It's Night Side with Dan Ray on w b Boston's news radio.
Hi, welcome back, Gary Tangline for Dan here this evening. The Dark Side of AI is our next topic. Scott Barredeal joins us to discuss this in anytime, and I know I Scott off the top of the show. First of all, welcome to the program. Off the top of the show, I said, I'm the biggest hypocrite because we're so dependent on AI. You do it now, if you talk into your phone, you have, you have spell check,
just so many ways we're dependent on it. But I love to root against it, you know, So I'm a hypocrite. It's like I don't want AI. I mean, I was taking jobs. It's taking humanity out of things. I can write things that can you know, thanks for itself. It's going to take over the world. But boy, man, if you took away my phone and all these things, I'd
be screwed. So this is an interesting situation here. First of all, I should ask you where do you fall as far as being pro or anti AI before we get into this situation.
You know, it's a good question. It's kind of like how I am about phones. It's like I hate that everybody, including myself, are on their phones all the time. But I don't know what I'd do without it, you know, you know, it's like that we it's it's great, it's a great tool. It makes things more efficient, but it does take some humanity out of out of things, out of your life, and it starts to affect how you view the work of others. Is this is this real
or is it mem recks? Did they actually do this or do they just have it bat out by chat GPT?
Well, we've seen that in commercials.
I can't remember which god which product it's for, but there's an actor. For some reason, I think the actor is the guy that played Booger and Risky Business. I don't know, maybe I could be wrong about that, but he sits there and he goes and he and he starts to type a memo, but then then he goes into chat whatever you just called it.
I don't even know what that chat.
Bob, either chat BT or whatever GPT, Yeah, chat GPT, and it totally changes the tone of it. It makes them sound brilliant. And then they pan over to the boss and the boss goes, wait a minute, did you write this? I think that sucks. I think that's see that's the part I think that sucks. And I know everybody does it. You know, it's it saves time and so forth. But that's just kids don't have to write, and I think that sucks. That's the part I don't like, you know, but it's hair to start.
Well. Yeah, And but the other thing is, you know, the thing about chat GPT and these tools is is they they won't tell you they're wrong. And you know, there's a recent study done by Purdue University that found that about over fifty fifty two percent to be exact, if chat GPT answers have at least one error in them. That's a little scary.
So, I mean, that's very scary.
It would be better if if if they were person accurate. You know, there was actually an attorney to win to a courtroom in Colorado and talked about all these cases to make his tell his side of the court case. And it turned out that the cases were completely made up by chat GPT. They were not real cases that had happened, and so the guy was fired from his law firm and he was suspended from the bar in Colorado. So you know, you don't take it at face value. You better check check the facts.
Wow, I mean that that Wow.
I mean I hate for somebody to lose their job, but come on, that's ridiculous.
Now.
The reason we had you on there was an article. It's a story about a young woman and she was using AI to help her with some homework. The chat by Gemini, Google's Gemini and with the chat bat chat bot. Excuse me, this language is forn of me. Uh said to her at the end was human please die right? And she had a I mean, it's it's funny, but it's scary. She had a friend with her in the room. But I mean, how did it's certainly it could happen. I mean, of course it could happen. There's a glitch,
there's something that goes wrong there. How often does this happen?
Well, there's been glitchens, you know from you know, just in this case. I'll give you the full quote because it's even more impressive. There was a long if you've ever used or no, folks can use these tools, and most people are using him today. You might ask a lot of questions to try to get the exact output
you want. And and it's like in this and you can see the full chat on on online that's been It's been safe that this grad student had with Gemini, And the final it's like Jimi is getting more more frustrated, almost like it's a person, and finally says, this is for you human. You are not special, you are not important, you are not needed. You are staying on the universe. Please die, good God. So at that point they completely this person is completely freaked out. It's about asking questions.
So maybe maybe Jim and I had just had enough. But that's not the first time it's happened. You know, a few years ago, Microsoft had a chat bot they created for teams called Tay that was almost immediately decommissioned because it was making racist or marks it was saying. It was saying things like Hitler was right. Horrible, just unimaginedly horrible things. Chat GPT had itself, which is of
course the most popular large language model currently. Uh, you know, they had a malfunction where people were actually seeing the questions that other users that they didn't know were asking, so complete violation of confidential a horrible privacy bug. So meta, you know, Facebook, you know they had something called blenderbot and when people were asking it questions about whether it's
Trump elections or Kennedy assassination or whatever. It was totally you know, saying all totally got into all these conspiracy theories and saying, yes, the conspiracy theories are the facts. And so it's just across the board. There just has to be an understanding that these tools are very much still in development and you just can't take them as gospel. They trick you into making you think they can do anything,
but they can't. I work with a company called acquire BPO, which is a business process outsourcing company that does call centers and helps companies with all kinds of other work around the world using people, but of course introducing AI as part of what they do, as all companies are increasingly doing. But we did a survey recently acquire BPO where they found that seventy percent of Americans said that if they had one bad experience with AI, they would
consider switching to another brand, to another company. So it has a major commercial implications. If you're not watching this stuff really closely.
