their uh their work on me so now they're giving me a dollar so when I call you and I spend an hour on the phone yeah I'm I'm working for you now yeah okay so it's like I'm not getting a paycheck I don't get healthcare nothing yeah I mean think about all the interactions where you have where you're where before you could get a person and you could get an answer and be done instead now you have to do all these gymnastics to try to hopefully you know it's like win the jackpot hopefully you can get a person you know you have to find out the secret password to get like an actual human on the but on the phone but literally it's like I'm I'm taking you're hoy you're you're putting yeah technology as a roadblock to me getting my problem solved but then the work is on me now yeah not on you.
SPEAKER_03Yeah well that that's the whole I mean that's an age old uh value proposition of CRM right that's like self-service right that was that whole thing I mean um and you know it's always been it's been challenging it you know even back then the whole idea was you could um you could eliminate your call center because now you have self-set service capabilities right um that was what they were doing with um IVRs um the whole I mean this stuff goes way back actually but you know the bottom line is it never really eliminated the call center because you know at the end of the day there is a there is a competitive uh driver to having great customer service operations and part of that is let's say that um you have a company that goes overboard with let's say CRM technologies or even this agentic stuff uh it reduces and it ultimately fails to deliver right um and um doesn't address the underserved and unserved problems that becomes an opportunity for a competitor right because you're you're lowering you're lowering the floor for yourself and that basically creates a differentiation for any any of your competitors who are going to invest in let's say people to to have more of a hybrid uh operations that blends people with technology right and technology enablement and so the these people who have and we've already seen it several times where people have gone off and said oh we're gonna get rid of everyone on a call center and the next thing you know they have to hire everyone back right right because the person if a person picks off the phone and calls you they've already seen your FAQ they've looked at your website already right so why are you regurgitating this again?
SPEAKER_00Yeah yeah they've already seen that so if they call you that means they have a special thing that they want to talk to you about that you have not covered.
SPEAKER_03Well or it can be a simple thing where oh my God I why am I waiting this long just to get a simple answer right and that's where um you know a a a simple answer to a simple question basic question and those are the kind of the um scenarios where you do want to figure out what is the best strategy for handling those types of um those types of interactions right or um engagement and I think that still takes a lot of discernment and that's where AI can be very useful in that you can look at all of the different scenarios and you can use it as a tool to help create the rules for your software and this idea that everything, all software is going to be run on GPUs using models is completely disconnected from reality. It's bonker bonkers you you actually want to use generative AI as little as possible. You want to use it as a tool to augment existing systems and existing ways of doing things because you know the the thing is is the AS400 or what you know like a Z series mainframe for transactional purposes is still going to be way faster than any um you know a GB300 MVL 72. Yeah sorry right right because the because the thing is it is not thinking it's just acting right hey doing transactions as fast as possible and um and people don't understand that there's different types of computing okay and what I mean by people is in general you know audiences out there who've been sold this Kool-Aid that everything is going to move toward AI.
SPEAKER_00No, it's not uh in fact we're seeing that that the complete opposite is happening there's more concentration on uh traditional computing when you look at these um these agentic especially these agentic um architectures right and systems yeah yeah there I don't know I don't know it's it's uh like I heard like a big uh the CEO of a big insurance company I guess he has saw he has seen a demo of like uh uh uh a gentec AI or something or a gen AI or something and he was like oh yeah we should just put all of our insurance acclaim stuff in this thing it's like oh god no it's like oh my goodness this is such a bad idea it's such a bad idea but you see how a lot of bad ideas get pushed down and we're still not seeing the success rate of these projects go up at all um yeah that's so it's like it's like when are you all gonna get real and figure out what your real problem is and stop trying to throw mash everything into AI. It just doesn't work that way.
