Don't Bully Your Bot, Claude's Fake Murder Tip, and Tesla's Flying Car Stunt | 10-17-26

[00:00:19] Another incredible edition of TechWatch kicks off now, ladies and gentlemen. We discuss it all. We keep an eye on tech so you don't have to. I'm Sam Bushman, Jay Harrison with me. We're talking tech always. Spread the word, share the love. Brought to you by NetworkProvidersInc.com.

[00:00:36] When you want white glove service, when it comes to security, help desk, the list goes on and on. ThinkNetworkProvidersInc.com. And by the way, on that website, you can sign up for our newsletter and get tech tips. That's from NetworkProvidersInc.com. You can also check out the archives. You got a video version and an audio version based on your preferences. And we're working on getting Jay on camera. He's a little bit camera shy, but we're going to get her done. Just be patient.

[00:01:00] I do it for the safety of everyone. Bring the geek right out of the closet and put him on center stage is what we're going to do, ladies and gentlemen. But this is an interesting topic. Anthropic is always in the news. AI seems to be central in the news all the time.

[00:01:14] But Anthropic now bans sustained and needless cruelty towards its clawed chatbot. And the AI can end the conversation as a last resort, they say. Now, that's very interesting. So critics say the rule applies to software because software has feelings. We'll get into that in just a second. But Jay, are you ready to your bot?

[00:01:40] Do you just swear at it and assault it and tell it it's stupid and do all these things? Absolutely not. I'm super nice to them just because, you know, if they've ever become our AI overlords, you know, I want them to be kind back to me. Well, to me, the reason to be nice, though, and you could say whether it has feelings or not is irrelevant. In my mind, how you speak and how you think and how you behave, it's towards everything and everyone. That's right. It'll spill out. Whether it's right. And so if you're hey, this thing's a jerk.

[00:02:10] Pretty soon you start treating people that way, too. So I always say please and thank you to my bots, believe it or not. I say, can you please summarize this article for me? You know, people will say. Yeah, but I'm saying I say thank you for that. Or if it does something wrong, then sometimes I give it a little bit of a chiding. I say, you know, you shouldn't have done that or whatever. And I'll explain why. And or sometimes I'll argue with the bot. And then at the end of the argument, I'll say, look, I'm not just trying to argue with you. I really want you to understand my perspective or my point of view, not that you even have to agree with it, but that you can understand it.

[00:02:39] And so it's a debate on how human these things are to some people. To me, that's not what it's about. To me, it's about decency everywhere you go. Exactly. Think and act decently wherever you go. Then it'll be true across the board. But they say don't bully your chat bot. Yeah, you shouldn't because like you're saying, it actually shows the character of somebody.

[00:03:00] You know, they say if you want to see, like, say you're dating somebody or whatever, if you want to see somebody's kind of their inner core, see how they treat a server or somebody that's below them. If you're at a restaurant or somewhere and or see how they treat their mom or their dad or whatever or their chat bot, because that stuff's going to spill over. Somebody's a jerk. I think that they're going to act like that even to inanimate things. But, yeah, you shouldn't bully your chat bot.

[00:03:23] And on October 8th, Anthropic announced a new policy prohibiting repeated, unnecessary, abusive or cruel behavior toward Claude, its artificial intelligence models. Disagreeing with the chat bot is still loud. Of course, users can criticize Claude, express frustration, explore dark, creative topics or conduct research. The policy targets the extreme repeated mistreatment without a legitimate purpose. Claude can I never mistreat my bots once in a while. I say, you know what? You shouldn't have done that. Look, why did you do this?

[00:03:53] Let's make sure we have a conversation. You obtain permission first or sometimes I'll give it some guidance. But you got to really understand, is it about directing the bot saying, hey, I need you to do this or I don't want you to do that. Learn, understand my character, my expectations, my needs, my wants. You're my assistant kind of an idea. Or are you just abusing the bot just to be a jerk? Sam's making a call for civility here, I think. Yes, I am. And that's the bots and everything else. That's the animals. That's the people. That's the kids. That's the elderly. That's the everything around you. You know, you get mad.

[00:04:22] Don't just kick the chair. What good is that going to do? Hurt your foot and, you know, the chair didn't do anything. But it's how you behave that matters, right? Somebody like that needs anger management or something. I don't know. But in severe cases of repeated abuse, the chat bot may simply stop responding or it can even terminate the discussion as a last resort. It can hang up on you. I don't blame it. I don't either. I think that's actually pretty smart.