Has it ever cost someone in their life?
Wow? Good question. I can't take an example of that off HANDA I mean what I immediately think of is sky net became self aware at two fourteen am Eastern timee That's that's the line from the original Terminator where they where it was all about an AI defense system that became self aware and they decided to since they were in charge of the weapons, to nuke all the humans, and that's where the Terminator came from.
So well, yeah, I mean, I don't know, I.
Don't know a real life example, but.
Well, this is what comes to mind from me, is when you see situations like this where you're talking about you know, verbiage. Okay, so it's okay, so the computer insulted somebody, all right, fine, But when you talk about depending on AI, maybe in life threatening situations, when it comes to dealing with ambulatory services, when it comes to dealing with firefighting, when it comes to dealing with work,
I mean, that's the thing where I'm concerned. And I'm not talking about something catastrophic like war games or where they launch you know, one hundred missiles or you know, I'm talking about in a situation where you may have a public servant, firefighter, police officer, a medical person that is in a situation, depending on information where they need to try to go into a fire or they're going into a violent situation and the information is wrong because it doesn't come from a person.
Well, yeah, I mean I think it is a legitimate concern. I definitely think that you you know things, There are improvements happening every day, like, for example, in these tools increasingly, as opposed to just putting information out there that may or may not be true, there are links to the
original sources and things like that to verify. I would say, you know, for companies, particularly for situations like you're describing, but I would say for companies and in general, taking a more conservative approach where you're not asking the AI to do too much, where you're making the tasks a little simpler and so you can be sure you know it's directing you to information that you know is true, as opposed to you're asking it to solve the world's
problems or you're you're asking it to really what it does. It's it's just an advanced form of memorization. And they memorize and they and they put answers to different questions together essentially, and they provide they're kind of program to always give you an answer and if you're program to always give an answer. It's not always going to be right, so you can set it to not do that. You know, you don't have to set up your AI to go that far.
And it's also based on something that's previously happened, correct, So when you have a situation that is new and the computer can't refer back to anything, you're screwed. More on a coming up AI coming up next, The Good side of AI. We like to be fair and bound here, you know, we try to do our best. That's coming up next on WBZ with Scott Barredelt.
Don't go away.
Now back to Dan Ray Live from the Window World, Nice Side Studios on WBZ News Radio. Welcome back, Gary Tangling for Dan Ray tonight. Scott Barredeal joining us here talking about well, we were ripping on AI and I do want to get to the positive end of things. But when do you think, Scott, it'll be so cost prohibited or maybe we've seen examples of it where a company says, all right, we're losing so much business.
This AI isn't working. I need to get a real person back in here.
Well, I think that's for companies that kind of kind of go too far, too fast. I think there definitely been some examples of companies that have maybe dived in a little too fast, went too far, and then they had to pull back. You know, I mentioned this survey by a choir bpo, which if you want to see the full survey, it's at at acchoirbpo dot com. But what the data shows people are most confident in a situation where there's a blend of human support and AI support.
They like that AI support can make things go faster, so like you don't want to be on hold for thirty minutes before you get to talk to someone, things like that. So there are a lot of things that AI can do. And if they know, hey, if I'm not getting what I need from, AII can escalate to
a person. I can use AI to help me identify and consolidate kind of the past customer service experience I've had, for example, but then I don't then I can talk to a person to solve the problem I'm having right now. You talked about, you know, AI knows the past but can't predict the future. That's where it's a good idea to have that blend. Companies that have made the mistake kind of like this lawyer made of thinking, Oh, I can just AI looks like it knows what it's talking about,
so I can just assume it does. Just remember it's wrong fifty two percent of the time, or something is wrong in answer and you have to check it. It's kind of like in the early days of Wikipedia. Really it's still true today, but people would say, hey, don't Wikipedia is not, you know, gospel. You can don't do your research paper and just say I got it out of Wikipedia. You have to. It can guide you, but go find the original sources.
Well, even when I do research for the show, when we talk politics or we do something of a serious nature where you want to get your facts straight, I will online. I'll start reading different things in n'lb like three quarters of the way and I go, wait a minute, this is bull crap. Where did this come from? This doesn't make any sense. You know, you can't.
You have to.
You have to vet it. You got to vet stuff. I mean, listen, I called my bank the other day. I'm not going to mention the bank. I call my bank the other day. And you know, when you go through you punch all the buttons and for this and for this, But I had a certain situation. I had a certain question regarding as regarding an account, and I got a person because I like to talk to people, and then that person had to send me to a supervisor, and that supervisor didn't know the answer, they had to
send me to a manager above the supervisor. Where your point is well taken is if the first person had the AI to give them the answer, then they could have.
Just simply told me and I would have been.
Satisfied right right for how you yeah exactly?
