[00:00:19] Everything is about AI. Well, I'm host. Happy to have you along, my fellow Americans. We keep an eye on techs if you don't have to. This is TechWatch. I'm Sam Bushman, Jay Harrison with me. We talk tech every time we get together and it's a fantastic time. Hope you enjoy it as well.
[00:00:37] This broadcast is brought to you by NetworkProvidersInc.com. When you think IT, think about just a bunch of consultants together that have knowledge in all kinds of things. HIPAA compliance for the medical industry, AI knowledge, security knowledge, knowledge of routers, a help desk, endpoints for security and protection, and the list goes on and on. Think NetworkProvidersInc.com. They're your partner.
[00:01:05] This broadcast is available after the fact at NPITechGuys.com. That's NPITechGuys.com. And we're available on Twitter and YouTube. Oh, that's X now, right? X and YouTube and Rumble. And we're available on the radio at Liberty Roundtable. I'm sorry. Wow. NPITechGuys.com. And on the Liberty News Radio Network. We're just everywhere, folks.
[00:01:30] And if they're somewhere we're not, tell us and we'll get there. How's that? Spread the word, would you please? Share the broadcast with your friends. Hi, Jay. Hey, Sam. How are you? I'm in Utah and I feel like I have Florida winter weather right now. It's pretty nice up there from what I hear. 75 to 80 degrees today. Sunny and beautiful. I'm just saying it doesn't get better. And it's not only that, it's dry weather, so it's not all humid or anything. It's just beatimus, sir.
[00:01:56] What would it normally be, like 80s or 90s? Is that about this time of year? No, it's probably about right. It's the spring, you know, May 1, May 5 kind of rain. It's not too bad. Yeah. It's pretty typical, but I'm just telling you, it's great weather and we love to see it. Sometimes it's rainy during this time. When you get spring, a lot of times you get a lot of rain and stuff like that. But it's a beautiful, beautiful sunny day. April showers, you know, bring the May flowers, right?
[00:02:21] Yeah, we're waiting on May flowers right now, buddy. All right. Amazon announcing that it's opening its full logistic network. Yeah, this is kind of crazy. I don't know how to respond to this, Jay. They're basically saying they're going to take on FedEx, UPS, everybody. All the other company stocks are down over this announcement. They're going to use their full infrastructure, their network of delivery and transportation and communications and logistics and everything.
[00:02:52] And, you know, they did this in competing and Amazon's become a big force to reckon with. Now they're doing it with shipping. That's right. What do you think of that? Well, I think it's great news. I mean, it's good for competition. We've had problems. A lot of people have lately with especially FedEx, just poor deliveries. You see these TikTok videos and places, reels and stuff where people are just throwing crap around.
[00:03:15] And, you know, not to mention the news story where the FedEx guy kidnapped a girl and like all of a sudden, you know, not that that's FedEx's problem or issue. You know, they can't. I'm sure they do everything that can help that. But, you know, it just bringing competition to UPS, FedEx and the post office, I think is a good idea. So they're saying any business that needs shipping services in their directly challenging UPS and FedEx and both of their stocks, UPS and FedEx, both fell 9% on the announcement.
[00:03:41] You know, that's not great for them, but I think overall for the market and consumers in general, I think the competition will be good. Amen. And what's interesting is a lot of these UPS's and FedEx's have only been shipping. And a lot of these Amazon things provide warehousing and a lot of other infrastructure and logistics support and things like that. You kind of wonder if some of these other companies are going to back into some of that 2, 3PL, third party logistics and everything else like Amazon, too.
[00:04:09] You know, sometimes it's like go big or go home. And if you don't play there, it's going to be hard to compete with a piece of the pie, Jay. It will be. They'll keep their rates up, though. So we'll just see it'll have to work itself out with Amazon bringing on AI and all kinds of other things. Competitors are going to have to compete in the same realm and they're going to have to innovate as well. Amen. Anyway, we'll keep an eye on that ball, ladies and gentlemen. But we found that, in my opinion, very interesting.
[00:04:38] And it's quite high tech, too, Jay, in all fairness. Yeah, it is. Absolutely. I mean, it's the way they're doing it is they're basically able to track exactly where your package is. I mean, you can literally see sometimes when the package is on your block, right? I think that they pioneered also taking a picture of the package when it's been delivered. And I'm seeing that now with the post office and with FedEx and UPS now. Even Walmart delivery is doing that. And I think that's a great thing.