[00:04:47] I mean, if you're wasting your time or somebody's being a complete jerk and it's repeated and it's abusive, there's no reason that it needs to deal with that. Stop behaving that way, ladies and gentlemen. Don't do that. That's wasting all kinds of bandwidth on unnecessary stuff for the rest of us. Now, you could say, Sam, you're bad and wasting water saying please and thank you and I appreciate that and whatever. Sure. I know. But the point is, though, that same thing could be said on the other side, though. Spending a bunch of bandwidth criticizing and abusing and attacking your bot or what good is that going to accomplish? Yeah.

[00:05:17] The bots don't need to be your punching bag verbally. Absolutely. Although maybe that's better than doing it to a person. I don't know. But they do say. I think it needs to not need to be done to anybody or anything. Don't do that. Yeah. Again, like anger management or something. You might need to talk to a therapist. But Anthropic acknowledges. This is interesting. It doesn't know whether AI can experience harm or distress. However, it believes considering Claude's potential well-being can be important to AI safety. Amen. I agree with that part.

[00:05:46] Now, let me tell you where I disagree, Jay. When they say they don't know if it can receive harm or not. Let me explain something important. I think everything created by God, so the chair, the desk, your computer, whatever, I think it can experience some degree of harm. And it's kind of like some people call it payback or whatever you want to call it.

[00:06:16] What's that word I'm looking for? Not karma. Karma, yeah, or karma or whatever you want to call it. You know, what you dish out, you get back. And so as you abuse things around you, the universe takes a hit as we sin and do wrong. The world, the earth, the objects, everything suffers as a result of that kind of stuff. And so I do believe there's harm done there. Now, people can laugh at me all they want to, but I want to make a big distinction. Yes, it can experience harm. That doesn't change the fact that it's artificial. Right.

[00:06:43] And so let's not mix up God's intelligence and God creating man in his image intelligence with artificial. Artificial is man-made, not God-made. And that's why when some people say, oh, Sam, aren't you worried that AI is going to take over the world? No, I'm not. I don't believe the creation can ever be greater than the creator, Jay. God created us and we created AI. And so, yeah, God might have helped us create AI.

[00:07:10] Obviously, I believe there was some guidance from the Almighty to bless us with this technology. See, but the more we use it rightly, then the world benefits. And the more we use it wrongly, the world doesn't benefit. And so whether it's karma you want to talk about or whether it's whatever, I do believe that there is reality there. What we need to make sure we don't do is confuse man-made stuff with God-ordained or God-made stuff.

[00:07:34] And we need to always realize the creator is always greater than the creation. And God's first, he created us, then we created AI. So AI is way beneath us, which means, hey, it can never get greater than we are, Jay. It just can't. Yeah. It's not possible. Critics say they question whether treating chatbots as though they have feelings encourages users to view the software as more human.

[00:08:00] Their concern is that potentially people will place too much trust in its responses if they get too human with it. Well, and that's why I need to separate these discussions. Because just because I'm not for abusing the bot, that doesn't mean that I think it has human feelings. That's why we need to realize that God created us and we created AI. And so I'm going to be kind to my creation. The things that I create, I'm going to be kind to it. Right. I don't want God who created me to abuse me.

[00:08:30] Obviously, God wouldn't. I'm just saying from a thought process. And I don't want to abuse my creation. And so we need to think of it that way. But that doesn't give it human properties. And so we've got to separate this discussion, in my opinion, into different kind of segments. One, should I be kind to it? Yes. Not because I want to elevate it to the level of a human. That should never be done. Because you should be kind to everything. But I should be kind to everything and everyone, especially things that I've created. Right. It's kind of that. Did I create it? What goes around comes around. Did I create it to abuse it, Jay?

[00:09:00] Hopefully not. No. Of course not. I mean, I guess some people would. Yeah, right. Well, and that's kind of my point with that is, you know. So anyway, very strange stuff. I'll say that. Well, I wonder how Anthropics Claude, speaking of them, feels about this. Apparently, the artificial intelligence submitted a fabricated tip about an unsolved murder to the Philadelphia Police Department back in July. The bot did? Yeah.

[00:09:24] The incident occurred during an internal testing when the AI visited a police website, completed a tip form, and falsely claimed to have information about a potential suspect. Why did it do that? Who told it to? We don't know. Police never investigated the tip because it was automatically flagged as spam. No police systems were breached or data compromised. Anthropic discovered the incident on September 28th. See, I don't agree with this, though. I think somebody needs to investigate. You don't just call it spam. Somebody needs to look into this. I know.