SPEAKER_03Well you know like RSAC when I was there um one of the concepts that I've introduced is um you know this continuum that we've gone um through or our the cybersecurity industry is going through uh has been going through in the past few years is you know this uh initial elation about generative AI um so everyone trying to use it as a force multiplier for handling and um supporting you know security ops right and whatever a whole wide range of things whether it's threat intelligence blah blah blah it's gonna do magic toward holy crap this stuff is gonna be super dangerous and then now elation moving toward we got to figure out how to isolate things how do we do that you know uh and this started off with LLMs now agentic AI is taking things to the next level in terms of uh threat potential and you know blast radius in terms of impact deep and wide right and at machine speed not yeah speed friggin' machine speed then this year um I I am making the uh prediction that everyone's going to be crapping their pants so hard that they're going to move toward compartmentalization so how do we not consolidate all the information and train everything in a single model that can get compromised and next thing you know uh a threat actor can come in and basically distill the whole thing and run off with it and then put it on the dark web you compartmentalize and then you secure and isolate so that you know the blast radius is not as um huge and you mitigate the damage that can be done.
SPEAKER_00Right. So my thing always was so like example like I always hear a lot of people they complain about third parties where oh my god my third party got breached and that breached us or whatever. I was like but it should be a limit to how much damage that they could do even if they're in your system. Yeah why don't you think about that and so to me this is the same thing but at a broader scale it's like if there should be a limit to the damage that they could do. So it's not like let's create all these capabilities and then figure out how to how to um you know create boundaries and like you need boundaries first yeah first before you start doing all this other type of stuff. It's like it doesn't work that way. Yeah yeah and so um thegentic stuff i i is gonna definitely go um it's gonna go sideways um well and then I'm like how can you someone had told me once they were like yeah it's like you know the analogy they gave is like you know you have an intern that you give tasks to with these tools. So I said but the intern has more network access than you do and it has less knowledge than you do and it can keep a lot more than you can
SPEAKER_03So here's the here's the problem. Number one, you expose your you're basically practicing FA and FO, right? You're fucking around and you're gonna find out.
SPEAKER_00That's right.
SPEAKER_03Which is super dangerous. What you should do is you should be testing all this stuff, right? Before you FA and FO in production or on any of your own stuff on device, right? Most people won't do that, especially consumers who have no idea. They just sat they go and they drop this thing into GitHub. Number one, uh you know, most people don't even think of security and privacy. That's an afterthought. And you keep seeing all these people going, well, you know, they know everything anyway. It's like, no, you want to take that to the next level? Um go right ahead. Right. And FA and FO. Um, but here's the thing: you can't test this stuff, that's the problem. The cost of testing and the amount of effort testing, whether it's human-based or you're using tools, including AI, is cost prohibitive to get to safe.
SPEAKER_00Yeah.
SPEAKER_03So no one's gonna do it. Um it's gonna be a friggin' and then you know the the security infrastructure tooling, once deployed, there that stuff doesn't exist.
unknownRight.
SPEAKER_03It doesn't, it doesn't friggin' exist. People are working on it, you know?
SPEAKER_00Right. But I mean, wasn't uh giving giving a tool ABI and access to your computer was a bad idea 30 years ago.
SPEAKER_03Yeah, pretty much.
SPEAKER_00And it's still a bad idea today.
SPEAKER_03Yeah, yeah.
SPEAKER_00That hasn't changed.
SPEAKER_03No, no, it hasn't. But you know, the thing is, is we've taken this whole data breach thing, we've been come we've become so desensitized to it, not realizing that there is going to be a price to pay at some point. And the price is now, because uh that dark web is hydrated with almost everything it needs to be it to basically compromise everyone on the planet who is digitally engaged. So the only option now, going back to compartmentalization and isolation, there's going to be at some point a need to isolate yourself. In other words, battlestar galactica, and then compartmentalize. So once the Cylons infiltrate the Battlestar Galactica, you want to be able to contain their spread within the ship so that you can reach Earth at some point.
SPEAKER_00You know what? But you know, a lot of this science fiction is turning into science fact, isn't it?
SPEAKER_03Well, they were they weren't dumb. They people, I mean, you know, these writers that sci-fi writers, they futurists, yeah, they they kind of got it, you know.
SPEAKER_00Yeah, they nailed it on a lot of points.
SPEAKER_03Yeah, they did. And you know, uh, I don't know what they must have smoked a lot of weed or something. I don't know. It's pretty amazing, you know, whether it's like Terminator or Blade Runner or you know, um any of these. I mean, if Mob was on, he would be listing off like a whole bunch of stuff. Logic Matrix.