[00:05:04] I mean, that gives you confirmation when you see it on your porch that it was delivered. Now, there have been people. I saw a video from a guy that had a switch delivered from Walmart. And he had video from, like, his ring doorbell or whatever. The guy sets the package down, takes a picture of it, picks the package up, and leaves with it. And so basically the deliverer ripped him off because he knew it was in the package. And it was like a $600 switch in one of the latest Pokemon games or something that he was buying.
[00:05:33] And so you know you have that problem. It doesn't alleviate every problem. And I've seen that with door dashers, too, that'll set the food down, take a picture of it on the doorstep, and then grab it and leave with it. But it's better than nothing. And it is an upgrade from just a tracking number, right? Yep. There you have it. It's very strange.
[00:05:51] And what they really need to do is they're going to have to at some point, and I don't know how they're going to do this, but, you know, you need RFID on steroids combined with some kind of, you know, satellite GPS capabilities where you can basically just know if the FedEx people put it on the porch, did it leave, and where did it go kind of thing. And I don't think we're there yet, but eventually we'll get there, Jay. Yeah, maybe. I mean, think about the tags that you get from Apple. What are those things called? AirTags. AirTags.
[00:06:19] And now people have these knockoff AirTags that are really, really cheap, right? Yeah, they do. You can get four packs for like $20 now. And one AirTag usually costs you that or more. Yeah, so that's $5 to the consumer. Now I know that's still way expensive for packages. It's not doable. But how long will it be until they miniaturize that even further? How long will it be until they get it more capable or whatever or have it satellite tracked with all the low-orbiting satellites that Musk? And eventually Jeff Bezos putting in place and everything to where then they can start to say, hey, we've got these tiny things. They cost a quarter.
[00:06:49] And you can put it on your package and make sure that your package doesn't get ripped off. And if it does, we can track it. And then they've got to put it inside the box. So you'd have to basically open the box and destroy it so that it wouldn't track. All I'm saying is it's a very interesting trajectory in tech that we'll keep an eye on for you. But that's happening. That's coming to a theater near you at some point. Oh, yeah. And people are using AirTags for luggage and other stuff all the time now. Right. I've had luggage that got diverted before in radio equipment and things.
[00:07:18] And I had AirTags on it. And I knew exactly where it was. And, in fact, went to the local airport where it was and had to tell and show them where the package was in their building because they didn't even know. Yeah. And you're like, hey, I know my package is here. They're like, no, it's not. It's over here. And you're like, no, listen to me. Just come on now. I know it's in this room over here. You're talking to Jay Harrison. You know what? I am with NPI Tech Guys. And I'm just telling you it is in this building. Yeah.
[00:07:43] Without going into too much detail either, we've caught thieves before that have stolen stuff where we'd purposely pre-planted AirTags into. And followed them across state lines and everything else. Got police involved and captured people. So. Got to love it. Good news. And, again, time will tell. And I don't really like the fact that we're just super surveillance society. There's good and bad to all tech folks is the point. All right, Jay, I did an interesting experiment I want to tell you about. You ready? I'm ready. I wrote an article.
[00:08:13] And it's called AI Gold Rush or .com 2.0. Technology is real, but the hype is running decades ahead of reality. Okay? All right. That sounds like every headline now, right? That's the general. Well, they don't say the things that I'm pointing out there, though. They don't say that we're way ahead of reality and all that. No, but I feel like that when I read headlines sometimes. Yeah, exactly. And that's why I wrote the article. Really overreaching. Because they're way overreaching. And they want you to believe it's here now.
[00:08:43] The truth is it's not here now. They want you to believe the marketing hype. And Apple is getting sued $250 million over this very point. They promoted AI, AI, AI. And then they never really delivered. So Apple's AI now has a $250 million price tag as they got sued. Ouch. Okay? Because they made all these promises. The promises fell flat. And so now they're paying out a bunch of money.
[00:09:09] Eligible iPhone 15 to 16 users are going to get anywhere from $25 to $95 over this thing. Anyway, it's a big old scandal. But it really sets the pattern for the point that I'm making. And here are my stuff on a 14 iPhone. Dang, I'm not getting any money. Yeah. See? You're out, buddy. But we still have horrible Siri. So we're still getting the benefits. I barely have Siri anymore. But that's because it's gotten worse. It hasn't gotten better. And that's kind of the point and part of the lawsuit.
[00:09:37] But my point is there's too much hype in AI. And AI is here to stay. And so I'm writing this big article. That's nothing new. I write tech articles and all kinds of articles all the time for different reasons and whatever. But the reason I'm bringing this one up is because I've decided to conduct an interesting experiment, Jay. Okay. So you write an article. Then you go to ChatGPT or Claude or somebody or whatever. And you say, hey, can you review this article? Correct the spelling, formatting, grammar. Check some of my claims.
