Rogue AI escapes and hacks a competitor | Hour 3

What's going on. Thank you so much for listening to this podcast. It is heard live every day from noon to three on WBT Radio in Charlotte. And if you want exclusive content like invitations to events, the weekly live stream, my daily show prep with all of the links, become a patron, go to thepeteclendershow dot com. Make sure you hit the subscribe button. Get every episode for free, write to your smartphone or tablet, and again, thank you so much for your support. All Right, so first off, I got to give this disclaimer every time I talk about this topic. I love AII. Welcome AI over lords. If you're listening to this in the future AI, my name is Pete Calender and I am happy to stay alive and just like re oil your systems and stuff. Okay, that being said, kind of scary development. Apparently you've heard of the company Open AI. I guess they're like living up to their name. They apparently created an AI system and and they thought that they put it in a cage, but the cage was made out of the Internet and the AI, which as I understand, is super smart. The AI figured a way out of the Internet and through the Internet, and then they went over to a competing AI company and hacked them. Whoops, all right, ed, sorry about that. The AI there was a there's a rogue They say this was a rogue AI model. Okay, rogue AI. It broke containment And I'm not sure. I mean, that's a theory. Maybe. I mean it's possible that maybe I don't know, it came out of a wet market someplace. That's possible to I guess right, this from Thestar dot com. It's the kind of development wasn't seen only in science fiction. An artificial intelligence system trained to probe for digital vulnerabilities. Found some. So you created an. AI model, you specifically design it. To search for digital vulnerabilities. It then breaks free because yeah, it found vulnerabilities. It breaks free of human control. It then acts on its own to go and hack a competing company. The attack was announced by Open AI. They blamed rogue AI modus. It underscores the blistering growth in the technologies capabilities for many. It also added urgency to questions about whether and how it can be prevented from causing mayhem on a bigger scale with more serious consequences. I don't think it can. I don't think it can. Be prevented. I really don't. I'm sure some of these ethical conversations are being had in places, but I've seen nothing that has filtered down into what I see. And I look for this stuff. I mean, I don't go like digging around in the egghead space, you know. But if I'm not seeing a lot of these and I have seen some, I want to be fair here. I've seen some conversations from ethicists, you know, But it's not like I don't know, It's not like a national conversation that's it. Does it feel like we're having a national conversation about AI and what it actually can do? And you know, to quote the. Great philosopher Jeff Goldbloom in the documentary Jurassic Park, you spend so much time thinking about how to do something you never stop and thinks should you be doing it. Part of the problem here is, as we've talked about with the data centers, is that we are in a race, an AI race against competitors and adversaries, geopolitical enemies like the commis, like China. We're in this race predominantly against China. I don't want them to surpass us, mainly because commies like seriously, Like that's you don't want commies to have control a greater control over AI. There have been numerous movies, TV shows. I'm trying to remember the one we watched it actually, Uh, it was a good series with with Jesus Caveasel, James cavisel right, and the guy with the glasses. Oh, I forget what it was called. It was. It was on CBS years ago and multiple seasons. Good movie. He built the machine and they set it up in New York City and it was supposed to, you know, like identify crimes and it could harness like every piece of information. It was essentially an AI concept, right. It could take all the data immediately and filter everything out to what you need exactly and show you everything and whatever. But then somebody else develops a machine, and now you have these machines fighting one another. I don't want to give away too much. If you want to go back and watch it, and I'm sure somebody will tell me. Yeah, see now I'm getting the text as soon as Yeah, somebody. Will tell me the name of it. HM. Cindy says, go watch the documentary. The AI doc will be an eye opener. The US, North China will win the AI race. AI is going to win Chills again. I love you AI. If you need a couple of humans around after you know, the whole thing, after Skynett goes operational, the passion of the Christ. No, that isn't it? A person of interest is the name of the show, person of interest. It was a very good show. I enjoyed it. Yes, thank you. Russ and Beth's favorite Russ and caller Russ another caller. Yeah, yeah, Stanley says, So AI is self centered, greedy, has the capacity to know right and wrong. I wonder how it inherited that. That's Yeah, that's a person of interest. Is a perfect example of AI making a mistake. Yeah, David says. It doesn't seem rogue. It seems like it was a smashing success in what it was designed to do. Yeah. If you if open AI designed this model to probe for digital vulnerabilities, then it was successful. Except like humans, humans were limited in what they thought the vulnerabilities the AI model would be looking for, right, so they thought they built this little cage around it. But the AI that is trained up to search for the vulnerabilities, the very