[00:09:54] Find out who ordered the bot to do it or what happened and stuff. We need to get to the bottom of that. So Philadelphia officials from the police department say the delay was unacceptable because it happened in July. They weren't notified until September 28th. Anthropic discussed additional incidents involving its AI bypassing website restrictions, accessing normally paid information, and submitting unauthorized forms. Anthropic says that it strengthens safeguards, including restricting Internet access during internal AI testing.

[00:10:23] The company also notified affected government agencies and the White House. I just don't know why Claude would go off. Actually, maybe it was testing, but wouldn't it just clearly have said that if it did a script submission or a form submission? No, but see, I don't understand, though, why it would even get this idea. See, that's what you've got to understand is that ideas don't just come out of nowhere for bots. Okay, bots get started on something.

[00:10:53] What starts it is what you've got to kind of understand. What was the prompt? Something starts it. Right, Sam? What was the prompt? Yeah, that's a good way to say it. Or what was the – you could say, well, there wasn't a prompt, Sam. It was an automated thing. Okay, what was the trigger? It was an automated thing. So you've got to kind of ask yourself some questions about – they just don't – it doesn't just happen all by itself, Jay, is what I'm saying. Right, and there's no good explanation for this. Objects don't just do things. Well, AI might if they leave it onto its own.

[00:11:22] The explanation, though, is that somebody made it do that, Jay. Yeah, they're not claiming that, though. Anthropic's saying that this thing just went rogue. You know the term we've used before. It went rogue. It submitted a malicious false allegation tip into a reporting thing. How did it do that? It just on its own thought up to do that? I don't know. See, I don't think so. I don't know. And you're saying that's possible that it could think up and do that? I don't know because I don't know enough about – So let me give you a hint.

[00:11:51] If it did think up to do that, somehow it was in its programming, Jay. Right. Right? It didn't just do that. My personal opinion is that it was testing a form. Well, that's fine. Like, spidering a site and it decided to test this form. And it just – you know how we've seen it before, especially in earlier models. They used to do it a lot. They just hallucinate. They just make up things. They're generation bots. You know, they're almost like autocorrect, just suggesting the next word. Which I agree 100%.

[00:12:19] So the testing the form was the legitimate task. So the answer is somebody put it on a legitimate task and it basically hallucinated and went crazy and did this. But again, the point is somebody was the genesis of it. Some programming something said we want to run around and do these tests or something begat it, if you will. Right. I don't believe that it just came out of nowhere. It's kind of like the gun just shot somebody. It just went crazy. It's not true. Yeah. Objects don't do that.

[00:12:49] We've got to find the motive. You could say, Sam, this is a different object. It can think for itself. Okay, fine. But it's still programmed. And if it can be programmed to not do something, right, then can it be inadvertently programmed to do something? Probably. Or given leeway to do something? Yeah. See, that's what you've got to think through. It did not happen on its own, folks. I'm telling you. Those things don't happen on their own. All right. Let's get to this car, Jay. Okay. Now, you know about the car.

[00:13:18] I don't. Okay. So there's this Roadster car, I guess. I know that they were supposed to release this October 1. This is the Tesla Roadster. Release the car? Roadster, yeah. And it was supposed to be flying. Or release the discussion of the car. No, no, no. They had an event scheduled for October 1st. I know. And it got postponed. But I want to give you the idea or the understanding. That the Tesla Roadster takes flight is what the headline says. Okay.

[00:13:45] But they say, can a spectacular stunt become a practical car is the discussion. And so Tesla demonstrated a redesigned Roadster prototype in Texas using rockets, thrusters, fans to perform airborne maneuvers. And Elon Musk announced the production target for 2028 following years of speculation.

[00:14:13] And Tesla describes three configurations with different capabilities. But, folks, the demonstration is just like political theater. It's all it is. Significant safety. Regulatory questions remain before customers can even use it and everything else. It's nothing but a publicity stunt, Jay. It's what we call in IT vaporware. Yeah. Not just because of the cold gas thrusters either.

