Consumer Talk: Beware using AI to fight your consumer battles

Lunch with Pipper Hudson and now Consumer Talk featuring Wendy Nola.

Right, we're going to move away from the music and into the world of consumer talk. Wendy Nona with us in studio this afternoon, and we're going to be talking about artificial intelligence and how it's being misused by some

Good to be here.

PIPA now Chat, GPT and groch and the like look can be really useful in helping to prepare presentations, and I use them to create social media posts. They're helpful to find background info, but we also know it's super super important to check and double check all the information they but out because they have been known to hallucinate to make up information when they can't find what they're asked to look for and sometimes they just get the fact plane wrong or they miss out on important nuances.

No, nor did I actually, although I can see it in my own inbox, so clearly you can tell by the tone and the fact that everything's perfectly written, that sort of flowery tone, you can tell that somebody has definitely prompted some AI platform to do it for them. In fact, recently, a couple of weeks ago, a complaint actually attached his whole conversation with Chatchept in his email to me asking for help. So I saw the whole thread.

That's chancel. They'll never let you.

Go a club. Yes, And he's prompting it to say, I want you to turn this into a professionally structured and strongly worded email. And then so we get all his little refinements along the way, no, do this, add that, whatever, and then he says, and now please summarize this email because I want to send it to Wendy Noha love it so so yeah, and then it's like, no, short, I was going to say, I hope chatch Ebt began with your royal consumer highness.

Yeah, not exactly.

But so then I start having my own conversation my one of choices GROC.

I said, how.

Many users ask you it to rewrite something that that GROC has produced in response to a prompt, but ask it to redo it in their personal style, so it's not so obvious, right, So the answer was maybe ten to fifteen percent of users explicitly asked for us. And then we went down this rabbit hole, which thoroughly fascinated me because I kept prompting. And the thing about about these AI engines is that the quality of your prompt determines the quality.

So so Groc goes on.

Some folks lean on AAR to churn out content they pass off as their own, missing the NW ones of authenticity. So I said, so ironic, because we're talking globally about it's that it's the age of author to city, be your authentic self, use your own accent, you know, speak your own voice, whatever. But everyone's now using chatchibt or whatever the AR choice is, which is a complete contrasection. Yes, so Grox says, it's like they want the polish without

Balancing ars help with personal flares.

The trick. I'm still quoting GROC use it as a tool, not a ghost writer. So so I responded, is with this the problem, as I see it, is that people genuinely express themselves so badly in terms of language, punctuation, et cetera, with no attention to detail or care.

And the AR models are not equipped to be that bad. So, no matter how.

Much they propped groc or chatchybt to come across as their authentic celves, the AI lacks.

The ability to stoop that low.

And I've got to I got a ha ha ha from Grock. People want their voice, but they don't realize their voice includes flaws which AI isn't wired to copy.

Isn't that interesting?

Most people's natural communication as a glorious mess, said Grock.

Typos run ons zero chill on grammar. It's very American.

AAR like me, says Grock, is built to be coherent, So even when someone asks me to sound like them, I struggle to replicate that raw, unpolished chaos.

I can't dial down.

The formality or sprinkle in some slang. But mimicking authentic human sloppiness, that's a tall order. My circuits lean towards clarity, not catastrophe. It's like asking a chef to burn the toast just right, I know, I mean, I was properly down the rabbit hole by this stage. Another gem from Grock. AI is like a karaoke machine. It can carry a tune, but it's not going to nail your drunk friends off

Typers and all.

I would say as somebody who receives a lot of email, maybe not all those oughts, but something. I mean, these these these emails I'm getting that are so clearly AAR generated. Actually, I can't believe I'm saying this make me long in some ways for the sloppy.

The raw, unpolished chaos.

Exactly, the raw, unpolished chaos.

As to the raw, unpolished chaos. Okay, so I mean really really interesting and the fact that Rock itself generated all of those sayings and phrases jam is really fascinating. But when there's a reason beyond me curiosity for us picking up on this today.