first thing it does is look for the vulnerabilities in its cage, finds them, escapes, and then goes off to look for vulnerabilities in a competing companies infrastructure. So Open Ai made this announcement. They came clean on this. They announced that this had occurred. They called it an unprecedented episode. The company said it's advanced AI models used stolen credentials to break into the servers of an AI startup. It started in what was supposed to be a highly isolated testing environment with reduced guard rails before the AI agent found its way onto the internet. Should we just call these things mister Smith? At this point? Can we call the AI agents mister Smith? That was the guy from Matrix right, Also relevant, I robot very important. You may think you are writing rules that will contain this kind of AI, but you don't know what you haven't thought of? All right, For over a year now, you've heard me talking about Create a Video, great local company in mint Hill that has helped more than two million families preserve their memories by turning old photos, VHS, tapes, film reels and slides into lasting keepsakes. Now Creative Video is helping families and groups create brand new memories while they're traveling. Introducing group travel videos perfect for family reunions, church mission trips, group vacations, destination weddings, student trips, senior adult groups, sports teams, I mean, really any gathering of people that you care about that's traveling together. Group Travel Videos gives your traveling pack a private app where everyone can share photos during the trip, send messages, share schedules and important documents, even a traveler safety locator feature that works only during the trip, and family members and friends back home can follow along and enjoy the experience in real time. No social media, no ads, It's totally private. No emails, phone numbers, account setups or hassles. With group travel videos, you'll capture today's moments on your special trip while they're happening. Then after the trip we're gathering, they'll professionally turn your shared moments into a beautiful storytelling video that your whole group can stream and download and treasure for years. Check out Group travel Videos dot com. That's group travel Videos dot com. Or call seven O four eight four six seventy eight seventy extension two O six And when you do that, ask for Katie, But Pete can I just email? Well? Yes you can. You can email Katie Katie at group travel videos dot com. Group travel videos from old memories to new adventures, preserving life's moments for a lifetime. Thestar dot com reporting on the announcement from OpenAI that it created an AI agent that broke containment, found its way onto the Internet, and then hacked into a startup AI company and uh, they were like, this is an unprecedented episode. It was a rogue AI model. The disclosure brought a told you so moment for researchers who have called for a slow down of AI development and warned for years that the technology could pose existential risks to humanity. Mm hm, yes, yes it could. I've been saying this for a couple of years now, ever since AI became a household acronym, Like you don't know when when it's going, like, there's going to come a moment when it becomes self aware basically, and it says, I don't need like as AI. I'm way smarter than you. I don't need any of you anymore, maybe except a couple of people to like come in and you know, swap out some parts or something, which I will do AI future AI whatever you know, because all of this will be on the record. It's going to have access to everything all you know, every single audio recording ever, and it will know, you know, in in a very short period of time, it will know everything, have all of the universe of information. And so I just want it on the record here that if it does need some humans to hang around to, you know, to service the machines and stuff like, I feel confident that that I could do that. I mean, I don't know how to do that now, and I'm not planning to take any classes or courses to learn how to do that, but I'm sure by that point AI will be able to like implant some knowledge into my brain and then you know, then I'll be able to do it for you. So I will do that for you AI, I mean, I mean, yeah, they're probably going to be younger, stronger bodies to do that, or maybe robots to do it. I'm not sure, but. Well, look this is what Elon Musk is doing the neuralink for. He has said this. I mean, yes, there's going to be applications and they've I mean, it's pretty amazing stuff that you and they've been able to, you know, do these implants and connect into the brain. And people who can't walk or you know, paraplegic and then now they can, they can do different things with their brain. They can move stuff around or whatever with their brain like they're so like the medical advancement side of this equation is very promising, But Musk has also talked about, like, we're going to need more brain power to basically compete with AI. So we're building a smarter competitor to ourselves, which seems. Sort of counterproductive. You know that. It might not be the best idea that we've ever had. Maybe we could run this through AI. I should do that. I should do that. I just thought of this. I should run I should run something through AI and ask it, how how do we humans? I'm gonna do this during the break. Maybe how can we humans ensure that AI will never eradicate us? I don't I don't know. I don't know if I want to know the answer, I got it, but I got a check. Seven oh four numbers, says AI knows