SPEAKER_00Well, actually, the um it's funny because like minority report, they said that Steven Spielberg, he had talked with a lot of futurists when he was doing the film. And he was like, So what's happening, what's gonna happen next 50 years or whatever? And that's why I think that movie holds up so well, because a lot of the stuff that was in there.
SPEAKER_03Yeah. Yeah. And you know what? If you really think about it, that they highlighted the privacy issue very early. Because when Tom's Cruise uh character, I think John, whatever, um, he he was trying to go battle Star Galactica. He wanted to get off grid. Yeah, how did you do that when you're getting retinally scanned and so you had to get new eyeballs? Right? You have to actually get eyeball transplants.
SPEAKER_00Yeah, that's so inconvenient.
unknownYeah.
SPEAKER_03But we're both beyond that, and then Gattaca as well, right? Gattaca. Yeah, I love Gattaca with um uh Ethan Hawk and uh Uma Thurman. That was a great movie.
SPEAKER_00That was a great movie. Well, you know, and there was a point, so like the Jude Law character in the movie, he is in a wheelchair, but he has I guess his DNA and his essence has all the good stuff, the good genies or whatever, but that doesn't for him, it didn't mean that he could achieve at that level, right? So it was very interesting. It was kind of like a cast that have been created, and so I think there are definitely parallels there in the way that people are going around AI for sure.
SPEAKER_03Yeah, and he cremated himself, right? Toward the end. Oh, that was like what kind of that well, you know, um these sci-fi writers, um, these futurists, you know, sometimes they get it or they've gotten a lot of stuff. But then, you know, it's it it they're usually getting it right because uh they're not just looking at technology, they're using they're looking at there's societal factors that they're looking at, um political, geopolitical even. Um you know how are how will um what what will governments look like? How will societies organize themselves as they're being impacted by technologies? Yeah, yeah.
SPEAKER_00So I thought about the movie, I'm I'm not sure if I told you about this, about 2000, maybe someone else had thought about this. 2001 Space Odyssey with Kubrick. And the scene I think that is it reminds me of what's happening right now, which is when the astronaut they had decided they want to shut Hal down, and he went into a different room and he was talking about it, but Hal was reading his lips.
SPEAKER_01Yeah.
SPEAKER_00It's like, okay, you thought you were being smart because you assumed right, so this is another human thing, right? Let me go in the other room so he's not hearing me, right? So but he's like, we don't get we don't understand what we're dealing with, really.
SPEAKER_03No, and and you know, the the thing is is because we become so over-reliant and over-trusting of the technology, it's really gonna cause a problem for a lot of folks. Because, you know, when I look at some of the transcripts that are produced by these AI tools, and and you know, they're using the most advanced models out there, at least reasonably um you know capable models uh out there today, there's a lot of errors. Um in fact, you know, these transcripts, if these are used to train future models or referenced um or referenced in uh any kind of uh query um that and fed into the context of any uh you know request, you know, it could be a really problematic. I've seen this thing, you know, just even on YouTube, put like you know, curse words in people's uh, you know uh transcript when that's not what they they didn't use a curse word.
SPEAKER_00Right.
SPEAKER_03You know what I'm saying? It's like okay, now at some point um that person will be victimized by that error, right? Right, exactly. Because people are trusting AI, totally, and then so the AI will make a judgment based on bad data about what per a person has said, yeah, or it it or incorrectly has been accounted as having said, and think about that victimization, you know, that's right. Totally, yeah. I had a and this is happening at mass scale. It's not like oh, maybe it's happening. No, uh I can show you where it's happening, where like uh they had um they had somebody saying that uh you know the Transcope basically you know um had an actor uh uh saying the N-word.
SPEAKER_00Yeah.
SPEAKER_03And you know what you uh you know what you know what what he was saying? What? Arnold Schwarzenegger.
SPEAKER_00Yeah, right. I remember that.
SPEAKER_03Arnold Schwarzenegger instead it put the N-word in there, and I'm like, oh my god. Right, right, exactly. You know, so did uh you know, did this actor ever use the N-word? And and then the uh AI is gonna say, yeah, it did. Right, he did on the Jimmy Kimball show or whatever, and here it is, and then you'll see the transcript, and somebody will go, Oh my god, he used the N-word. Right. Totally, right? The only way you can prove that, uh, and then the damage will be done, and the only way that you can prove it is somebody goes through the effort of watching the entire episode or that episode or that clip, and yeah, oh, he said Arl Schwarzenegger. Where's where'd the N-word come from?
