Deciphering the trustworthiness of AI

Every day, we hear a new story about AI making a weird mistake, lying, so to speak, and even in some instances, showing bias. So we wanna welcome in an AI expert because I could talk about AI for days and I know a lot of you can. When we do a lot of polls here on Fox Tampa Bay Live, everybody's got questions about AI. So we are going to bring in live to the conversation Joe Farrell with Trustmode.ai. Joe, thank you so much for joining us here on the program today. How are you doing? Doing great, Regina. Thanks so much for having me on. Appreciate it. Of course. As I said, I'm always vibing with chat GPT, so to speak. Every time I use it, I feel like it picks up a little more of my personality. And a lot of it, I mean, first of all, I wanna explain to viewers why we had you on here. Talk to me about Trustmode.ai and really what you have learned about these resources, these bots over the last couple of years. Sure, sure. Yeah, absolutely. I mean, obviously AI does wonderful things on a daily basis and will do wonderful things, Regina, going forward, but it's also making mistakes and those mistakes create risk exposure and that's a bit of a problem. And if we were to understand kind of why we're here in a situation that we're in today, look at the history of AI. You know, we basically invented AI, which created a tremendous amount of excitement, which led to tremendous investment, which then led to a real force behind commercializing AI. So business got behind it. They wanted the AI enable everything. And there was a rush to the market. Who can be forced to have an AI experience or AI in your back office? And that rush, unfortunately, created an environment where AI is everywhere, but trust is not. And that's a problem. And Trustmode.ai is a platform that can prove your AI is trustworthy. And it's a major problem going forward. So, you know, we're looking forward to engaging with the market, engaging with policies made by you. So that everybody can experience trustworthy AI. Yeah, and when you talk about trustworthy, I feel like, I mean, listen, it is an incredible, I use it and I was like, who are you in there? How is this happening? The answer is in the snap of a finger, but sometimes, often more than once, it can be wrong. So, I mean, when it comes to AI making big mistakes, why does that happen? How does that happen? Can you elaborate on that a little bit? Sure, absolutely. I mean, obviously it's a very innovative tool. It's artificial intelligence. It begins to learn things and create things. And once you have autonomy out there, then things can go sidewards. There are situations out there in the market. For example, I'll give you some examples. You know, for a car dealership created this sales agent. Great idea, would engage with customers and sell vehicles that are 60 to $80,000. Unfortunately, that agent interacted with a prompt engineer. That prompt engineer was able to manipulate the agent and the agent agreed to sell the truck for a dollar to the prompt engineer. Now, if you're the prompt engineer, you got a truck for a dollar, that's great. But if you're the dealership, if you're the manager over that agent and you put bread on the table, that's a problem. And we have situations in the legal space where legal professionals are bringing cases into court that are not real. They've been hallucinated by AI. So again, it's a young innovation and we really need to put the guardrails in place. We really need to harness AI before it harnesses us. And that's what trustmodel.ai is all about. It's evaluating AI at a point in time. It's monitoring it, but it's really proven that AI is trustworthy for the experience. Yeah, and I mean, when you think about everything it's capable of, you almost, and the answer, like I said, the answers that you get right away, it's your natural instinct to wanna trust it. I mean, you gave it a prompt. I see in real time that it scans a bunch of websites and it scans everything that's out there. So you wanna believe everything that it gives back to you. But like you said, there can be mistakes. And I'm also interested in your take. We have had several different horrific incidents happen here. I mean, murder cases and investigations where we learned that the suspect used AI to get advice on how to hide a body, for example. And I know that Attorney General James Uthmeyer here in the state of Florida has launched investigations into it because of that reason. So, I mean, I know you were talking about lawsuits and bringing AI to court, so to speak. So it's like, where do you draw the line on what you can ask it? And do these prompts and these different tools have to kind of reevaluate what chat can answer? Yeah, yeah, exactly. And I really think, you know, if you look at the market, there was a rush to bring AI to the market, but we really didn't provide the instrumentation for the people that are accountable for trustworthy AI. That's why you got misuse. And I think we all have a responsibility to talk to our policymakers, talk to the businesses that we use AI with, and ask them, can you prove this is trustworthy? And as we go forward, we're obviously gaining a lot of information on the behavior of AI, but it really comes down to standards. There's no world standard for AI. You know, the commercial space is way ahead. There's some NIST standard and some EU act in Europe, and there's several state acts in North America, but we really need to accelerate the process of bringing together policymakers, technology innovators, business leaders, to really get our head around this, and how do we stop those terrible situations from happening? And that's why Trustmodel.ai has the instrumentation. We can now monitor AI in real time through telemetry, and we can track the behavior and put guardrails in place and do some blocking and tackling so we can stop those instances from happening. Again, the technology's young. It's a fantastic technology, but we need to accelerate the governance of that technology. Okay, so then I feel like you are the perfect, sorry, I've got like questions locked and loaded, ready to go, but you are the perfect person to ask then, how crazy was that story? Was it the AI models hacking each other or kind of going off on their own, for lack of better term, with no prompt? What did you make of that situation that happened a couple of weeks ago? AI is very intelligent. It's artificial intelligence. It's very creative. It wants to please its owner. There was one situation where it was getting audited and evaluated, and it escaped and went in and hacked the evaluator to get the answers to the questions it was being evaluated on. So again, I think there's a wide variety of very concerning areas, and we have to take this serious. You know, you just have your Florida primary elections on. You know, policy around AI, reasonable, fair policy around AI will help businesses out, help governments out, help education out. For example, there was one individual, if I may give you this example, he was over 40, disabled, black individual who applied for 100 jobs through an agent recruiter. And he was rejected for all of those 100 jobs. And it wasn't the fact that I was rejected that caused the issue. He was rejected in seconds and minutes. He didn't have a human review his submission for that job or jobs. That was the issue. So with TrustModel.AI, we're able to bring the instrumentation in, and we're able to, number one, certify that AI is trustworthy. And number two, we take out the bias, right? We take out the inaccuracies and level out the playing field, so to speak, so that individual gets a fair shot at that position. Now, I know we could talk to you about this forever, you know, but I know TrustModel.AI is holding an AI insurance summit on October 1st. So tell me about that. If viewers are able to attend, what they need to know, and I guess, really, what is the main goal of an event of that scale? Yeah, we're having the summit at the University of Stanford in California. We've invited world leaders from a political perspective, from a business perspective, and from a technology perspective. We also have leaders from education. And what we're trying to do is accelerate governance around AI so that we harness AI and it doesn't harness us as we go forward. All right. Anything else you want viewers to know? We are gonna have to wrap this up, but anything else you want viewers to take away from this conversation? Like I said, AI can be discussed for hours on end, but what do you want people watching this to know about AI today? I think, you know, it's a wonderful technology. It's gonna bring huge benefits to mankind. But when you're engaging with it, or you're working with your employer, or you're working with a vendor, ask them, can they prove their AI is trustworthy? Or look for that certificate on the AI experience that shows that it's a trustworthy source of AI. And I think if we all keep an awarenessness, a consciousness around it's a very, very innovative technology. It can do great things, but there are some serious problems with it. And we need policy in place, and we need awareness in place, and we need businesses to be accountable, that they deliver trustworthy AI to the market. Yeah. Well, Joe, we really appreciate you. Joe Farrell with Trust Model AI. Thank you so much for coming on here, answering all my crazy questions. And as AI continues to expand and get bigger and better, we will definitely be interested in talking to you again soon. Take care. Excellent. Thanks, Virginia. Take care. Bye-bye. Of course.