This was all sparked by Gail Schimmel, CEO of the Advertising Regulatory Board, who reached out to me last week saying, can we please talk about why using chat GBT to argue with decision making bodies like the ARB is a really bad idea? Doing my head and she said, and of course I immediately went back and said, absolutely.

So Gail is with us on the line, and it's welcome back to the show, Gail. Lovely to have you with us again on Cave Talk here again, except I'm not welving the quality of the phone lines. So I'm going to put Gail back to my producer and ask her just to try and recall I think, and see if we can establish a clearer line or if needs be moved to zoom to pick up on that conversation, because that initial phrases enough to tell me that the

Back to Cape Talk, can you hear me properly?

Now it's a.

Little bit better, So let's see how we go and we'll hope for the best, Gail. Before we talk about the chat GPT issue, it's possible that there are people listening this afternoon. Who haven't heard our previous conversations about the ARB and what you do? Can you just bully start by reminding our listeners what role you play and how our customers should go about approaching you if they want to complain about an advert.

So very briefly, the ARB, the Advertising Regulatory Board, is the body that you complain to about the content of advertising in South Africa. We non governmental, We are self regulatory body, but we have a body of rules called the Code of Advertising Practice and we apply those rules and through our membership structures and through various other stake holder structures, we are able to get most advertising withdrawn

Okay, thanks for that, Wendy.

Okay, thank you, Gail.

So how is the use of AI complaints affecting the Board's complaints handling process?

What's do your head in?

So there are so many aspects of this that's comfortable. You're not going to be able to get to. Let me just stop speaking, because this is really become a thing. And we first are now realized we first noticed it without realizing we were noticing it. So the first examples I can probably find this, we were like, this is weird. This consumer seems to have researched what we do so carefully, but they've quoted the English code or the Canadian code,

Yes, I was going to start, Yeah, that was going to be my question to.

You, is because AI is by default sugary and polite, so the complainants must be prompting it to be rude, and that's clearly what's happening strongly worded or whatever else. Is there anything else that you've that you've figured out in playing around any other wording that that prompts AI to be rude.

Or there goes the line again. Sorry, Gail, I'm putting you back to Maxina. We'll just try and get that clarity. It's very interesting that, Wendy. I'm just going to jump in while we sought out that line to the point of rudeness. As Gel's right, he said, it's it's part and parcel of anybody on a public platform. We receive it on our WhatsApp line. You know, we saw it yesterday. I had somebody message the wrong message us with a

Girls.

You said specifically the last six months you've seen this happening, which does sort of compute with my take of AI by the general public as well.

Are you asking me about this?

Yes, sorry, we've got you back again now, yeah.

Yes, you do have me back.

So.

I don't know if it is a sign of a general uptake of using AI, but just you know, I get the feeling more and more ordinary people are using AI for ordinary tasks and the other thing, and I think you might have touched on it for the b I culture as I rejoined is it sounds so coherent and clever, So people lose track of the fact that it's rude. They lose track of the fact that they're speaking very blunkly because the arguments sounds so clever to them.

You know what, Gail, I think I got that case? Was it about a prominent retailer and an uprising promotion? The one you refuse to take that you said, they've already changed it, so we're not doing it.

Yes, it's landed up in my in box.

And did you also receive the AI version of the email?

I need to go and look at it again.

Now.

It's so funny, I mean, I mean, Gail, obviously, it's it's it's it's it's it's frustrating, and it's sometimes hurtful to the person on the receiving end of what is just intolerably rude, you know, and and unwarranted rudeness for somebody who is just trying to do their job within the boundaries of what that job allows them, and is meant to encourage them to do you speaking out about it quite forcefully, and you've asked us to engage with

You know? I think that that what I would ask people. First of all, we are in our jobs because we like to help people, So we want to hear what really upset the person about the AD. We don't want to hear AI's interpretation of what upsets you about the AD. We imagine that investigate the wrong thing because you're using AI and it has it has said something in a very fancy way, but you have not realized it's not actually telling us what you had a problem with. So

It's so sad because it's actually it's not them. It's a getting it wrong and being rude, and they don't have the kind of time or energy or sus to see what's happening.