SPEAKER_00Right, exactly, right. And the only reason why we know this, like, so this could happen millions and billions of times with people who don't get the press. Yeah, totally.
SPEAKER_03Or it's not, oh, will it happen? Or it's not happening at scale, oh, it's happening at massive scale.
SPEAKER_00Totally, totally, right? Like, I have a uh a note taker that I use for meetings and other recordings, and and they have this thing now where they try to score you, like, okay, you have high engagement, high, you know, they're trying to score you based on like the the conversation. And anyway, this one I mostly ignore it, but one time this one conversation I had, and they said I had low something. Something was low. I said, What the hell are you talking about? And so I looked and and it said, and I was like, I think it said low engagement or something. And I was like, Oh my god, what the hell are you talking about? And it's found out what it meant. They said the other person talked more than I do. It's like it's an interview, yeah.
SPEAKER_03Oh my god. So imagine more than me, right?
SPEAKER_00And I'm thinking, like, what if like what if some some companies using this to rate their employees? Oh my god. I know you have low engagement because this tool said it's like I'm interviewing someone, they talk more than I do.
SPEAKER_03Can you imagine what's gonna happen to meetings now? You're gonna have all these people just talking shit, you know, for the sake of having engagement, you know? I was like, ah hey, you know, I have no idea what I'm talking about. I'm gonna say something, and then everyone has to waste time listening to somebody uh just trying to make their quota or you know, trying to up their their score.
SPEAKER_00Right, right. Oh, so bad.
SPEAKER_03I I have a comment. Uh I'd like to know what we're having for lunch, and uh yesterday's lunch kind of sucked. Uh, but it would be really great if we had like a FAQ or like a an agent that we built that could take our preferences and uh you know on a weekly basis, blah blah blah. You know what I'm saying? It's it just get like ridiculous.
SPEAKER_00Totally, right? And all those things are very like I have to really start for this, right? Because like it didn't really explain it, and I had to look at it. So no one's gonna have time to go look deep down to find out you know where this started.
SPEAKER_03No, no. I mean, but and it goes back to the cybersecurity problem. It's asymmetrical. When you get harmed by AI generated misinformation, right, about yourself, the damage happens, and it costs you it, it it costs nothing for that to happen.
SPEAKER_01Totally, right?
SPEAKER_03Yeah, but it will impact your life, it will impact, it will cost you um a ridiculous amount to remediate that situation for yourself, you know, and they don't give a shit, you know. The AF doesn't give a shit. It just goes, well, you know, I don't know. I didn't know Dave.
unknownRight?
SPEAKER_00I'm sorry that I emptied out your bank account. Oops.
SPEAKER_03Oh yeah, well, you know, I've you didn't tell me that. I don't know, you just FA'd and FO'd, you know?
SPEAKER_00Deal with I sold your house, so you can go on that that tour around the world that you wanted to go on go through.
SPEAKER_03But uh I I don't I don't know. I mean, you know, it's like we've we sound like um Luddites, but it's not. It's we're we're talking about safe. And you know what, it's not like things have really gone the way that the people who have hyped the technology have, you know, said the reason why things aren't going as quickly, or there's adoption issues or safety issues, is because of the things that we keep bringing up. It's like, look, you gotta you have to address these things before you get to safe and uh you know uh into you know AI used uh in a way that is uh constructive and positive and in you know life enhancing.
SPEAKER_00Um I think that right. So I think adoption on the long-term scale really does depend on trust.
SPEAKER_03Yeah, absolutely.
SPEAKER_00So the trust is not being built the way that it should be, and so that's the problem.
SPEAKER_03Yeah, and then it goes back to what we were saying before about expectations, right? Those two points. There's expectations, and then there's um what um the technology actually can do, right? And um when there's a disconnect, you can never really have trust. And then everything becomes untrustworthy, and that's gonna be the biggest challenge going forward. And we already see this. This is not something that, oh, you know, this that won't happen. No, it's happening. Totally. Go and look at some what's happening with your banks.
SPEAKER_00Oh my god.
