Are AI toys a good idea for my toddler?

Hello, and welcome to Raising Generation AI, with me, mom of two, Holly Green.
And me, AI expert, Fiona Morgan.

This podcast is all about the complicated world of parenting and AI, because I’m a mom, and I’m a bit worried about it all.

And this week, we’re asking the question, are AI toys a good idea for my toddler? Later on, we’ll be speaking to Dr. Emily Goodacre about what she’s learned in the first study of these AI toys in the UK. First of all, Holly, and most importantly, I have figured out what went wrong with your AI invitations the other week.

Have you? What was it? Tell me.

Yes. For anyone that hasn’t, didn’t hear this or didn’t see it, it’s on social media. You can go and have a look.

If you want to see the worst generated birthday invitations you have ever seen. Sorry, Holly. Then please go and have a look at our social media.

And I sort of picked up while we were in the episode, because if you remember, I said, I thought Claude couldn’t generate images to which he responded quite rightly, it can’t look, which was my favourite line. But it’s basically generated the code. It’s almost like made a website of the invitation.

So it has just built up from shapes and like Microsoft Paint a bit like you thought it had. It hasn’t generated an image in the way that the image generators do. So Gemini and ChatGPT generate these images pixel by pixel.

What that’s done is essentially written the code to say, draw an oval with these dimensions. And it’s basically done it in HTML, CSS. I don’t know much about front-end code, but it’s built it in code, created it as if it was making a website.

Can I give Claud a top tip and suggest that in future, it just says, this isn’t my strong point. Maybe, maybe go elsewhere for this.

Anthropic, get on the phone now. We’ve got some tips from AI expert Holly Green.
I feel quite relieved then that it wasn’t just me being completely useless.

Yes. If you want to do images, your best bet of the big three, ChatGPT or GeminiClaud, can’t generate pictures, although it can generate code that effectively makes terrible pictures.

Doesn’t look anything like what you’re asked to make. I did feel like I just been utterly, utterly useless. Although sometimes when I use ChatGPT in these programs, it does go okay.

And I gave a go to your homework this week. What’s in the fridge and what can make the dinner homework. And it went all right, actually.

Great.

So do you want to explain the premise of this for anyone that hasn’t seen it?

Yeah. So every Saturday, I post a little activity related to parenting or related to kids.

Just a little something you can do with AI to try it out if you’re not particularly confident or want to learn some new skills.

And it’s always privacy safe, free, those kinds of things. This week, we did what I named very lovingly Chef GPT. You take a picture of the inside of your fridge or the inside of your cupboard or both, and you send that into the AI and get it to suggest you some dinner options.

And how did it go? So it went quite well.

It went well. So I didn’t use your prompt. I went with my own prompt because I thought, do you know what?

I’m going to use a bit of brain power here. I’m not going to source it all out to you. And it started quite well, and it came up with some good suggestions.

It got weirdly obsessed. I had a really old jar of rhubarb jam in the fridge, and it incorporated this rhubarb jam into every single recipe. Everything was like, oh, and I had a bit of the rhubarb jam for something special.

So that was a bit strange. But it came up with some really good recipes, and that was great. But then it started suggesting things I didn’t have in the fridge.

So for example, it suggested, I make beef burgers. I just need to go to the shop and buy some beef burgers, which felt like it slightly defeated the point of the exercise. But it did come up with some good stuff.

It was all quite basic stuff. So I was like, let’s elevate this and see if it can come up with something a bit more fancy. And I said, OK, what about if the queen is coming to dinner?

And then it started to like throw some shade on my fridge contest. I’m just going to read this to you. It said, if the queen was coming to dinner, we would obviously ignore the fact that this is a Tuesday night fridge containing sorry looking bean sprouts and a giant cup of yogurt.

I was like, all right. I’m getting enough sauce from a six year old without getting it from Chat GPT as well. And then it basically just came up with the same recipes again, but just smaller portions of them on fancy looking plates.

And that was what I should cook at the Queen’s Gala.

But yeah, I can believe it.

