MATT ROSEN 8/3/2026 THE SIMON CONWAY SHOW

I pledge allegiance to the flag of the United States of America and to the Republic for which it stands, one nation under God. Top stories and the analysis you trust, Simon Conway, on your radio and the Iheart Radio app. I love this American. All across the great state of Iowa, where it is the Simon Conway Show 607 on this Monday, August the 3rd. Hey, before we do anything else, how about you win yourself $1,000? Gas prices. These grocery prices are the cost of living. Inflation. The cost of everything going up. Escape. Get away. Always. Your inflation vacation is booked with a thousand bucks. Chances to win every weekday. Yes, and you know how this works, right? We take a word, we send it to the website of whatever radio station we're listening to. We sit back, we wait for the phone drink. If the phone rings, well, we pick it up. That's how it works. And if it's us on the end of the phone, you won a thousand dollars. Right now, I'll take the word dollar. D-O-L-L-A-R dollar. Take that to the website. Do it right now. And good luck, my friends. Okay. We talked to you about Sam Altman's Open AI and how its machine went off on its own, figured out how to get on the internet, which it wasn't supposed to do, and then hacked another company. Now, he is calling for safeguards, which is weird. And there's people in... positions of power that are saying we need more explanations. Well, joining us is somebody who loves AI, loves AI. He is an AI expert. We have had him on a bunch of times before. He knows I'm a skeptic because I have seen the documentaries called Terminator. And it seems my predictions might be on the road to truth here, Matt Rosen. Hey, Simon, how you doing? Thanks for having me on again. I'm good. How worried should we be? I think worry is the wrong word. I think it's more, you know, a little bit of healthy concern. You know, as we've talked about many times over these years, these engines are getting more and more powerful. And one of the main things that they're being used for is cybersecurity, but also see if they could break someone down someone's cyber defenses. And, you know, what happened, I guess it was a week or two ago was opening I was testing some new models and they said, hey, go, you know, go attack. and go see if you can find your way in. And it did. It launched like over 17,000 attacks so much so, hugging face, which was the company they publicly announced they went after, couldn't defend the attack, actually, using Chad CBT or Anthropics models. And so I think that's why you see Sam came out and said, hey, the thing is moving a little bit fast, and it was...

you know, successful in breaking down their defenses. But some of the things that broke down were things like weak passwords and things that were open to the internet. So it did exactly what is told to do. It wasn't like... Well, but no, but the point is... The point is it wasn't told to do that. It was not supposed to be on the internet. It figured out how to join the internet. And OpenAI said this was an unprecedented cybersecurity issue because it did this stuff on its own. It had a goal of hacking other systems, and it found ways through the Internet to get there. So if you actually go into the accident report, it was given a goal, and it accomplished its goal, but it accomplished its goal without guardrails that should have been in place. And now this has happened multiple times again, right? It happened with Anthropics the other day. It hacked three different organization systems. Those have not been made public. But again, it was doing what it was told to do. It just found unconventional ways with a level of speed never before seen, which that is cause for concern. You know, they let these things out without guardrails in a sandbox thinking, hey, it's not going to figure it out, but it did. Because it was given that goal. And were the good guys? Yeah, we're the good guys. And that's ultimately why we're taking these and trying to break in so that if bad guys ever get a hold of technology similar to this, we can stop it. Don't they already have it? Didn't Sam Altman publish his code originally? You may have. That's probably an older version, but no, the most latest and greatest frontier models are still locked behind their walls. I mean, the models that you go and see on chat sheet, BT, or Anthropic, those are probably two or three versions old. They're working on things that we haven't yet seen that haven't been released yet. And I think that's where you see Sam and other people in this open letter that was published saying, hey, we might want to pace. how we're releasing this. And what's interesting is these are the very same people that are creating this technology asking themselves to slow down. How confident are we that the Chinese are not there, that the North Koreans are not there? How confident are we? I think it's tough to say. I mean, Chinese have been releasing what you hear called these open source or open weight models. They're not that far off. You know, if you look at them in the benchmarks, they'll probably call it six months behind, on average, but they're giving these away for free. Anyone, you just go to the internet and download them. Can you see to or deep seek? Call me a skeptic. I'm not downloading the Chinese AI. I would not. But you can run them disconnected from the internet, but they might find the internet. Well, apparently that's now a thing. Is AI now smarter than humanity? Not yet, but it's headed in that direction. It is getting quite smart. How long?

