Is AI going to kill all humans in the next decade?

Coming up on today’s episode of Raising Generation AI.

Meta’s AI has been asking some scary questions.

Are your kids smarter than AI?

And does it even matter after this week’s headlines that AI could wipe us all out in the next decade?

Plus your questions. This week, we’re helping you block inappropriate images on your child’s phone.

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, you may be more worried than normal, because apparently, AI is going to kill all humans in the next 10 years.

I mean, what a week for AI news. And I don’t know about you, my feed has been full of this, absolutely full. Not only is it going to kill us all, but I’ve had to have every single person’s opinion on it.

I really want to go into detail about this. But before we get there, can I talk to you about the video I sent you this week from the American mom, Kalie Robins? So for anyone that hasn’t seen this, doesn’t know who she is, and a lot of people won’t, of course, she calls herself a travel mom creator.

She’s based in Utah. She doesn’t have that many followers, a few thousand. But something quite scary happened to her this week.

Take a little listen to this.

“Yesterday, I posted a completely innocent video of me and my daughter singing in the car. And Facebook put a Meta AI question underneath it that said, who’s the child passenger?

First of all, why in the hell is that even a suggested question about a minor?

But I clicked it. And this is where I started getting really uncomfortable. It starts pulling completely separate information for each of my kids.

Their names, birth information, pictures and videos of them. It pulled a newborn picture from my mom’s page from years ago. And then it pulled this.

I deleted this picture years ago. I’m not going to pretend that I know how it works. I know this picture has not been sitting on my page for years.”

And then after that, AI put up another question, which was, where does Kalie Robins live? And she clicked on that and it pinpointed her location. Okay, so off the back of this, she’s removed all images and videos of her kids from social media.

And she’s issued this reaction video as a warning basically to other parents. But oh my goodness, this is terrifying. I mean, there’s so many elements of this that scare me.
One of the main ones is the fact it’s not only dug up things that are on other people’s social media pages, but also things that were deleted, things that as far as this mum knew weren’t even on the Internet anymore. I mean, that’s terrifying.

Yeah, absolutely. When you delete things off the Internet, doesn’t actually always mean that they’re gone forever. You’ve got to go through.

So where are they? It depends. I mean, even when we’re talking about AI chats, the companies keep them for 30 days, even if you have private chats set up.

So often they keep them for legal reasons and things like that. But it’s the same. You can reach out to companies like Facebook and request permanent deletion and stuff like that.

It’s a bit shocking though because you imagine once something is deleted, especially something you deleted years ago, that it is gone and it just goes to show that it isn’t. I mean, my other question would be, well, I have several questions, but a big one is why on earth is AI doing this? Why is it suggesting that people find out this information that they really don’t need?

I am on your side and I think that this is a mad use of AI. When you build AI products or any products for that matter, one of the processes you should go through is thinking about how could this be used in the worst way possible? That’s how you should ethically approach building new products.

That goes for technical products as well as physical products. Quite often, I’m not sure that companies go through these really in-depth testing and ethical thought processes for everything that they put out there. On top of that, the people that are building these products, I don’t know who has built Meta AI at Facebook as in the individuals.

I don’t know what the team makeup is like and stuff like that. But unless you have representatives from all walks of life, these things don’t get thought about because mum in Utah, having her location pinpointed is an awful thing to happen. But for a lot of society, they’re like, well, people can know where I am.

There’s no threat there. There’s no concern. There’s no reason to be worried about that.
So it doesn’t come up as a risk. Do you see what I mean? So it depends on the team that are there thinking about it.

I think most people would be appalled at the idea that strangers off the internet could find their location at the click of a button. I think most people would be horrified at that prospect, especially if you’ve got young kids. And again, there’s photos and videos of the kids.

I just think the whole thing is so, oh, it feels so invasive to me. Is there anything we can do to protect ourselves from this sort of thing happening? I mean, even if you’re not posting pictures of your kids online, as many people don’t.

If you post yourself, presumably, this might happen to you and someone, for example, could still be able to find out your location. Is there anything we can do to protect ourselves or are we just open to this now?