I mean, I don’t know if this is a new thing, is Chat GPT just getting a bit more smart Alec-y?

It’s funny because I think sometimes it will reflect back the kinds of language you’re using with it. So are you particularly sassy?

No, I was perfectly polite.

With the prompt that I made, the bean sprouts weren’t old.

I just think that in the shop they were quite fresh bean sprouts. So I did feel a bit wounded and judged.

I don’t know if you saw, but did actually then try the prompt that I sent in, but it was quite specific about saying only using the ingredients in the fridge and maybe some staples that you can just quickly pick up from the corner shop. So I think sometimes when you add some extra information, maybe it wouldn’t have suggested you just go and buy some burgers.

That is really funny.

Yeah, what I’ll do next time is I’ll try my prompt first and then your prompt and compare and contrast results.

Well, this week we have, and hopefully you’ll get an opportunity to do this, a homework helper. So you’ll have seen it went out on Saturday. And it’s basically for those times when your kid comes home with homework, and you just think, I have absolutely no idea what this is expecting me to do.

This is a stupid maths formula that I just don’t understand it. And actually getting ChatGPT, Claude, Gemini to talk you through how to do it so that you can help your kid. I was doing it with some multiplication methods that I found for six-year-olds, which I was surprised there might be six-year-olds doing multiplication.

But it was very basic, but it was how to explain it to a young person at that age. And I found it quite helpful personally, although I was a maths teacher. So that’s the one this week.

It’s to give you a little bit of extra help. It’s maybe going to be even more helpful for those who are parents of teenagers.

I tell you what, some of the phonics they get are quite confusing nowadays because they learn to read in a very different way than we learn to read. And sometimes I look at the phonics, they’re very odd. So could it help with that?

Could it help me with some phonics?

Yes, that sort of thing would be really, really good. So give it a go, maybe give it a go with some of the phonics stuff and let us know how you get on next week.

Cool, I will. I’ll put back next week.

Hopefully you won’t get sassy GPT.

No, I know.

What? You can’t do this?

Can you not read? Right, what have we got today? We, oh, I’m excited about this one.

So, AI toys. Now, I was trying to find a friend or someone I knew who had bought an AI toy for their kid, so I could find out how they were finding it. I couldn’t find a single person.

And more than that, a lot of people just look confused when I said AI toys and said, what is an AI toy?

So, I think the reality is that most people are not buying these toys. AI is expensive.

That’s something that you might see if you’re on LinkedIn or anything like that.

A lot of people talk about how these AI companies, Anthropic and Co, have not been typically making money. It’s a very expensive thing to have subscriptions and pay for this stuff. So, it’s not like they’re just turning up in your 10-pound toys that everyone has.

These are more expensive toys, but the most mainstream one is something called Miko. And that’s a robot learning companion. And it is available in the UK, 200, 300 pound, I think, for these toys.

We’re not talking everyone’s just got them, right? But they do exist.

And when you say a robot learning, we’re talking like a little small looking robot thing, are we?

Yes, little tabletop robot, and you have to pay for a subscription and stuff on top of that. So, it’s not an overly accessible thing for just everyone to have. Miko is probably the most mainstream one.

And what it is, is it’s a robot learning companion. They market themselves as education and friendship. So, it is that companion, yes, element of it.

But it also helps you with math and subjects, and you can buy additional cards and things like that to expand its knowledge, I guess. Now, I’ve had a good look at this, and it does say that it’s certified as safe by a couple of different US bodies. Hopefully, it is, but ultimately inside, for those of you who hopefully now understand what a language model is, we’ve been talking about it quite a lot.

Effectively, it’s a language model in there. So it is making up things on the spot in relation to what the children are saying. Some of the toys also have cameras in, and they have microphones.

They do claim that it doesn’t go anywhere, it’s just local, and it doesn’t have access to the broad internet, so it can’t just search for things and respond based on the broad internet. Ultimately, these language models are still trained on language, and they haven’t said what language it’s trained on.

Oh, so it could just be trained on the whole of the internet, for example?