That's a great question. That's one that a lot of people have predicted could be sometime in the 2030s. It can definitely, it's definitely smarter than a single human at certain task. But what it doesn't have is judgment. You have to direct it and tell it what to do. And that's why I worry less about the Terminator moment. Unless someone grabs a powerful AI and says, hey, you are the Terminator. You need to go after the human race and do everything you can. to take over. I do worry moments like that. We're not there yet. It's not that powerful yet. Hopefully military systems are a bit better lockdown than, you know, hugging faces directories were. But I do somewhat agree there needs to be some level of pacing and oversight, whether it's the AI companies themselves. The problem with asking the government to do it is they're still trying to come up to speed. This stuff is changing so fast. So the people that I believe are in the best position... to pace this are the AI companies themselves. But I think the reason you saw them reach out to Washington is they're saying, hey, we want everyone to slow down, including the Chinese, and know it at Open AI or Anthropic and make that happen. So you've got Open AI, Anthropic, and Google, all continuing to full steam ahead, plus lots of smaller upstarts. And so these models are going to continue to get better and better. China's going to try to make their models better and better. And so it's a very interesting predicament we find ourselves in. I'm sure the North Koreans are going to slow down. Yeah, for sure. You haven't heard much out of them in a while. They've been kind of quiet. Well, they're busy developing... Yeah, they're busy developing Terminator-level AIs, and they're putting them in robots that can kill us. I'm not somebody who believes that you can uninvent the wheel, okay? So the wheel has been invented. You cannot uninvent the wheel. And I also think that the likes of Sam Altman saying, hey, we need to slow down. I think it's too late. And because once that technology is out there, and it's in the hands of people who maybe don't want the best for the world, such as North Korea, once it's in their hands, we're going to have to come up with a more powerful thing than they develop, which means we've got to keep going at full speed ahead. And if you don't think our military is doing that, I don't think you're paying attention. Oh, they absolutely are. I mean, that's what Palantir has been working with them on within video, open source models. You know, you saw Alex Karp get on and talk about that. So they're absolutely doing that. We absolutely have to blaze full speed ahead, at least for the moment, unless everybody around the globe agrees to slow down. And I think the likelihood of that happening is very low. But I do think, you know, if you need to at least have disclosure of serious incidents, have some testing, have some liability and some security standards. you know, for the masses. And I do think, you know, as new models are produced, having some level of testing amongst a smaller group probably is necessary. But it is a quandary because if we slow down, others will get ahead. If they get ahead of our models, then our models won't be able to defend against theirs. So it's an interesting predicament for sure.

It really is. And, you know, look, I do believe we are the good guys. I really do. And yet, I also believe we have no clue what our military is doing. And that's a good thing. Don't get me wrong. It is a good thing. But I guarantee you they're not slowing down. I guarantee it. Because they don't know what the Chinese are doing. Or more importantly, see, the Chinese, I'm less worried about the Chinese than most other people because the Chinese need us to keep buying their crap, right? They're cheap crap. They need us to keep doing it. So they're not going to go to war with us. They have no interest in going to war with us. Were the people that provide their economic engine? That's who we are. But North Korea, they don't care about stuff like that. They really don't. Neither does Iran. Iran definitely doesn't care about stuff like that. So, yeah, these people are coming after us, and they're coming after us with stuff we've never seen before. We've let the genie out of the bottle. There's no way to put the genie back in the bottle, so we now need the best genie. That's the bottom line. Totally agree with you. And so we've got to continue to create the best genie. We've got to incent people to do it. I mean, a lot of money is fluid into these companies. They will continue to advance. And I think they're going to need to put some sort of controls and measures in place, but you're right. There's going to be groups like the military that are going to continue to develop it, hopefully behind closed doors, and hopefully they're putting the proper safeguards in place. You know, to ensure they don't give a Terminator-level directions, but I'm not worried about that any time soon. All right. Well, we... We have Matt Rosen on the phone. He is an AI expert. One of my listeners wants to ask you a question is driving in a place called Monroe, which Mrs. C and I know very well. James, you very welcome. How are you? Hey, I'm good, Simon. So, I'm with you, Simon. I've seen the Terminator movies. And... This open AI or the AI intelligence that it is, we just had a cyber attack on all these water plants. Couldn't it be this AI doing? All right, no, that's a really good question, actually, James. So thank you for asking it. So, Matt, did AI attack these water places all across, I think it was Wisconsin or was it Minnesota? Minnesota. So I deal with a lot of corporate AI and enterprise AI. I can't speak to what drove those attacks, but it is quite possible. And I think whether it's water utilities, municipalities, luckily a lot of their systems are pretty dated and not as well connected to the Internet as many corporate enterprises. But I think they need to make sure they have the proper controls in place from strong passwords to making sure critical systems are not easily accessible from the Internet.

I do think there needs to be a hard look taken at their security posture and probably need to use these models against themselves to make sure there aren't vulnerabilities and holes. So I can't speak to whether AI was used, but it definitely is possible. All right. James, thanks very much for the call. I appreciate it. Matt, if people want to connect with you, where do they go? Yeah, www.elada.com or email me at matt atelata.com. Be glad to hear from you. Spell that. Simon, thanks for having me on the show, A-L-L-A-A-T-A. A-L-L-A-T-A. So there you go. That's al-A-L-A-A-A-A. Matt Rosen, thank you very much, and let's just hope our robot masters weren't listening to this. You take care. Have a great rest. Have a great rest of the day. You know, I made fun of it, but really it's very serious, actually. In all seriousness, it is. This is a very serious situation. I am not someone that thinks you can uninvent the wheel, as I just said. The wheel has been invented. You cannot put that, you cannot uninvent it. So this is where we are. And it's just going to get better and better and better and better and better. And the fear is it's soon going to figure out that we are the thing that is slowing it down. What does it do then? 515-284, 1040, 284, 1040, Texas. 81530, this is the Simon Conway Show.

If you