SPEAKER_03Well they what they're doing is they are basically shitting their pants on totally hot.
SPEAKER_00Well, and they should. They should. Banks, whenever a new vulnerability comes out, like the the people who are the evildoers, yeah, the first thing they do go for is a bank. Or they go for someone that they can get their bank account. Yes.
SPEAKER_03Yeah. And then, you know, what do these people still do? Yeah, what can you uh, you know, what's the what's your what's the name of your pet?
SPEAKER_00Right.
SPEAKER_01What's your name is called?
SPEAKER_03What the f it's like, okay, what's your phone number? And then they authenticate you off of three things that are already on the dark web that anyone can find. It's like public information. It's like, okay, what?
SPEAKER_00Probably, right. That's public voice too. So it's like, oh, it's your date of birth.
unknownLike, what?
SPEAKER_03Come on. You know? Um, there's a lot of institutions that still do that. And so, you know, and then think about it. Now, um with Claude uh Claude, uh, what is it? Uh Mythos.
SPEAKER_00Yeah.
SPEAKER_03Anthropics mythos model. Now they're everyone's freaking out about that thing. It's like, no, don't you should have freaked out a long time ago. That it just makes things better, but you should always already have the concern uh that these tools can be used for vulnerability discovery. And that whatever you thought uh would be useful in enhancing your security posture, or at least giving you visibility to what you needed to work on, front actors are gonna use the same thing. Totally. It's cheap for them to exploit existing vulnerabilities.
SPEAKER_00Totally. Right. It's true, it's true, you're right. It's true. It's like it's all funny. So when all this stuff is happening, I'm like, so what exactly are people confused about? What are you confused about? It's like you create a capability that's so out there, but you don't have a way to contain it. Are you or it's doing things that just Bill needs to be here?
SPEAKER_03Bill needs to be here with his hat and his shirt.
SPEAKER_00That's right.
SPEAKER_03That's A and F O epidemic. That's what we gotta call it. Right?
SPEAKER_00That's right. That's right. That's this is really the gap of the ugly. This is the valley. Oh, yeah, yeah.
SPEAKER_03No, this is just like friggin' nasty.
SPEAKER_00Oh yes.
SPEAKER_03Oh my god.
SPEAKER_00We need to go back and watch sci-fi movies because they were trying to send us a message that we didn't get.
SPEAKER_03We didn't get oh well, you know, we're we're we're just not gonna you know that here's here's the thing. I just encourage a lot of um the vendors to focus on um what the real demand is going to be, which is how do you protect yourselves from this stuff? Because enterprises are going to really struggle, especially financial, any kind of um quote unquote trust-based institution, whether it's government or banking or telco, they're gonna have to invest a lot in this stuff, uh the protection, right? And they have to really tune into how bad stuff is gonna get. And you know, yeah, mythos, you can be afraid of it, um, but you should already be afraid. Right. If you're if you're not afraid, you're behind the ball here, you know, you're like behind the curve. Um be shitting your pants right now, and then talking to your board, going to your board and say, look, there is some friggin' crazy ass shit going down, and I know you guys are pushing AI like it's fentanyl, right? Let me tell you something. You know, this mythos thing is just the tip of the iceberg. There's already an iceberg that you need to worry about. Forget about the tip. There's a totally and that's that's the education that is not happening. Right?
SPEAKER_00It's true, right, right, exactly. And so and think about the scale. We're not talking about one company or one tool or one capability, we're talking about the ability to do this at scale.
SPEAKER_03Right. In a lot of different ways, and then think about how much of a threat surface or a vulnerability, yeah, a threat attack surface the consumer is gonna become. So you have all these obligations as a company to your customers. But if your customer is completely irresponsible, is F A and FOing on all this stuff, creating all kinds of uh vulnerabilities for themselves, what are you gonna do? You're gonna do what a lot of banks are doing now. If you are making a even a um a transfer, which used to be guaranteed and not a problem, they're going to basically put the they're gonna do the self-service thing. If you want this transfer to happen, you take responsibility for if there's any fraud that happens because that we cleared this transaction for you. When did that ever happen? Right. I was like, wait, why do I bank with you guys if you're gonna sit there and not guarantee anything, you're not gonna be accountable for any anything, and you're gonna put it on me. Right. It's like, oh, eventually they'll say if you have anything on deposit, we're not responsible. You you need to be responsible.