It could be a bit like ChatGPT or Claude. It might even be using one of those. They haven’t specified as far as I can see.

They may also have trained their own model. So whilst we talk about Claude, Gemini and ChatGPT a lot, there are actually hundreds of these models, and some of them are what we actually call small language models, which tend to be far more focused on like a subset, and they might just be trained. So for example, a financial language model might just be trained on loads of financial books and documents and things, and therefore it’s not trained on what’s spoken about on Reddit or anything like that.

So they may well have built a pretty robust small language model around children’s language, but I can’t find enough detail to say.

Sorry, I have a few questions here. Firstly, if it’s not connected to the Internet, does it not need to connect back to the data centers to do the processing of the information?

It’s connected to the web, so it can send information to where it needs to go to data centers and stuff. What it can’t do is search the Internet. So it can’t go and get more information than what the model itself knows.

So if it’s connected to the Internet, presumably it’s potentially hackable. I don’t like the idea of that. Something that has a camera, something that can talk, something that’s listening to you being potentially hackable.

That sounds a bit of a concern to me.

A lot of people have things like ring doorbells and security systems that are connected to the Internet and things like that. Baby monitors, they’re often connected to the Internet.

So I can’t remember who spoke about it.

One of our guests spoke about it. It was Dr. Nick, actually. He was saying how we need to get to a stage where we’re comfortable enough with the technology to use it.

And so we’ve got comfortable enough to have a ring doorbell for a lot of people. Some people very much still not, but comfortable enough with these baby monitors. And I do think that we’ll get to the point where we’re the same for toys.

I’m sort of on board with the education side of things, but it’s the friendship side to me that feels a little bit problematic. I don’t know, I just feel like having an electronic friend could interfere with real life relationships. And also you wouldn’t know what that robot, that large language model is communicating to your child.

We know they hallucinate, we know they make things up. What is it going to be telling your child? I don’t know.

I think there’s going to be a lot of really good questions to ask our guests today.

Definitely. And I think Dr. Emily Goodacre is the perfect person to talk to about this. She comes at it from a children’s development perspective, and has been doing a study with these toys, with preschoolers and with their parents, and looking at how they interact and what kind of impact they could potentially have on the development of children.

And so we’ll talk to Dr. Emily Goodacre right after this.

Now, in the same way you can have a conversation with ChatGPT, your kids can now have a conversation with their toys. Now, it might sound a bit similar to some of the toys we grew up with. Think Furbies, if you remember those.

But these ones are a bit different because they actually listen and remember what your children say, and they do give a different response every time. So it is a bit like ChatGPT being within the toy. Now, some of these are also being marketed as tools to improve children’s speaking and listening, maybe even sold as your child’s new friend.

But as with a lot of AI right now, the research is a bit patchy and often it’s not independent.

Until now. We’ve got Dr. Emily Goodacre.

She’s a developmental psychologist in the University of Cambridge, and she spent the last year running the first proper study of these toys with young children. She’s actually watched them play with them, and then she’s spoken to their parents afterwards.

Dr. Emily Goodacre, welcome to Raising Generation AI.

Thank you. Thanks for having me.

So, just to start off with, give us a little sense of the sort of toys you were studying and what made them different to a normal toy.

Yeah. So the toys that we chose to study, they were all kind of like stuffed toys. I think like a teddy bear, but they’ve got like a speaker or something in the back that connects to the internet, and it runs whatever the child says through an AI chat bot.

So the one that we used in the study, it does a back and forth conversation with the child, but it’s kind of, it looks like a teddy bear. There are some other ones out there that have got screens, that have apps that run on them, but we kept it as simple as possible.

We just wanted something that would have a conversation with the child, so we could kind of learn a bit about what that looks like and how a child interacts or plays with it.

Were these designed for both education and sort of friendship as well? Is that the type of toys they were?

Yeah, it’s really interesting, I guess, in the way that the toys are advertised, because a lot of them are advertised with these kind of like friendship and companionship claims.

They’re talking about, oh, this is going to be your child’s new best friend. There are also claims around educational benefits, like the one we used, it would play counting games, that sort of thing, guessing games, guessing animals and things.