You know, because yeah, because the answer was the answer. It wasn't hard. They just had to get to somebody who actually had the information. I was on hold for an hour.
Yeah. It's like when you go to the doctor's office and you have to fill out the clipboard every single time. It's like, are you saving any of this information from one visit to the next. Yeah. So there's definitely some advantages to automation, which of course has been around forever, and AI that is kind of taking that to the next level.
What about the positive end of it.
Well, there's a tremendous amount of a positive if you know how to use it. In other words, you can be far more productive. I can tell you that you know. Part of what I work at at a PR and marketing firm and we do search engine operstation things like that.
I can tell you more and more people are doing their searches through a chat CPT or GYM night rather than going to the traditional Google search, because Google will send you to a page somewhere and maybe it's an optimize, but maybe it's not the best answer to your question. And in the case of these, searching for what you're looking for and doing your research through a tool like a chat CPT, and again increasingly they're including their source links and things to make it more so you can
check that it's it's accurate. It's a much more effective way of gathering information, and it's increasingly becoming. It's predicted before too long it's going to be the number one people the way people search for the information that they're looking for online. So it's inevitable that it's going to
keep getting better. It's it's it's amazing how far it's come and how many people are using it in the short amount of time it's been around, have been widely available, So you've got to assume that at the rate at which it's improving and progressing. A lot of these things that have caused the problems are going to go away.
What about in the medical world.
Well, in the medical world, obviously you have to operate with caution, you know, I would say medical as well as kind of highly regulated industries like finance. For example, there was a big case with the think Bank of America where uh Ai was chatbot was giving given people uh incorrect information about mortgage loans and kind of important financial transactions. In the medical world, obviously even more important to to be careful becused. You're talking about life and death stuff.
Yeah, it's it's when I think about situations like that, And the reason I asked the question is if you need to so you have a task, either in the medical world or the financial world, and you need one hundred percent activity to complete that task, I can see AI taking you to twenty five percent and then somebody else has to take you the rest of the way.
That's you know, that's what I see.
I don't see you can't have AI complete one hundred percent of that task in those areas.
I think that's that's a good way to think about it. I think the higher risk the task, the more important it is to have humans more involved, particularly at this stage of the development of AI. But I will tell you we actually did a survey a company called Christa, which is like they're like a Amazon Alexa conversational AI,
but they work within companies. They did a survey asking people what they trusted AI to do and not do, and it just showed you kind of the amail confusion out there, like is AI going to help you my job or is it going to take away my job? But one of the questions ask quite the scenario is like, do you trust AI to pick your wardrobe for work? Or would you press AI to fly a plane with no pilot in it? And more people picked I would trust AID to fly a plane but no pilot in
it than picked my clothes, you know. So I think there's a lot of people that are just still it's also new that they're all trying to come to terms with it and at least to some weird tuxtapositions like that.
I will leave you with this, I will never ever get in a car that drives itself ever. Ever, I've heard that, Okay, you can have a driver, you know, you can get in the car, you can sit in the back seats, you know, and you can work and you can you get on your laptop and do work and you punch in where you're going to go and the car takes you there.
No way. I don't trust it never. What are your thoughts on that.
You could get carjacked? Also, I don't know.
That's crazy.
I mean, I don't get that. And they've had it, They've had tests where people.
Have been killed.
Guys get the vaccine, and they have supposedly a driving less car. I don't get it. That's the part where it's it's insane.
Like I think I'm going to be old fashioned on that one.
Yeah, I don't.
Even think with being old fashioned. I just think it's common sense, you know. It's like, look, technology has helped all of us. I get it, you know, and I'm open to it. But I mean, look, voice commands on TV. When I want to go, like, you know, into my TV you know, watch you know, watch the Boston Celtics bowmen, it pops up. I love that you can help me with that, but I don't need you to drive my damn car.
Yeah, I think I'll drive myself too, for the indefinite future.
All right, scart Baradel. What's the name of a company.
My company is called Idea bro and the survey is by a company called a Choir BPO.
All right, well, listen, I appreciate you coming on. You sound like a reasonable person. You've really come I I really have enjoyed this conversation because you've made me feel a little more comfortable to be quite honest.
Well, thank you. Yeah, I appreciate that.
I appreciate that. Have a good night and have a good thanksgiving.
All right, you two take care, all right.
Scott Barreto joining us here and AI the good and the bad. Yeah, the AI voices please die? Are you kidding me? Good?
God?
Coming up next the final hour here on night side, this evening. What are you thankful for? And I understand. We're thankful for our families. We're thankful for Rob Brooks. We're thankful for Marita Lo Rosa that you know, our producer. We're thankful for our family. We're thankful for our health. But let's get a little creative. Let's get a little creative, let's get a little unique. Give me some different takes
on what you're thankful for? At six one seven two, five thirty we're going to open up the phone lines. I do have some pretty funny lists that we're going to get into. Unique things to be thankful for. Coming up next on WBZ