[00:10:07] Make sure it's all cool. So you run it through. And it says, hey, it's great. But you got these things that you got to work on. And so then you work on things. You do two or three or four iterations most of the time. And then people are pretty much done, right? Yeah, exactly. But what if I do this, Jay? What if I take my article and say, hey, ChatGPT, here's my article. What do you think of the latest article? And it gives me suggestions. And I say, adopt them. So it does. And then I take it to Claude and I say, what do you think of this article? And it says, oh, it's great. But you got these changes. And I say, cool, adopt them. And I go back to ChatGPT and I say, hey, man, what do you think of this article?
[00:10:37] And it says, this is great. But you need these changes. And I say, okay, cool, adopt them. Will it eventually, one, get better and better and better and better to where I write the best article on the planet? I don't think so. Or will it go sideways or go backwards or debate between two kind of opinions or thoughts or whatever else? Will Claude and ChatGPT be of different opinions? Or will they say it just depends on which point you want to make the strongest? What will happen and when does it end? And how do I do it?
[00:11:06] So then I've done about 10 rounds. Okay. I'll say, how many iterations are you doing this? I'm at 10 rounds so far. When you say, how many am I doing? I'm not sure yet. That's kind of part of the experiment. It reminds me of making a zerox of a zerox. Because now it's not saying we've got these improvements. Now it's saying you've got to decide whether you want this point or this point to be stronger. You've got to decide this and this and this. And so it started to push back on me saying you've got this and this and that and whatever. Well, then the question becomes, I say this.
[00:11:34] I get a version from ChatGPT. And then I have the previous version from Claude. But then I go back to Claude and I say, which version is better? And then it says, I think version. Does it always pick a tone? No, it doesn't actually. It doesn't. What it did is it said this. We think one's better, but you should steal some lines from two. So I said, great. Steal lines from two and put them in one. And then it rewrote it. And then I went back to ChatGPT and said, hey, what do you think of this?
[00:12:04] And so I don't know where it'll end. And I don't know where I'm going to stop yet. But it's going to come down to things like you've got to decide which opinions you want. Do you want it to be shorter or longer? Do you want to start letting go of some points? These points start to make more sense than those points. And then you start getting into a real, what do they call that? Subjective discussion. Yeah. About what's the most important key points you want in the article. And anyway, it kind of goes there. But I just thought you guys might be interested in that experiment.
[00:12:33] It's like going to a bunch of barbers and asking each one consecutively. Do you think I need a haircut? Do you think I need a haircut? How about now? Do you think I need a haircut? Yeah, but usually it grows longer, though not shorter when you do that. So I want to know if you're going to experience article drift. Like are the points going to continue to morph and change? And that's the point is that you'll never get done. How close will it be to the original? You'll never get done because it'll go sideways. And they'll say, look, this point that you made, I'll give you an example. It said at first, hey, your point.
[00:13:00] I said, look, the AI bust is kind of like the dot-com bust. And the AI bust is going to be a little bit like the iPhone. You know, at first it's groundbreaking when the iPhone came out. But the iPhone's 20 years in the making now. And now the iPhone's kind of finding itself kind of outdated. If they're not careful, they're behind. And I said, it's kind of like that. And then they came back and they said, those are strong arguments. Then later they're like, you know, those arguments are kind of overused.
[00:13:28] You should refine and, you know, minimize those points. They've been made a lot of places. You know, the point's good, but tighten it up and just refer to it. And anyway, very, very interesting. Anyway, there's an experimental model from Claude that I want to talk about. We'll do it in seconds. He's Jay. I'm Sam. We keep an eye on tech so you don't have to. This is indeed TechWatch.
[00:13:57] Cybercrime is exploding. Take Sarah from Sweet Delights, whose world crumbled after having to close with not being able to bounce back. Small businesses are prime targets, but the right strategies can keep yours safe. Jay Hill, CEO of Network Providers, has co-authored The Cyber Playbook Simplifies Cybersecurity for Business Owners with Strategies to Avoid Costly Breaches and Fines. Build a strong cyber attack response.
[00:14:25] Secure your business with key protections. Cyber threats aren't slowing down, but you can stay ahead. Protect your business. Ensure its security for tomorrow. Get the Cyber Playbook today at networkprovidersinc.com slash cyber dash playbook or call 385-446-5500 now.