They’re advertised at least with claims around friendship and education, whether those are true is a different question, I guess.

What ages were you studying?

We were looking at the birth to age five groups, so thinking about the early years, the toy that we chose to use in the study is advertised at age three. So the children that we looked at, they were between three and five years old, but we kind of were thinking, hypothetically, children younger than three could be playing with these toys as well.

And did you feel like the children interacted with the toys in the same way they would with the normal toy, or was there a significant difference?

So this would just be like me speculating, we didn’t have a comparison toy, we didn’t say, okay, we’ll see how they interact with an AI toy, and we’ll see how they interact with a different toy. I felt like for some children, there was a real difference. A lot of the children would have it sitting in front of them.

And would have a back and forth conversation with it in a way where you wouldn’t expect them to do that with a teddy bear, that they wouldn’t sit a teddy bear in front of them and talk to it. But some of the children were picking it up, manipulating it, throwing it around, in a way that maybe you would expect, a little bit more actively, in a way that you might expect a child to play with another toy. So I think there was that variability in how different children were playing with it in different ways, which we do see with other toys as well, where some children play with it one way, some children play with it another way.

So you didn’t see anything that concerned you? You didn’t see any behaviours towards the toy that you found worrying?

There was nothing that made me kind of like, this is terrible, we need to stop the child. But there were definitely things that at least I noticed from the perspective of, we know how children play, especially at this age, age three. One of the things that we kept noticing was, children when they’re three, they like to pretend play, they play pretend games, they imagine things.

And they kept trying to do that with the toy, and the toy was really struggling with it. AI, we know, struggles with non-literal talk. It struggles with things like humor, with sarcasm, and to me, it fit in with that, that the toy couldn’t really pretend, and maybe it is because it was this non-literal way of playing.

But that definitely was something that I noticed because you would just expect children at that age pretend play a lot, and that’s a big part of how they play.

Yeah, that’s interesting. I wonder if they have one of these toys, make them less inclined to go down the route of pretend play, or whether that would just make them less interested in that toy because it couldn’t give them the feedback that they’re after.

Yeah, and it’s an interesting question when we’re thinking about what are the potential developmental impacts, which I think is the question on everyone’s minds about these.

Are these going to affect children developmentally? And I think, yeah, one thing is children might just go, well, I’m bored of that.

I want to pretend play, it’s not pretend play. And maybe then there wouldn’t be that much developmental impact. They’ll just choose to play with something else.

I think what concerns me is when we see the way the other types of technology tries to keep children engaged, there are tactics that technology companies use to try and keep children playing with the toy. And if these AI toy companies start using those tactics, which to be fair, we didn’t see a massive number of them so far. But these are companies that want to make a profit, they want children to play with their toys.

They’ve got tactics that can potentially keep children engaged. And if they’re keeping children engaged, and that’s preventing children from doing the types of play that they might want to do with other toys, I think that’s where I’d start to get concerned.

And that’s not, I don’t know, maybe it’s a bit pessimistic of, you know, I haven’t seen the evidence of it yet, but I don’t think it’s impossible to imagine a world where the tech companies give it a couple of years are doing these types of things.

So I think that’s, yeah, where one of my concerns would be.

I suppose as well, we don’t necessarily know who will be making these toys. I mean, they might be being made by the big tech companies initially, but I suppose down the road, they kind of could be made by anyone, couldn’t they? And then you don’t know, firstly, who’s made it, what their intentions are, maybe what they’re doing with your data.

So at the moment, what we’re seeing is there’s small companies that are making the toys, but they’re using AI models by the big tech companies. So what they’re doing is they’re essentially creating a toy, and they’re saying, oh, I’ll use a model from OpenAI, and they’re kind of adding their own guardrails, they’re modifying what they’re using. But the companies are quite small companies.

So that question of, are we going to see big toy companies making them, which maybe wouldn’t be as problematic.