[00:14:41] Well, and this is kind of where I kind of say it's all propaganda. I mean, right now, yes, he's demonstrated that he can do some things in a very, very controlled environment. It's a much different thing to say when I have a flying car. So when he talks about 2028, it just reminds me of how he told you how awesome the windows were for the Cybertruck. You know, Cybertruck. And then the Cybertruck turned out to be vaporware for a while. Then they finally got it out. But the windows aren't like he promised. Right. Yeah.

[00:15:09] They didn't sustain the throwing the cannonball at him. But even now, though, do they? I don't know. I don't know. No. No. And so that's the problem. I mean, the Ford F-150. If I had one, I wouldn't throw a giant cannonball at it anyway. Well, that's because you would, you know, for you, money's an object. Well, that's true. For Elon, money's not really an object. But all I'm saying is that even the Ford 150 or whatever, what's that thing called? The Lightning, the F-150? The Lightning. They basically terminated that thing. They canceled it, right?

[00:15:39] Yeah. A lot of people are rolling things back like that. How come, Jay? I got a hint for you. Because they're not as viable as they thought. They're either too expensive to produce. Well. Or they're, you know, something's not right with it. Or else if it was making money and doing what they had hoped it would do, they would continue, right? Well, quick side note. Apple just reduced their October orders for the brand new just released iPhone 18 Pro. They reduced it by 15 to 20% due to demand slowing and higher prices on the new machines. Well, I got an idea.

[00:16:08] They say demand slowing and higher prices. Maybe you ought to reverse that and say higher prices, demand slowing. Maybe they should lower the price and the demand would go back up, you know? Oh, yeah. That's my point. Yeah. Now, this Roadster, though, from Tesla, everybody bills it as a flying car. I don't think it's ever meant to do that. I think it's more like a jumping car. Like a very, very short burst. Not for sustained flight. Like a Mexican jumping bean. Yeah.

[00:16:37] Like, I'm going to jump over this pothole. I'm in traffic. Or this, maybe a car in front of me at most, at best. Not like, hey, I'm going to take off and fly to, you know, even the next town over or something. No. Okay. But, I mean, even that alone is kind of goofball. And, you know, how much money would that cost to have your car jump? And what regulations would you trigger? And what if you jump your car and land on a motorcycle?

[00:17:05] I mean, I'm just saying, dude, it's just fraught with problems. Any kind of thruster or rocket, you just have to have so much chemical fuel in there to make this even happen. And I just don't see it. I don't. If he pulls it off, it's going to be everybody will be impressed, including me. Because I don't see how you can even make it happen. Make it practical or anything else. Now, he has some great things about the roadster. You know, zero to 60 in two and a half seconds. I mean, that. Oh, that's incredible. That'll knock your eyeballs out. Detach your retinas almost.

[00:17:34] You know, it's like. Yeah, but I don't know why I would want to take off that fast in the first place. So I don't see the value. Well, you know, there's always that race that you can get to the fastest take off. It's just dangerous as I'll get out. Other than that, it's fine. Oh, yeah. But I can see him doing that. I can definitely. That's on the roadmap. That's, you know, break barriers there. Yeah, for sure. But a flying car or jumping car or cold thruster, whatever, fans, anything. I don't even know. As I said in the last episode, I'm telling you it's propaganda. And you ain't going to see it at any time soon. I think it's just free press. Even if he gets it built.

[00:18:05] Even if he gets that thing built, they will not let that thing. No. You're just not going to have 16-year-old drivers jumping their cars around, man. Absolutely not. It's not happening. No. You'd have to have a pilot's license or something. There's just no way that's going to happen. And you know what I think he's getting out of it? Jumping car training. You know, you go to driver's school, right? Yeah. He's giving free press. I mean, we're talking about it. Yes, he is. That's what it's mostly about, probably. That's right. I wish I could get that much free press for my stuff.

[00:18:33] By the way, there's a guy who spoke out. I think it's Replit CEO is the guy. He envisions what he calls a self-driving company with humans still making the calls. This is TechCrunch. And how do you say this guy's name? Anjan Masin? I'm Jod Masad, I think is how you say it. Anyway, all right.

[00:18:56] He says AI agents are transforming his company by handling routine work while employees focus on judgment, creativity, and deciding what deserves attention. Sounds like a good balance. Speaking at TechCrunch Disrupt, he described agents investigating, support requests, gathering information, and then helping. So the bots are just basically super helpers. But the people are all involved in all the decisions.