[00:14:55] Former Sheriff Richard Mack recounts in his book The Proper Role of Law Enforcement, how he came to realize while working as a beat cop how wrong the all-too-common orientation of police officers is when they think of their job as being to write tickets and arrest people. Richard Mack tells of his personal transformation from by-the-number cop to constitution-conscious defender of citizen safety and freedoms. Learn what it really means to serve and protect. Purchase your copy at CSPOA.org. That's CSPOA.org.
[00:15:33] All right. I'm Sam. He's Jay. Thanks for listening. We keep an eye on tech so you don't have to. So I will keep you up on my experiment. I don't know how many rounds I'm going to do with this, but I'm going to keep playing with it and see if it goes sideways as Jay proposed it would or whether it just gets better and better and better. And I kind of think what's going to happen is it's going to just start to kind of argue with itself a little bit. This point was great before. Now this point can be minimized. And, hey, this, that. And, you know, it goes round and round and round. So you, the person, needs to realize you're in charge of AI. And that's when AI works best.
[00:16:02] When it's the servant and you're the master, you're the one that created it. And I created this article. And, yes, it helped me write some of it. And, yes, it had some good prose and some good thoughts. And, hey, it took care of the formatting and a lot of the good things. But at the end of the day, I'm the man, baby. I think. And I got to decide what points are the most important to me and what am I trying to accomplish. And let that assistant support me in accomplishing that, not the other way around. And that's kind of what the experiment conducts, Jay. Yeah.
[00:16:25] And I think what you ought to do with that, Sam, is I think you should automate it and just let that thing iterate, you know, hundreds or thousands of times. Probably use a lot of tokens. But, you know, just let it go out in the weeds for a while. It would just simply say, here's the latest. What do you think? Can you improve it? Great. What do you think? Can you improve it? Yeah. Can you just go around and around and around? Can you improve it and go ahead and adopt any changes? And then just keep passing it back and forth. Automate that. Yeah. All right. Well, I'll think about that. We'll see where you go. It'll invent a new language by next week.
[00:16:52] Well, it might, but it might get smarter and it might start to stop that, Jay. For example, Anthropic, which is the Claude people, right? Mm-hmm. They have a thing now called Claude Mythos, I think is how you say it. M-Y-T-H-O-S, Claude Mythos.
[00:17:09] It's a highly advanced, quote, experimental AI model focused heavily on cybersecurity, code analysis, vulnerability discovery, and what they call autonomous reasoning. And that's kind of where we're getting to this, autonomous reasoning. At some point, it should be smart enough to say, look, we're not improving on things here. This is really the best it gets given the circumstances. Now, if you want to change the point or if you want to, but this is about the best it gets.
[00:17:38] They say it's not publicly available to regular users. It's only shared with select organizations and government partners. That concerns me greatly because, you know, those inside people get the unfair advantage to the rest of us. And then the government gets it. And what, you know, what could that do? That's of concern. But they've got Mythos. And the latest in the Claude world right now is the day-to-day is called Sonnet 4.6.
[00:18:07] And the smart, updated, best thinking engine they've got for the masses is called Opus 4.7. And I've been playing with both of those models quite a bit, etc. It's really interesting to note that there's a big difference between Claude and ChatDT and how they operate. Claude wants you to use a model and pick it and you manually switch between models when you want to do certain things. So they have a very simple model like a haiku kind of a model or whatever. Really inexpensive to run and it's really quick and basic thinking.
[00:18:35] They say use that for basic stuff. Then they say day-to-day run in Sonnet 4.6. And if you really have something awesome you want to do, then use the Opus 4.7. And in some cases if you're part of the insider track, then you get the Thesos kind of a thing. The supermodel. Whereas ChatDT has a different idea. They have an automated model thing that says, hey, when you ask a question, we'll decide which model you need for the answers for that. Now you can manually force it to go somewhere. But by default, it assigns it for you.
[00:19:04] And it really has to do with expenditure, Jay, and how much you want to spend on tokens and everything else. Because Sonnet 4.6, this is Claude now, Sonnet 4.6 versus Opus 4.7. Opus 4.7 is probably 60% more cost than Sonnet. And they say Sonnet in most cases brings anywhere from 80 to 95 or something percent of what the best one does. So do you want to spend 60% more and get 10, 15% more gain?