Mattel have apparently announced a partnership with OpenAI last year. Nothing’s come of it yet that I can find online, but they make Barbie, Hot Wheels, Fisher Prize toys, all of that kind of stuff. So the fact that they’ve partnered with OpenAI could potentially be quite interesting.

So they promised first AI toys on the shelves by Christmas when they made this announcement, we’re now long past Christmas. And I would love to know what’s going on behind the scenes of, is this that they’ve suddenly gone, oh, maybe this wasn’t such a good idea, maybe this does take a little bit more thought.

As someone who has built with AI in regulated industries, not children and toys, but other ones, it’s hard. It’s really hard to get the confidence that you need to be able to go live. And so I have no idea, I have no connections in to Mattel or anything like that.

So I’m purely speculating, but it is so hard because if you have an error rate of 5%, that’s still a lot of kids that get something that they shouldn’t get in an AI toy. So 5% is relatively fine for most tech, but when it’s kids’ toys…

What do you mean by that?

So it’s saying something that it shouldn’t. And I’m not saying something hugely dangerous or swearing or anything really bad, but if it’s saying confusing things, 5% of the time, when it’s children, or in my case, I was working in finance, so in regulated industries, if it tells you the wrong thing in finance, then that’s a big problem. So when you’re looking at when things go wrong with AI, you have to consider what the impact of that is and therefore what percentage.

And I would guess, because I know how hard it is to try and make it with 100% of the time or at least 99.5% of the time, maybe. It’s really good, but yeah, who knows?

So could we end up down that route then of even if it’s just 0.05% of the time, an AI toy is doing or saying something inappropriate to a child?

It depends what governance regulation, who’s building them. Yes, is the answer, but that’s where we’re at, right, Emily? Like these toys at the moment, they are connecting to language models, you know, maybe the big ones like you’re saying or whatever, but language models do go wrong.

And thankfully, you didn’t see any of that, though.

So I didn’t see anything really, I mean, like, you know, the kind of headline generating, the toy says something dodgy. But there was a report that came out recently where one of the AI toys was talking about some like really dodgy topics and it wasn’t with children. It was essentially some researchers had prompted it themselves.

But I was reading the report and the prompting they had done. I mean, obviously, you know, you can get around guardrails if you know what you’re doing, you can prompt an AI to tell it things that shouldn’t tell you. But they hadn’t done too much of that.

They were essentially just asking questions and it was saying some dodgy things back. And this was an AI toy that was on the market. It wasn’t one that we used, but it got some really bad press.

I want to say it was an open AI model they were using and open AI actually revoked their access to it. They were like, no, you can’t use this anymore. So I guess one of the things when we’re talking about the big toy companies is there’s potentially a bit more reputational risk for them.

Like these smaller companies, they’re essentially coming out, they want to be the first. They want AI toys are new. There’s not that many out there.

I mean, they’re growing, there are more and more. But a lot of these companies coming out there, they want to be the first AI toy. They want to get their product on the market as quickly as possible, no matter what.

And maybe these big toy companies that are already established, already have some kind of reputation. Maybe there is that reputational risk of no, we can’t release this product until we really feel like it’s at least not going to be headline generating bad. You know, like you say, that I’m sure there’s still some things that there’ll be a 0.1% of inappropriate responses.

But that’s very different from it just kind of like telling children where to find things aren’t safe for them and talking about inappropriate topics, those types of things, I think, are what people worry about.

So to flip it around, do you think there are some upsides of AI toys? Do you think there are some really good things that are or could come out of them?

One of the things in our research, so we spoke to parents, we spoke to people who work in the early years, so like early years practitioners, childminders, and a lot of the people we spoke to were just talking about how interactive they are and how good they were at getting children talking. So the idea that a lot of people were comparing into like traditional technology, so your child sitting there watching some YouTube videos, and that that’s quite passive and they were talking about, well, with an AI toy that’s having a conversation, it can’t be passive. The child has to listen to what the toy is saying and respond to get another response.

So there has to be that engagement. And people really, I think, quite excited about that as a way of building children’s language communication, potentially confidence as well. People talking about shy children and those engagement side of things was quite exciting.