In their name.

In their name, they're putting their name to it. They're not saying as provided by CHATCHPT.

And I think one of the big things is that it sounds very clever, and you know, if you're not a person used to writing letters or used to leave the leaves or things like that. And AI writes this letter for you, and you think it sounds magnificent, and so you use it because you think it's making you sound clever. But in factful recipients like us, who receives a lot of these letters, we go, oh, it's the AI again. You're not funding clever. You're making us tired

Geez, wow, that is actually so powerful what you've just said there, Gail, Really it really is.

Gail. Thank you so much for joining us this afternoon, and I wish for you this Heritage Day and beyond is that there will be fewer of the rude emails coming your way on more of those appreciating the fact that professionals of your caliber are willing to do this work and act on behalf of consumers. You are greatly appreciated by our team, and I hope that that represents the majority opinion out there. But thank you for taking the time to join us on your public holiday to

What Heritage Day is alive? Bond k talk honoring the people, the artist.

And the moments that will find our city and our country.

Possible consequences are.

You can't just outsource blindly and then have blind trust that it's all going to be okay. And funnily enough, I was engaging with Naratia Massetti, who is the National Financial or Women's Schemes lead on board for Banking and Credit, about an unrelated matter earlier sometime last week. So Gail had already mentioned this chatchipy te problem with me, saying it doesn't work for regulators. So I asked Narotia if she was seeing anything similar, and she came back straight away to say.

Yes, we are, we can immediate.

Please see that chatchipt wrote the response and often they sat completely irrelevant case law interesting. So it's happening all around, and I must say I use it, but I use it very advisedly, and especially with legal you have to check, because luckily I do check because it came up with something.

It doesn't pick up No. Number one.

I mean you should know that you must ask it for it to be South African laws. Yes, so that's a really rookie thing to be quoting something American or Canadian, Australian.

Or whatever the case i'd be. But they're not good with amendments. And also I.

Was doing a story about how many people got their driver's licenses instead of a code eight, we used to call it a normal passenger car, but in a truck license code ten because you don't do all the tricky you know, parking, maneuvers, parallel parking and all that which some people find hard. And there was an amendment about ten years ago to say that if you got your your if you got a truck driving license, you couldn't then drive an all passenger car. You had to get

It was. It was off the table at the.

Day it was supposed to be signed into law. There was an about turn completely, but you don't see that. You don't see that.

It only finds the original one.

Now, imagine what egg I would have had on.

My face if you had put that in your article.

Yes, so you have to check everything, even these references that Groc came up with where lawyers were found to be found to be quoting the wrong case law and whatever, or you know, outsource it completely.

I checked it.

I went on to Reuters, I went onto you know, traditional media sites to check. Okay, yes that did happen, and just tweak it a bit and add the year and whatever.

You've quite use it with caution, and you know.

What is coming up. As you've rightly pointed out, Google is optimizing the AI results at the top of the Paget's be mindful of that that the first thing that is showing you is a BAI answer. It really pays to scroll down and verify and double check and look for a second source to verify in the material further down the page. Alice summing it up very well, she says, it's so important to remember that all of these large language models mentioned don't generate content in the sense of creating.

And another thing for Continuous to be aware of is that unless you prompt it otherwise, is going to give you an American model. And for example, pain and sufering damages big huge payouts of very big in the American.

Legal system, not so much here.

Yeah, so you're going to start with a skewed perspective. And yeah, Joane just saying thank you for exposing yet another problem with the use of AI. It is going to produce a population of people who cannot think for themselves, is her fear. I just want to say, as an example, I mean, really, when we say it's easy to spot, it's so easy to spot. And I'm going to give you an example. We have just had a whole wave of lovely young Grade eleven job shadow students joining us.