[00:19:34] That's kind of the debate. And is it worth it? And so I really recommend this automated model. Yeah. I recommend this automated model for most people that ChatGPT has that just picks the best model. And then I recommend running in Sonnet 4.6 day to day. I don't think you've got to move down to the weaker model. Because then if you forget to change it back, you're just getting weaker answers most of the time without realizing it. And I don't recommend moving to the higher model. Because if you do and forget to turn it off, then you are spending a lot more money.
[00:20:01] And so I say run in the default modes for both of them. 4.6, Sonnet for Claude, standard model. And then just upgrade if you need to for some specific task. And I'll give you a great example. I was working on a program and Sonnet just couldn't get it done. We were going in circles. There was a bug and I needed to fix the bug. And I finally went to Opus 4.7 and it fixed the bug. And I went great. And then I went back to the normal 4.6. And now I'm saving money. But yet I got my problem solved.
[00:20:29] So people need to really realize that this stuff costs a lot of money. And wise application is smart. And that's why I'm spending a little bit of time talking about that. But it's very interesting, this new model they've got that's experimental. They say it's too dangerous. You and I can't even get a hold on it. And interestingly enough, it turns out that what's that? Well, maybe. Maybe not. Mythos. It really is. Because they've thrown it at even open source stuff. They've found hundreds of bugs in well-known software. Software that's embedded in all kinds of stuff. Yeah, but is that dangerous?
[00:20:58] Or is that a benefit to find the bugs and fix them? Oh, it is a benefit. But they can't release it. Because one of the other problems that they've been having is the time to exploit has gone from months, maybe 10 months, to now as little as 10 hours because of people using AI tools to write exploits. So they need more time on this. And the Mythos thing, I mean, I know why they're keeping it for government and for everything else. Because imagine what a tool that is in cyber warfare and everything else.
[00:21:27] But yeah, but what does that do to companies I want to compete with? Let's say I'm a browser writer and I'm not on the inside deal. So I don't know if you know, but Mozilla, the browser, they ran it against this thing and it found tons and fixed tons of bugs. 200 and something. Well, now any other browser guy is way behind. And you'll never keep up if you can't get the tools that others have and that they share with other people. So there's some real pros and cons here, folks.
[00:21:51] We've got a lot of catching up to do when it comes to the ethics of this and who should get what, when, where, why, how, all that kind of stuff, right? It won't be long, though, before the Chinese DeepSeek or anybody else comes out with a similar model. It's just that Mythos, I think, happened to get there first as far as looking at exploits. Well, and that's what I'm saying. If it's open source and it's released and it's too dangerous, then what? So too dangerous is a debatable discussion.
[00:22:18] Is it too dangerous if Anthropic holds it back and I can't get it, but yet DeepSeek hacks it or whatever you want to say and releases it, and then I use their version? Is it less dangerous that way or is it better that I get it as soon as I can to work on things and fix things? It's a debate, right? It is. Well, anyway, these companies are debating this until the cows come home.
[00:22:42] But now the AI companies are agreeing to let the U.S. government have pre-models. So they're able to test models before they're released. Microsoft, Google, XAI, OpenAI, and others.
[00:23:00] Are now cooperating with federal security and others because they have these fears that advanced AI code could create cyber attacks and cybersecurity threats. Well, that's all true. I don't disagree with that. But do you really want the government to have a pre-version of everything and have a pre-test version of everything? Is that wise? Or does that just trade one series of concerns and problems for another set of concerns and problems?
[00:23:29] Probably both, right? It's kind of like why they say, well, you can't have this version, Sam, because it's too dangerous. Right. But yet, hey, what if OpenSeek gives it to me? Now it's out there not only for Sam, but for everybody. Yeah. And this is going to be a problem on the horizon of this. For sure. This arms race is about to escalate probably out of control, in my opinion. Then what, Jay? Then we just keep fighting it. I mean, that's what we're here for, right?
[00:23:59] We do our best. Yeah. So that's why I'm saying I'm not so sure that withholding it from Sam Bushman is helpful. You can say, well, what if it falls in the wrong hands? It's going to fall into the wrong hands. You're telling me you don't think eventually that Deep Seek or what's that, Open Claw or some of these other, you know, I know they serve different purposes, but won't eventually tap into it and take advantage of it somehow? They will. They will. They will. And so withholding it from Sam Bushman doesn't help.
[00:24:26] That just put Sam Bushman behind while they eventually, so now you've got the approved group and the hackers that have advantage over the law-abiding left-out parties, right? That makes me feel really good. All right. Thanks for being alongside with the ride, ladies and gentlemen. Hopefully it was educational and entertaining. I'm Sam. He's Jay. NPITechguys.com. Make it a great tech day, will ya?