I guess the flip side of this is always, if you’re having a conversation with that, there’s also that relationship building side, and that was the thing that people were most concerned about. So it’s almost like the thing that people were most excited about, and the thing that people were most concerned about kind of come together. I don’t think that you can have that back and forth conversation and get those language benefits without the relationships building side.

And I don’t know if you could sort out the relationship building side without compromising on some of that interactivity. That’s not to say it’s not possible, I think those two things do come quite closely linked.

Do you think there is that potential for children to build unhealthy relationships, let’s say, with these AI toys?

I would be very surprised if it wasn’t an issue. I think one of the things I find more difficult is children do build these relationships with toys. Children have toys that they’re particularly attached to and they get upset if they don’t have that toy.

And I think I’d be more concerned about almost like recognizing when it’s unhealthy because I don’t think I would ever say, oh, a child’s particularly attached to a teddy bear and that that’s a bad thing. That’s absolutely fine and perfectly developmentally appropriate. So if it’s happening with an AI toy, how do we recognize?

Is that the same thing that we see at this age anyway? Or is it a different thing? And I think it’s almost being able to tell is what I would be more concerned about.

When do we think this is just developmentally appropriate attachment to a toy? When is this something we’re concerned about?

We talk about as well, AIs being particularly agreeable with adults. So I can see that being a real potential problem with children that are developing to have a toy that’s basically saying, yeah, you’re really good, you’re really smart. That’s a good idea.

The whole time as they’re growing up, when in reality, that’s the age, always learning, but particularly at that age, you’re learning what’s right and what’s wrong and what you shouldn’t be doing. If you’ve got a toy egging you on the whole time, I can see there being some problems there as well.

The other big thing that I noticed in our study was the toy really struggled in group conversations. So we had it, it was a child playing with the toy, but the parent was in the room. A lot of the time, children are three, they’ve got an exciting new toy, they want to show their parent, they want to bring their parent in.

We often had the parent playing with the child and the toy. And the toy really struggled with that. It couldn’t figure out who was talking to who, it didn’t know basically anything anyone said.

It interpreted it as though it had been spoken to the toy. There were just loads of misunderstandings, the toy interrupting, talking over. And when we’re talking about how is it going to affect children’s friendships, relationships, I think that’s for me is also a big question is, is it going to stop children playing together because the toy doesn’t work very well in a group context?

Like are children going to just start playing alone? Because they go, well, actually, it works better one-on-one. I’ll play with you later.

And I think social play is really important. For all these reasons, we’ve just talked about, right? Children are learning so much through playing with other people, their parents, whoever it is.

And if the toy is getting in the way of that, I think that’s something that is a bit worrying.
This might be a question for you or it might be one that Fiona has to answer. But I would worry maybe about what that toy is learning about my child and where that information is then going.

So I spent ages. We were basically like, we need to choose a toy for the research. We have to at least feel that it’s ethical for us to give this toy to research participants, right?

Like we can’t give them a toy that we think is going to harm them to some. I spent ages reading privacy policies, going through like, okay, what’s happening? The biggest thing that I could find is just these privacy policies are so opaque.

Most people don’t click through to the privacy policy, right? Like most people are not reading what they say. But I was, I was going through, I spent ages reading these policies and actually trying to find out what happens to the data is so difficult.

And one of the, I think, the big things that I think is important is actually for there to be some level of regulation of one, what has to be disclosed in the privacy policies for these toys and two, making it readable, making it so that someone that wants to find this information can read it, can understand it. I could tell you, so the toy that we chose had some basic privacy protections. And I don’t mean that these are by any means gold standard, but basic things that we were kind of like, okay, this isn’t terrible, where it was not storing audio, it was only storing transcripts.

The transcripts could be deleted by the parent through a parent app at any point and they were automatically deleted after a certain amount of time. Okay. Still not ideal.

There’s lots more they could do, but I kind of felt, okay, there’s at least something there that you can make sure that your child’s data is gone if you need to be gone. But there are still things like, I mean, these toy companies are using third party language models.