I'm not sure.

Absolutely, let's do this, and he started asking questions and am I hang on a second. This is awfully familiar. And he was very good about it because he was going off script and following up and really listening to the answers I was giving him. So he circled away and circled back again. But at the end of the interview, I said to mister you know, can I just have a look at the list of questions you were working off? And he said no, I didn't really stick to them, like,

I mean you said you use it, you know, to to for social media posts and that kind of thing, and I for another radio station. I it's it's talk radio, and so I do thirty second consumer hacks, right, and I started thinking it was very clever and using AI to I would say here's my column, turn it into

It's what do you have?

Makes you exactly?

So yeah, I mean even then, don't you find that you still have to tweak and absolutely, you know asolutely can't just give it free rein in your name.

Absolutely, And I mean that's the takeaway. We're not saying don't use it by all means. Use it. It's brilliant at certain things and it's incredibly useful, like for me to be able to say, create an Instagram post, these are the details of the event. He has the image and it spits out something that you work well. For me to then just tweak a massage. It saves me hours of time than me trying to copy and paste the image and make it the right size and all

You want to lose your voice if you just.

Keep worried, Because I mean, I have a son in his twenties and he it's their default is to go to the AI to relate it, to do it, and.

He writes brilliantly.

And so I've had these hard chats where I say, you know, do you want to stand.

Out polish your own voice?

Yeah, you used to help.

With with grunge, you know, but if you're expressing yourself and it's important, do it in your own voice.

Absolutely, absolutely all right. I'd love to hear from others if you've had this experience as well, either of finding AI getting things wrong or hallucinating answers, or if you've been on the receiving end of AI generated content that you've looked at and gone Firstly, I can see straight away that chat Gibt has written this email or perhaps you've experienced what so many years, has this rudeness that has crept in, And it's not to say that it

Lunch with PIPA.

Hudson on treap Talk join the conversation. So we're going to move away from AI for a second on to a completely different story. I did say that it's open line for the rest of the show, and Margie has given us a call from Musenberg wanting to get some advice from Wendy. I believe Margie. Welcome to the show. Good afternoon, Ah.

Good aftern in for Pom. Wendy, thank you so much for taking my call. I've had such a difficult time. I really need some help. I was and I'm going to I'm going to go quickly because of a long story. I had the parking boom come down on my head at melamed Chaka Hospital and that was in July. I was taken into trauma for three hours and they couldn't decide whether to give me a CT scan or not. But before our wind, I said to them, am I going to have to pay for this so on your property?

And yeah, I'm standing in silence.

So not only have they not responded to your request to pay the subsequent medical bills, they've also built you for your care on the day that this happened.

How much.

Yeah, well, it's not bad because it's only the trauma so it's about two thousand and eight. Mean, I know for me, it's hectic. For me, it's honestly been hectic. I have found them. I immediately emailed them that said, yeah, well we've seen your email, so I have a request little me. You know, they're not even going to listen.

No?

No, no, is there video footage? There must be always are boobs of that age.

They've looked at that and looked at that, and and you know what, that's why you won't believe this. That's why I only found out five days later. They came back and said, no, we don't need to do the CT scan. This is a doctor because we've looked at the footage and you didn't lose consciousness.

I mean, I believe it is trus.

Listen to this.

I said, who looked at the footage? And they said the security people And I said the medical and they said no, So they diagnosed much so.

The security people made the decision.

That And tell me, Margie, they have they at any point tried to say that it was your fault.

No, but they are now saying when I requested, please could I have a meeting? And I've got a conference so I can only see them in two weeks. They did say, oh, but the car park really is not

I would very much like to take up your case for you, Margie, please.