It’s really difficult and I don’t know how easy would be to completely remove everything about you from the internet. There are things you can search for like, right to be forgotten, which will remove you from places on the internet. But I don’t know if we’ve passed this now by posting so much about our lives for the last 10, 15, 20 years on the internet.

It brings me back to a point I made last week actually on the podcast when we were talking about typing information into AI. There’s nothing right now that suggests that that’s a huge privacy risk. Because your data could be used to train other models, but the chances are it will be so diluted and things like that.

Unless you’ve got incredibly unique names that are linked to incredibly unique names of schools and you’re putting copious amounts of data in there that it manages to surface, I think the risks are relatively low. Well, are low. But who knows what it will look like in 10 years’ time.

I think just taking a super cautious approach to privacy is the way forward really. I think there’ll be things that we can’t control. There’s a lot about, especially in America, we have it less so here in the UK, but in America, they’re using facial recognition cameras a lot more than they are in the UK.

They are in the UK, but extensively in America, and there’s a lot of protests against the capturing of that kind of information. There are some things that we are limited in what it is that we can do, but where we can control it, maybe not putting information on social media. If you are the kind of people that share information on social media, it can be done in a safer way in the way that you can post your holiday pictures after you’ve come back home.

And do the sensible things like that. But in terms of piecing together your life, I don’t know if there’s truly a way of getting out of this.

Well, that’s nice and depressing. And I’ve got another lovely depressing line from Sam Altman. I saw this video on your Instagram this week.

Should we take a little listen to this one?

“You know, like a kid growing up today will never be smarter than AI.”

There we have it. So Sam Altman is the, he’s the founder of OpenAI, isn’t he, Sam Altman?

Yes. And this was at a G20 meeting, which is basically where government ministers from the world’s biggest economies kind of gather and talk about technology and things like this. This particular quote came from an interview with a member of the US government.

And actually the argument Altman was making in the wider point was that countries can’t afford not to adopt AI. So he was very much trying to convince countries to adopt AI and was using this as part of but this is a quote that he has said multiple times. So OpenAI have just launched their own podcast.

Better AI podcasts are available, everyone. But they have just launched their own podcast.

And he repeated this quote or this idea in the first episode of that podcast as well, which was that it’s more that our children are growing up in a world where they won’t know anything other than AI and everything will be smart in the term of technology, smart technology, and it will be expected by that generation that things will just be smart and know about you.

The quote specifically was, a kid growing up today will never be smarter than AI, and that is what has caught the headlines.

Yep. As a parent, it feels quite, almost offensive. I sort of get what he’s saying, but if I was a tech bro, no current plans, I would be trying to bring people along with me and embrace this technology.

What I wouldn’t be saying is, AI is going to make your children obsolete, which is effectively what he’s saying. What’s your take on this? Is AI going to be cleverer than my children, Fiona?

Well, I put it back to you. What makes someone clever? How do you describe what makes someone clever?

Because in many ways, I suppose, computers are already cleverer than us, for some things. A calculator is cleverer than us in some ways.

Oh, yeah, definitely. If you ask producer James about my math skills, he’ll tell you immediately that a calculator is far better. But you’re absolutely right.

The internet already has far more knowledge than I could ever dream to possess. Also a lot more nonsense, believe it or not, than what’s in my brain. I mean, there’s different levels of intelligence or smartness.

I think it’s important to make sure that our children grow up understanding and knowing that value. I think the biggest issue with this headline, as you say, is that our children are reading this, they’re seeing this and they think, well, what is the point? If AI is just going to do everything, what is the point?

I’d imagine if I was a kid hearing that, I just think, well, why am I bothering my GCSEs if what’s the point of putting the effort in?
I can completely see that.

Do I believe that AI will be smarter than our children? I think it will be able to process information faster than our children, but computers can already do that. It does something at the moment which people call reasoning.