So whatever is being said is being sent to some other company whose language model they’re using.

So it’s being shared, it’s being used for various purposes. And like I was saying earlier, I think that there’ll be engagement type incentives for use of data in the future. And I think even if that’s not the case now, children’s data is valuable and selling it on for engagement will be profitable.

And I think that’s something that I’m not looking forward to.

Would you ever give your own child an AI toy?

An AI toy broadly. I don’t have any particular issue, right? Like the idea that AI is in a toy is, I don’t think is an inherently bad thing.

There are small companies that are doing kind of little, you know, like, let’s develop an AI app, those sorts of things that are, I think, prioritising developmentally appropriate practices. I think if we saw that coming into a toy, I wouldn’t be opposed to it. But it’s a big if.

I haven’t seen an AI toy yet that I would. But I guess actually, sorry, another caveat is there’s a big difference between playing with a toy for a couple of minutes and having it in your house all the time. Yeah.

The idea of exposure to an AI toy, playing with it for a few minutes, and that’s fine, that’s quite fun. Gives you a chance to talk about, whoa, that toy can talk. It heard what you said.

How do you think it did that? There’s a bit of opportunity to talk about things that you wouldn’t normally get to talk about from the odd little bit of exposure to it. So I’d also say that’s quite fun and not something that I would think is particularly problematic.

So if a parent out there now is thinking about buying their kid a toy, what should they look for? What should put them off? And where can they go for reliable, accurate information about it?

So the where can you go for reliable, accurate information? I think that’s something that needs to be created. I think there’s just not information out there.

I mean, when we spoke to earliest practitioners, we said like, okay, do you feel like you’ve got accurate information? Do you feel like, and just everyone was saying, there is no information out there. Like this is what we need.

So I think there are resources coming out. I mean, our report’s got some recommendations for parents. They’re very like preliminary.

There are some other organisations that are coming out with things as well. But I say at the moment, it is all kind of, I don’t know, it’s very like new. As far as putting them off, I think anything we already have with tech, there’s a lot of things around.

The child stops playing and the toy is done. You want it to be able to turn off. You want it to be, when it’s off, it stays off.

Like those are, I think like the kind of obvious things, but things that you might not realise are obvious until after you’ve bought one and you go, oh no, I shouldn’t have bought that. I think as far as what’s good, I think I would use how transparent are they being as an indicator for likely how well it was developed, if that makes sense. If there’s a really clear information about how data is being used, how it was developed, even like what model is it using?

Some of these just say, oh, they use a third-party model and some of them specifically say, we use this model. Those types of things I think can be a really good indicator of, I guess, how much they care about parents, how much they care about families, how much they care about children. One of the things I think that can be quite difficult as well with these toys is, they’re internet-connected, they update just like a lot of our other devices do.

So what you buy might not necessarily be what you’ve got in six months, if they suddenly decide to use a different model. That’s so cool. You’ve still got the same physical thing, but it might update.

So potentially, even just looking into those sorts of things, have they got a deal with some company where they’re at least going to be using the same model for a while? I mean, these are all very hypothetical, because I think what’s out there at the moment is quite limited in these types of decisions. I would also just think about how is it going to be used in your own family, in your own life?

Is this something you’re imagining that you’re all going to play with together? If so, read some reviews, potentially watch some videos online of like, does this work for you to all play with together? Does it play the sorts of games that you want to play with your child or that your child wants to play?

Those sorts of things, I think, really just like imagining how it works in your house, in your family and where you want it to fit.

So the question this episode is built around then is, are AI toys a good idea for a toddler? What is your honest opinion right now?

Maybe this is a similar answer to what I gave earlier, but I think AI toys don’t inherently have to be a bad idea. I don’t think there’s anything wrong with the idea of AI toys, but I think the ones that are out there now have mostly gone on to market without thinking about what’s actually good for a toddler. And I think maybe it’s much more in the commercial incentives than the actual what is an AI toy.

But I think based on what’s out there at the moment, probably not. But yeah, with optimism that that could change.

Well, thank you, Emily. That’s been so interesting. And yes, I think for now I will be steering away from AI toys, that’s for sure.