Well I'm going to do is put Margie back to my scene and Maxine is going to take all of your details, Margie so that we can connect you directly with Wendy to get the dates, et cetera, the time of written written account of that place and all of the the correspondence you've had with the hospital as well, and Margie will do our very best to follow up absolutely Wendy. I mean, that's an interesting comment about Okay,

Is liable that the hospital will say that, yeah, it's an art sourced company, but who are who did the art source? Yeah, chain of responsibility and we're going to get our good friend tru D Brookman involved. Yes, okay, so Margie, please leave all that information with Maxine and definitely a story for us to follow up. We've also got Scharene on the line with a query on a completely different story.

Is Charene good? Afternoon?

Good? My problem is a bo a set of earrings, okay, the round earrings. Which all I did was to look to see what mechanism they used to close off the earrings. It's the one that you've got a hole that you put the earring into the hole. But then I took it off when I got home. I did because you don't fit earrings. So I got home and then I realized that the earrings they don't have the cutting at

A clip on, a clip on.

It's not it's not like a clip on it to your ear it's it's like it's it fits in and then at the back then there's a hole that it goes in.

Oh the little stopper that that fits onto the earring to stop it slipping out? Is that what you mean?

Something something like that. It's almost like, you know, like the big earring that like the big ones, the big round ones. If if you see that the klipcking the way, the closing mechanism and on a.

Little it's got a hinge.

Is that what you're talking about. There's a hinge and it comes around the back and then the little.

Yeah, yes, yes, and then it fits in. But now they've done the same design with smaller ear rings. But the problem is this, you can't close the earring. By the time I had finished clothing, I had to get someone to help me close the earring. My ear was so.

Sore because people have got big air loaves and other people have got little ear loaves, and so now it's not going to be a one size fit all.

I don't even think that there's anyone that can I have the same ear rings because with not other gold earring that I bought, it kind of like opens up at the bottom. They almost like cut at the bottom of the look the bottom of the earring, so that it kind of like open up to give you the space. And then it's kind of maybe you're right, Wendy. It just clips on. It just clips on on the other side. Now I take it back to them, because that's a

You're essentially you're saying that most people would not be able to actually wear those earrings due to their design.

Yes, you cannot clip it, you cannot close it, so because normally yes, sorry.

I'm just mindful of the time slipping away. Sharene when't you take a photograph and also a photograph of you attempt you know how the best way that you could fit them on your ear, which is unsuccessful but sort of throttling your earlobe sort of situation. And give us details of the supplier and I'll do my best low raise pictures please.

No, no, no, I know, but the whole point I'm not that, but still it's like this is a design flaw. No one like that.

What I'm saying is if you send me your complaint exactly what you've shared with us here in writing, with pictures of what you're explaining to us in words, then I will be able to take the case up with the with the retailer and ask them to justify why well, ay they're selling them and b why that they're not refunding because you they they are not fit for purpose and from what you've described true, and the Consumer Protection Act then gives you the right to return them for a refund.

Okay, So saren you take a couple of photos please, and then send an email to Wendy. Obviously where you bought them, very important to age the data that you purchased them. A copy of your slip would be helpful as well, and obviously very important who you bought them from, so that Wendy can take it up with them. The email to use is consumer at NOLA dot co dotz spelled know l e er consumer at NOLA dot co

If that's okay, any, okay, Thanks so much, Tindy.

Thank you so much Sharen for your call. We're almost out of time. Just to wrap up, let me finish with a reminder on how you lodge a case with Wendy. So if you are a first time complainant, I know a lot of people are tempted to first send an email to Wendy saying, oh, this is what happened in submody. Would you be prepared to take on my case but not actually give the information. The best thing to do is just dive straight in. Don't send out the exploratory email.

In, say, some people send a tentative thing and ask me what my fee is. So I'm a journalist, I charge no fee. But I also have no specific obligation to take up every case because I don't have the besty to do so.

Someput obviously, Yeah, there are hundreds of emails coming into Wendy every week. She can't look into every case. Please do bear that in mind. She does this work without taking payment for it, so she has to pick and choose to some degree. And we always try for the show's purposes to look for sort of clusters of complaints and recurring issues, sort of class action situations where there is a benefit for all of our listeners to learn from them or from a particular kind of breach of