I hate using the term reasoning because that makes it sound like it’s having actual thoughts and making decisions, but it will be able to process things and give steps to complete tasks and stuff like that. But where I think AI is terrible is that it feels very formulaic and very robotic. I actually posted on LinkedIn this week about how I think AI has completely ruined my ability to write.

I had to write an article for a magazine this week, and I wasn’t using AI to write this article because I don’t believe I should. I write our newsletter every Friday from myself. If it’s an article, I very much do not use AI.

I was finding I was almost replicating the patterns of AI language when I was writing. I think it makes it sound boring and horrible and robotic. I think what I managed to break out of and get through from this was to try and sound more human.

I think that in itself, it sounds like I’m bigging up myself, which I’m not, but I think that skill is a very human skill that I don’t think AI will be able to have. That’s just one example of where I think, no.

That’s really interesting because I’ve noticed this a lot. Things I read and I think to myself, I reckon that’s AI. But then I wonder if it’s exactly that.

It’s just people have read a lot of AI and are almost internalizing that cadence and that rhythm. Because it is very distinct. It’s that whole, it’s not X, it’s Y, and I don’t know exactly how to explain it.

It’s something you just feel from the writing. It’s very broken up ideas almost, laying this idea out very clearly, and then this idea out very clearly. And there’s just something very specific about it.

And I do think more and more people are starting to write a bit like that in real life. So that is so interesting that you said the same things. I haven’t heard that anywhere else.

I think that’s just one example of where humans will always be better than AI, because AI can’t invoke that feeling of really what happened when you’re describing that story. It won’t automatically relate it back to other times in your life. That kind of building up a whole thing.

Being street smart is the other thing. AI doesn’t have any concept of existence and life. I think being street smart is one of the most valuable smarts that there is.

Being able to understand how to exist in the world. So I think it’s our job to make sure that the kids know that.

Yes. I think you’re exactly right. It seems like, I don’t know, reading body language, isn’t it?
I assume AI will never be very good at that sort of thing. Like you say, it’s making sure kids grow up with enough human interaction and human learning and human teaching that they don’t just adopt these AI ways.

So hopefully all this has been useful today, but let’s be honest, we might not even be alive in 10 years to make use of it.

Yeah, turns out a big new story of the week is that AI potentially is going to kill us all in the next 10 years. So talk us through how this has all come to light, Fiona.

So it all started with Jacob Coxon. He’s an AI researcher that was working on Claude.

He left his job and then posted a Twitter thread about why he left.

In it, he claimed that neither Anthropic, who are the people that make Claude, or his previous employer, OpenAI, who make ChatGPT, are acting responsibly. And then a current senior safety researcher at Anthropic basically quote tweeted Jacob’s thread in agreement, saying that he felt there was a greater than 10% chance that AI will, and I quote, kill all humans in the next decade.

Okay. Firstly, my question is going to be, when people are talking about AI killing everyone, what do they mean by this? Are we talking killer robots or are we talking, I don’t know, interrupting water supplies?

What is it that these researchers think that AI is going to do?

There’s been a few different people saying different things over the past, I don’t know, 10 years because AI has been around for a while, it’s not been so much in the headlines. The idea, yes, weapons and robots used for physical warfare, I guess, for lack of a better word. That is a risk and we talked about this a little bit when we were talking about the humanoid robot games.

Yeah.

Because it did invoke those kind of feelings, especially in the opening ceremony.

The robot armies, yeah.

Exactly. That is a genuine threat. I think there’s been a lot of stories recently about hacking.

Open AI hacked Hugging Face. You might have seen that in the news. We did briefly discuss that as well.

There’s a risk here that the AI becomes, they say super intelligent enough to just do lots of hacking.

But why would AI want to? This is the issue for me. On the one hand, AI, as you tell me very often, is basically just maths.

Yes, it’s very complicated maths, very complicated models, whatever. Why would it want to kill us all? I just can’t draw the line between those things.

Back in 2003, so think about how far back that was. There was an Oxford philosopher, Nick Bostrom, who invented this thought experiment, which you may have heard of called the paperclip maximizer.

I don’t know if that’s something you’ve heard of. Yeah, go on.