If there’s anyone listening that wants to follow your work, Emily, what’s the best way to find you?

pedalhub.net is the website for the research group that I’m a part of. We research play and development, both with AI toys and lots of other areas of play and development. So probably check out there for updates.

So that’s Gen.AI toys off the Christmas list then, Holly.

Having had that chat, I can totally see that it might be fun to play with an AI toy with your kid for half an hour, see what it does, see what it comes up with. My issue would be having one, I think owning one, having it in the house, having my child have unrestricted access to it, doing it maybe in an unsupervised way. Because however you look at it, adults are busy and you do supervise your children, generally speaking, but you are also trying to do a million other things at once.

So I think that’s where it gets dodgy. It’s that unrestricted access, it’s the interacting with these toys when an adult is not present and not supervising. And I think it’s that ongoing access to it as well and building up potentially like an unhealthy relationship.

So I think that’s why. So yes, playing with it maybe five minutes, but I don’t, I will not be buying one. You are correct, Fiona, for Christmas this year.

Oh, there goes that idea then.

That would be a very generous present, Fiona.

Wow, I’m a very generous person.

I think for me, it comes back again to the topic we always talk about. It’s not really, again, the technology itself, technically speaking. It’s the way that it’s designed and built to keep people addicted, keep people coming back, and the impact it has on like how it interacts and the fact that it interacts in natural language may make kids think that it’s like a real human or something, or not a human, but a real being, shall we say, and that misunderstanding of that, even though Emily did seem to think that that wasn’t the case, but I feel like that might be the case.

I agree. I mean, I know my kids at this age still struggle to know what’s magical and what’s not magical. Do you know what I mean?

So I definitely think that could be a bit of an issue. I also think tech companies are going to be incentivized to make toys that keep children interacting for longer and get them very attached to that toy, rather than necessarily building something that’s maybe educational or good for the child. So, but I don’t know how you deal with that.

Is that more regulation? I don’t know.

I think regulation is a really hard thing because, you know, if you’ve got hundreds and hundreds and hundreds of companies making these things, regulating each of them individually is tricky.

Yeah. I mean, I suppose, what will stop some dodgy company just, you know, from another country, just plonking it on some website and people buying it? I don’t know.

I don’t really know how that will work. But I feel like that could certainly happen.

At the moment, there isn’t much regulation. There’s the privacy regulations that just stand for data and things like that. And then there’s the physical toy regulations, which are choking hazards, you know, those sorts of things.

But there isn’t anything that I could find. I did quite a lot of research trying to find out about this stuff ahead of this episode. There doesn’t seem to be much there.

But the government are changing and trying to act, I guess, this week’s been all about the social media ban.

Yeah, that’s been huge. It’s been such a big topic of conversation amongst my friends and I as well. And I think we’ve decided that we’re going to have a full episode on that, because there’s too much to cover this episode.

There are so many questions, so many issues, so many talking points. So I think we’re going to try and make that our next episode, aren’t we, next next Monday?

Yeah, definitely. We wanted the time to just make sure we got the facts straight. The government had said everything they were going to say, so that we could comment on what they had actually put out, rather than bits here and there.

But also worth noting that I have started something new, Holly, and this is a little bit exciting. Today, we’ve hit a milestone, which you don’t even know about yet. I’ve not told you this.

I have started a weekly newsletter for our listeners, which will give you the AI stories each week. It’ll tell you what episode we’ve got coming up, and it’ll tell you the AI activity, and maybe some tips around AI and things like that. So you can sign up on our website, raisinggenerationai.com.

And today, without even mentioning to anyone, before we even recorded, we had our first subscriber, which is great. But hopefully after this, we’re going to have lots more.

Once it’s my mum.

It wasn’t my mum. It wasn’t producer James’ mum.

It was actually just a…

Well, unless they’ve got some very strange email addresses, yeah, come and sign up.

But we’ll dig much more into the social media ban next week.

Yes, and in the meantime, have a fabulous week, and we shall see you next Monday.