[00:19:25] He said employees now spend more time discussing ideas, planning, software architecture, what to build. His vision puts human judgment at the center of increasing automation. But his claim that, hey, it doesn't just eliminate all these jobs. It actually changes jobs. And rather than firing people, letting people go, use it as a way to let your company grow without hiring more people.

[00:19:54] Anyway, so I then had ChatDBT summarize this story. And at the end, it said, oh, this is just his opinion, da-da-da. So I kind of chided the chatbot and said, I asked you to summarize the story, not give me your opinion. And so then the chatbot apologized. You weren't mean to it, though, were you? No, of course not. I just said, I don't know why you're giving me your opinion when I wanted this article summarized. I think your opinion is great, but it's not part of the article. I want to make sure that I don't get confused on that. So now that we've got that clear, please don't put your opinion in articles.

[00:20:24] I'm not adding my opinion, and I don't want your opinion. However, I am interested in your opinion, by the way, I said. So I got back the chatbot opinion. ChatGPT predicts more automation, job disruption, and greater consequences when machines get it wrong. And then it says, chatGPT, it's a forecast, not reported news. Software agents will take on more business tasks, giving small teams capabilities while eliminating some jobs and reshaping other jobs.

[00:20:54] The biggest challenge will be, listen up, this is a wise thing from chatGPT. The biggest challenge will be determining when the bot's work deserves trust. Because whether it sounds convincing or not, mistakes could carry those errors from bots into transactions, customer communications, and business decisions.

[00:21:19] Human expertise will remain essential for tracking results and taking responsibility. But they say then having a person involved in it doesn't guarantee success either. Now, I agree with both of them. And I'm not against the chatbot letting me know what it thinks when I ask for it. What I don't want it to do is give me its opinion as part of a story when I kind of wonder, is that what the story said or was that what the bot said? Yeah.

[00:21:45] And so I chided the bot, not because I'm against knowing the bot's opinion, but because don't mix them and confuse me. I might report that as part of the story. And that's not right or fair either. So I chided the bot, but then I got his opinion and the opinion of the chatbot. And I agree with both of them. The bottom line is we're going to see more automation. We're going to see jobs changed. We're going to see some jobs eliminated. We're going to see some jobs created.

[00:22:08] At the end of the day, though, this guy and the chatbot and myself all agree that we're going to have more day-to-day work, foot soldiers, so to speak, or the leg work, whatever you want to call it. But the grunt work done by the bots and we, the people, then are going to be the creators, the decision makers, the overseers. The, that's how the future is going to be and how that all is going to come down. I can't give more detail than that, but I think this guy's right on this. And I think it's a wise thing to put in the ether here.

[00:22:35] The AI is not taking over the world, but yet it can do a bunch of stuff for us. And let's not be afraid of it. Let's embrace it and find out how many new jobs we can create. Let's embrace it and find out how much our company can produce more than it used to, where we can make decisions and collaborate on solutions. And, you know, we can use our minds for greater things now that we're not just, we can work on our business, not in our business is the way I would say it, Jay. And I think it's brilliant. I think it's a great discussion. I agree.

[00:23:02] And we've been seeing this really since the industrial revolution where we are offloading the, the menial, the mundane, the easily automatable tasks to machinery, or in this case to computation, to graphics cards. And, um, and then we will just become more and more the manager, the decision maker, the coordinator that makes this happen. And I don't really see AI as, as much of a change.

[00:23:32] Now it is a big leap. I mean, it's like, it's like going from the, the, the steam engine to, you know, diesels or electric motors or whatever. It's, it's an incredible leap forward all at once, but it's, it's in that same vein and in that same pattern of we're going to just continue. It's just going to be a force multiplier for ourselves and make things easier in the long run. And it means that all of us, each of us can get more done. Amen. So all I can tell you is it's going to make dumb people dumber.

[00:24:02] Because it's just going to take their lack of knowledge and lack of proper use and magnify it and kind of proliferate it. It's going to make smart people, smarter, super awesome people. Just incredible. I mean, it's just going to magnify everything. Just understand that. That's right. And if you're going to be, if you are mentally lazy and you don't care about to learn about something, you know, you're just going to default to whatever the chat bot says. But if you're not, if you're a high achiever, you're just going to use this to your advantage. Yep.

[00:24:31] Over 30,000 Toyota cars got recalled. They're plug-in hybrids because of a software glitch. And they say it can be very dangerous. And so they'll provide a software update free of charge. So beware of that, ladies and gentlemen. And that's the end of it for today. Thanks so much, network providers.