The idea behind this is there’s a super-intelligent AI, and it’s told to manufacture as many paperclips as it possibly can. Now, I just want to be super clear. This didn’t actually really exist.

This was a thought experiment that the philosopher came up with. Don’t read into this too much. It’s not a nightmare that is playing out in real life.

It runs out of steel. So it’s tasked with making as many paperclips as it can. It uses all the steel.

And because it’s really focused on that task of making paperclips, it ends up killing all humans and turning their bodies into materials for its factories.

So what you’re saying is potentially AI could kill us all in the next 10 years?

Not quite. A big part of this quote that a lot of people haven’t really focused on. The quote tweet that was there from Evan Hubinger, he’s the guy that’s currently at Anthropic.

So I want to make it clear that Jacob Coxon did the original tweet thread. He has left, he quit and said that he couldn’t work there anymore, whereas Evan Hubinger, who is a lead for safety at Claude, and he’s still there. So he said that he believes Anthropic is trying its best, but they don’t think that they found an alignment for super intelligence and are not on track to.

However, he doesn’t think that the current models are the ones that are the threats. He thinks it’s going to be future models. He thinks the current models are not the threat, if you see what I mean.

But they’re obviously still training future models and trying to make better models. And he thinks they’re getting close to those models being good enough to do something.

That’s interesting that someone who is still at the company is saying that as well.

You work for ITV and if you suddenly came onto Twitter and said that ITV have killer robots working in the studio that are likely to kill people and any contributors over the next 10 years, I think ITV might sack you, regardless of whether it’s true.
It’s not true, just to be clear. Definitely not happening at ITV. Exactly, exactly this.
I find it very, very strange. And that did make me wonder whether there is an element of, and I’ve heard this suggested, that it is marketing.

And I do think that to an extent. So both the people behind Claude and the people behind ChatGPT are hoping to trade publicly on the stock market over the next six months or so. And obviously, if that happens, they want to make their product seem incredibly valuable so that people buy stocks, right?

I don’t know much about the stock market. I’m not going to pretend that I do. But ultimately, the more valuable the product, the more people buy stocks for higher value, et cetera, et cetera.

So it’s in their benefit to say, we have this most incredible product that you should come and buy into basically. And actually, I met with a friend for Coffee just yesterday, and I spoke to her and she was just like, they’re going public? Definitely, I’m going to buy stocks.

And that’s exactly what they want, right? They want to make the world think that they’re so good and they have to get a piece of this. So when they go public, it does well, right?

So I do think there is a part of that. I also think there’s a part that we were just discussing in the last segment where there’s a value in making the product seem sort of slightly detached from what we could ever understand. Do you see what I mean?

And distracting us with this idea of killer robots. Well, we’re here chatting about killer robots on the podcast. We’re not talking about the fact that there’s real lawsuits currently happening in the US where people have taken their own life because chatbots have suggested to them they should, or they’ve committed a crime and things like that.

We’re not talking about the stories because we’re busy chatting about whether AI is going to kill us in the next 10 years. And I think there’s actually a couple of very well-known ethicists, Emily Bender and Timnit Gebru, who are very, very well-known for, this is their view basically, they think that it’s all about distracting. They call it the deflection.

It’s all about distracting the world so that we’re all distracted by this premise of Terminator style wipeout. And we’re not looking at the labor conditions of the people that are having to train the AIs, the data centers that have been built. I mean, I’ve seen protests or petitions this week in Torrington and Devon, where I used to live because people don’t want data centers.

And we’re not talking about that. We’re talking about this massive headline. So I think it’s part of it.

I do think it’s part of it. But there are others. There was a guy called Yann LeCun, who was Meta’s Chief AI Scientist until 2025.

He basically said that because the current systems have no model of the physical world, no persistent memory, no real planning ability, basically they’re not that useful. And there are some really respected people that suggest that it’s all a load of nonsense and it’s all hype. And that’s the difficult bit, is that we have no idea because there are respected people saying that it’s all a load of nonsense.