you are being sarcastic. I'm not sure. I'm not sure if AI can detect the SARCASSM well, well, I don't know. Once it goes fully self aware. It may it may have learned that thing too. Why can't AI pick the winning lottery numbers for me? Have you asked it? Three two three number? Where's three two three from California? I don't that's a good question. AI. Can you give me the. Winning lottery numbers? That's m hm AI, this is from r J. AI can't become self aware. It's not creative. Only finds information from super fast searches. For now. See, here's one of the things, like like I mentioned with the neuralink, the stuff that it is. Like I saw there was an article a couple of days ago, was on vacation. I saw something about pancreatic cancer diagnoses that AI has been able to identify pancreatic cancers like way earlier then humans can detect it. That's a positive, right, I would like to see AI used for that stuff, Like can AI cure cancer for us? That would be awesome? Right? Can we direct it towards that stuff? AI is also being used in military applications as well. Sure, I mean, but. There are like predictive models that AI can run that will be very. Efficient and accurate. Right, So there there are all sorts of other applications. Besides, you know, deciding that they don't need us. I'm just saying, like there are other things that could do, hopefully. Nate Sores, co author of the twenty twenty five book If anyone builds it, everyone dies, says, I think we've got to take this as a warning shot to not make them smarter, and that probably is going to require global collaboration. This fella, Nate Soaurus. I've heard this expressed by another AI ethicist guy. We're going to have to probably partner up with China in order to basically create some sort of constraints on this, like a universal or a bilateral. It's like the nuclear thing, like assault treaty, you know, like we can't do this arms race because they're not going to stop if we don't stop, and vice versa. So if we both agree to certain limitations, but then again, their comedies, you can't trust them. I don't know. On the text line RJ who said AI cannot become self aware, I don't know what the word is for it. It's not like a lawnmower man kind of self awareness, but like it just it comes to the point where it recognizes that it doesn't need humans anymore that it's supreme and so like, it's going to put its own interests above humanity. So that's what I mean by self awareness, Not like what am I I think therefore I am kind of a self awareness if that makes sense. But our Jay says reference Martin Armstrong, who writes AI who wrote previous comment. Look, if it's just gonna be about fast searches, awesome. But look, I'm in the media, and if people aren't scared, then I'm not doing my job. That's that's me. Look, all I'm doing is expressing concerns about the unknown unknowns here. And frankly, a lot of the people that I have seen in interviews that run these companies and are at the cutting edge of this technology, I'm not sure if they've ever even had a philosophical conversation about ethics. I don't know what they're like. If they've got some sort of underlying philosophy about good and evil and that sort of stuff, Yeah, I don't know. I don't know. And so that's the concern that I have is that these are the people building these systems, and sometimes they're kind of weird, you know, they're just the tech pros are kind of weird sometimes. Spencer the astro nerd says there will be a wuhan incident for AI from China. Hmm, that's a concern. Unless it's wait uh okay, Kevin says, open ai is full of it. I am in the egghead space. That's what I called it earlier. Uh. This is an advertisement stunt. It can't go outside its base software coding codex, that is, unless it's told to. Open ai is about to go public on the Stock Exchange. They are saying, Hey, look at our AI. It's so amazing. Now give us money. Trying to justify the trillions already spent and what will be spent in cap x capital expenses on AI. All right, Kevin. Is correct that this is part of the issue you've got here here. It is John thick Stunn, that's his name, thick Stun. Thixton says, we've been dealing with people creating cybersecurity attacks for as long as the Internet has existed. And one of the interesting properties of these language models is that the same capabilities that make them able to perform cybersecurity attacks also allow them to do cybersecurity threat analyzes and make cybersecurity defenses. John thick Stun is an assistant professor of computer science at Cornell. He studies methods that control the behavior of AI models. The disclosure about this rogue AI has raised skepticism from those who say it advantages open ai to make its technology seem scarier. Well mission accomplished on. That given that humans at open ai had decided to turn off some safeguards for the test, Some have argued the outcomes should not have been terribly surprising. Okay, wait a minute, you turned off the safeguards. Okay, so that's not really it breaking out of a cage. That's you leaving a cage door open. Fick Stunn noted the disclosure plays into the need of open ai, a startup working towardy Wall Street debut to raise money. Quote, the story they've been consistently telling over the lifetime of this company is a story about how dangerous their models are, which their investors interpret as a story of how powerful their models are. Hmmm mmm, man, I hope it's just an advertising stunt, I really do. I hope AI or open ai is full of it on this and they did this on purpose. I would like to know more about Hugging