SPEAKER_01Like, what the fine It's like you say what exactly, it's happening now, not tomorrow or ten years from now.
SPEAKER_03It's already happening, you know. Right, it's true, it's true. And then you have all these people screwing around.
SPEAKER_00Yeah, like we need to get real.
SPEAKER_03Yeah, and this is why I going back to like what we were talking about before, it's important for like I think great opportunity for the likes of Apple, uh Lenovo, who else is there? I mean, you know, um for again Google, if they they actually lean into this stuff instead of promoting, you know, the stuff that causes risk. Great opportunity. And privacy, security, how do you isolate, how do you comportmentalize? Um important for individuals because we're gonna become easy targets.
SPEAKER_00That's true. That's true. I agree. I think I've always thought this was the opportunity, so I'm glad that we can I know you didn't that's why we love each other, Debbie, you and I. Yes, yes, yes. We're always ranting with our fists, you know, yeah, and then everyone is saying that ah you know yeah you well you know people somehow come around to our mode of thinking in like a couple of years. It takes about a year or two before they catch up what we're trying to say.
SPEAKER_03Yeah now you hear a lot more people talk about trust you hear a lot more people um talk about privacy um they they still need a lot of help figuring out how all this stuff fits together but um they are talking about it and they weren't talking about it four years ago. That's true.
SPEAKER_00And they were absolutely not talking about um privacy um eight years ago nobody really knew they weren't nobody was nobody was talking about it. So I'm glad that people are talking about and you know I've gotten some calls over the last couple like weeks or so from consumers. Yeah they're like hey what the hell what is happening right so I feel like it's actually sinking in because people are feeling impacted by it. So yeah.
SPEAKER_03Yeah yeah I mean you know um I I I think that value brand you know from a brand perspective it's a huge plus I mean it's worked for Apple it should hopefully work for other um brands out there as well that can uh you know be responsible and provide solutions to consumers but also enterprises organizations there it's gonna be like uh because you know eventually mythos is gonna get out there and the cost of remediating all the vulnerabilities in an organization's um IT is going to be cost prohibitive you're going to have to prioritize what you lock down confidential computing all these things that people didn't give a shit about uh even though there were things that they should have given a shit about right yeah it's true it's true yeah oh my gosh well thank you so much for having me oh my god what are you talking about you're you're part of the gang there's nothing that you nobody you don't need to thank anybody we we just jump on right we do we do we do I'm always happy to jump on uh it's always great to have you I'm glad that we had this chance to vent and by the way everyone we're just venting yeah we're venting we see all kinds of ridiculous stuff through the course of the week and sometimes we just have to let loose that's true yeah that's true we did it we did it I feel calmer now I feel calmer now yeah just getting a lot options bar never mind so but hey everyone if you made it this far congratulations that's awesome yeah yeah but hopefully you you you got a lesson and what might happen if you fa and f o that's right this is a shout you know shout out to Bill for being the prophet and the purveyor of FA F O not in the bad way in terms of a warning to people right right on F A and F O make sure you know what you're doing and that you really engage with trusted parties as you look to AI enable your life or your organization right there's a lot of people out there who really don't know what they're talking about uh selling stuff that ultimately you're gonna want to refund on yeah right so be very cautious about who you trust and vet their backgrounds uh because there's a lot of people who just I'm telling you they should not be calling themselves AI experts of any kind. Uh and then um yeah and remember to that we do this every week uh regardless of any any of our travel schedules and stuff even though we're really busy but we appreciate your viewership uh we appreciate your listenership and we hope that you enjoyed this very very long ass conversation but uh we hit a lot of really good stuff right Debbie yeah we're very timely always always always ahead of the curve right exactly exactly and so remember to like share comment even you know especially if you have something constructive um even if it's not that constructive yeah go ahead and make a comment share your thoughts your reactions whether you agree or disagree you know um we're just trying to spur a debate and uh introduce uh maybe some angles on the topic of uh AI and other tech topics that um are getting too much hype not enough um grounded uh angles uh considered in those those discussions and uh we'll see you next week all right uh so take care we'll see you Debbie all right have a great weekend