There’s respected people saying, I think this is very real and everyone in between. I spoke to a friend of mine earlier this week who’s very involved in AI and I’ve not spoken to him in ages. And he used to be very, oh, it’s all nonsense.

Like there’s no AI doom. And he was just, I’ve changed. It’s only been 18 months, maybe, well, maybe not even that, about a year since I spoke to him and he said, actually, I am now, in his words, a full blown doomer.

So you might hear people refer to the term P doom, which is the probability of doom, which is 10%. It’s a known term within the AI industry. What is your P doom?

So what do you think the chances are that AI is going to wipe out the world? So it is a very spoken about thing. And some people think it’s nonsense.

What I do find really annoying is we have been warned about this for years. We’ve all seen Terminator. We’ve all seen what happens when AI gets out of control.

And yet we’ve done it anyway. There’ll be sci-fi writers out there saying, I told you so.

There’s so much value in what AI can do though. It’s made miles of progress in a lot of science. And I was listening to a podcast quite recently about what it’s doing for weather forecasting.

Yeah, yeah, yeah, sure.

The actual technology itself, when it’s used well, is good.

It’s the people.

It can do all the most amazing things in the world. It could cure cancer, but if in 10 years it kills us all, it’s irrelevant because we’re all dead anyway. Do you know what I mean?

I can totally see if it does amazing things and the human race continues to live unexterminated, then that’s wonderful. But I don’t know, a 10% risk for me is too high of a risk if that is indeed the case. Anyway, shall we go on to our last thing for today, which is our parents’ questions?

I have had a question from Sophia in Aldershot who says that she was on a train, got sent an unsolicited, let’s say dodgy photo, and is really worried that this could happen to her daughter, or I suppose her son. What can she do to protect them from that? Is there anything?

Actually, yes, there is, and it’s quite current. So on an iPhone, you can go in, and as an adult, as a child, as anyone, you can go in on your iPhone and go into the settings. And if you scroll down to screen time, in that menu, there is something called communication safety.

Now, when you first open it, you’ll see a little pop-up that gives you some information about communication safety. But ultimately, if you go in there, you can turn this communication safety on. And what that does is it’s an on-device AI that prevents your phone from showing nude or inappropriate images.

Oh, that’s good. Now, it is. The way the iPhone does it specifically is that it does it on device.

So there’s no cloud involved. So if a child takes an image of themselves that they should not have taken, then it wouldn’t then scan that picture in the cloud to establish that it was an inappropriate image. It would do that all on the device, which is the point of it.

And it pops up a little cover pop-up thing that says, this is an inappropriate image or this has been identified as an appropriate image. Are you sure you want to click through? Now, this is what happens more when it’s an adult.

But if you also have a parent-child setup on your iPhone, so when you set up your child’s iPhone, you can connect it to the family iPhone and set it up. You can set up so you can’t bypass it. So on an adult’s phone, you can say, yeah, I’m cool.

I’m happy to see this. And just kind of click on through. It still gives you the warning if you’ve got this setting turned on, but you can then click through it.

With a child’s phone, you as a parent can have a control over that so that they can’t click past that. And they’re given options like, this is an appropriate image, you should speak to someone about this, would you like to message someone? And it can prompt them into messaging you to talk about it and those kinds of things.

So there’s some various things in there. There’s a lot of information if you go into that setting in your iPhone, you can click on it and kind of go through and read about and see what the screenshots look like as well. So yeah, and a very similar thing on Google phones as well.

So you can have a look at those settings.

Oh, that’s really good news. I didn’t know you could do that. I will be going to implement it on my own phone.

Something cheerful to finish up the podcast with. Gosh, that feels like it’s all been a bit heavy this week, doesn’t it?

Yeah, it’s been quite a week. As soon as I saw that news break, I thought, well, there we go. That’s the podcast for this week.

Well, I am very much hoping we’re going to have something a little bit lighter and more enjoyable for next week. Probably not, but hey, that’s why we do the podcast.

And if you are enjoying listening to them, please do leave us a review. It really helps other parents find us and we’ll love you forever.