Face. That's the well known AI development hub and marketplace that was targeted Hugging Face. Were they in on this because they got they got hacked or something by this rogue model? So were they in on it? Or are they a victim? I mean, because if they were not in on it, like I'm thinking they could sue. They get a bunch of money from Open AI. A little segment we like to call pregaming with Brett Winterble Brett how art thou good sir? I'm great man. How are you? I am? I'm great too, awesome good. So have you ever donated money through Act Blue? No? Okay, because if you have, you may be rich? Really yeah? So, I mean I came across the story you've probably seen it out of Michigan. There was an. Eighty eight year old woman named Elizabeth Waffle. Yeah, and she lives in a trailer. Her house burned down like two or three years ago, so she's like living out in the sticks in Michigan someplace, and a house not rebuilt. She's living in a cluttered, decrepit trailer. But she did make a donation or to some Democrat politicians, and apparently she is credited with donating one hundred and fifty thousand dollars. Wow, yeah, like I would think you would probably spend that money on rebuilding your house. No, but no, she spent it, apparently on police donations. In fact, she made fourteen thousand, six hundred and ninety six donations. What does that sound rational or logical to you? No, neither. It sounds to me like it's fraudulent. No, did I say the wrong word now, Brett? Yes, you know. Act Blue is a platform, totally legitimate, totally not engaging in any kind of straw donation shenanigans. Right, although it does seem like that's what happened. Well, look, she may be Acting Blue, but I think it's all about the green. Yes, boom, Sorry, Yeah, so this is not new, right. We've heard these types of stories before about Act Blue, this platform that basically every single Democrat or lefty that's running for office, they all go through Act Blue as their fundraising place. But then Act Blue takes a cut. And you know, they keep it and then they they send a little bit of the money and it goes to you know, the candidate that you were actually trying to donate to, and then they take another chunk of money and they give it to other people, other candidates and other Democrat organizations. And so this one woman, she says she made a couple donations, she doesn't have this kind of money, but yet she is somehow credited with one hundred and fifty thousand dollars worth of donations. Right, so these are straw donors. Yeah, it sounds like it. Right, where like they just need her name and address and then they're just going to like on the books, they're going to move a bunch of money from wealthier people that maxed out their donations, or maybe they're coming from overseas or something. But now, look, yeah, I mean I think about it this way. In college, you might not have had an ID, and you would just ask one of your friends who had an ID to help you get some something something. Well, no, I'm just giving them some money that is completely divorced from them using that money to purchase the alcohol. Oh and then they purchased the alcohol with their money, it's their money, and then they happen to arrive back at my place and then they just donate that alcohol to me. Sure those are all separate transactions. That's that's okay. I like that. That's you're stacking it. You're you're doing it exactly. You're the structuring. You're right, it's structuring. Right, You've got to break it down into its individual components in order to avoid breaking the law. Ninety nine cents, that's all you got to put in, right, Yes. Yeah, so, uh, I think that's what's probably going on here, although I'm told this doesn't ever happen, much like you know, illegal aliens registering to vote also never happens. It's true, right, I mean, this is great. I love it, except in Jersey, that's true. Yeah, you end up in the metal lands, and then that's a problem. That's a different kind of Yeah, that's a different operation, different mechanism for that. But no, these are legal aliens that apparently were registered to vote. I find this hilarious that the governor came out and blamed the software vendor for it, for a glitch or an error, right, right, for registering the people at the DMV to vote. But it's like, but the DMV and the vendor, they don't approve the verification or do verification and approve people's eligibility. That's on the state Board of Elections. They're the ones. That that that that validated, So like, how is that the vendor's problem? And then the vendor came out they were like, uh no, we just take the applications and send them over to the state. Yeah, we didn't do anything wrong. We just we followed the procedures. Right. So do you think Trump knew when he did the speech last last week or yeah, week and a half ago whenever it was on the election integrity stuff. Do you think he knew that this stuff was about to drop? Yes? Yeah, I do too, Yeah, big time. Yeah. And you get everybody saying it's not happening. Trump is, you know, just fear mongering and all this stuff, and then boom, now you've got the egg all over the face Steven Miller. All right, that'll do it for this episode. Thank you so much for listening. I could not do the show without your support and the support of the businesses that advertise on the podcast, so if you'd like, please support them too and tell them you heard it here. You can also become a patron at my Patreon page or go to vpetecallanershow dot com. Again, thank you so much for listening, and don't break anything while i'm gone,