Original Broadcaster Captioning (Enhanced). Treat it as a searchable index of what was broadcast, not a quotation record.
00:00:00trends, apps, gadgets, algorithms algorithms and internet obsessions
00:00:02obsessions shaping how we live, work,
00:00:04work, think and connect now on BBC
00:00:06BBC News. The interface. Jess, is
00:00:12is there any spooky technology in
00:00:13in your life? Oh yes. I have an Alexa Alexa that loves to randomly wake
00:00:17wake up at around 2:00 in the morning morning going I'm sorry I did not
00:00:21not get that. I don't love that.
00:00:23that. I don't love it either. I've
00:00:26I've got a kitchen scale here. Oh,
00:00:28Oh, and you brought it. You're listen, listen, you heard the beep. So it's
00:00:32it's got a touch
00:00:39Screen. And when you touch it, it beeps. And sometimes it just like like turns itself on and just starts
00:00:43starts beeping like someone is touching touching it. And I can't I cannot cannot figure out why. Like, my only
00:00:46only theory is that it's like the the humidity in my house or something, something, but even that doesn't
00:00:49doesn't really make sense. I don't don't like that. But I do like that that we're bringing props on the
00:00:53the show now. I think that's a good good I think that's. That's a good
00:00:55good trend for us. So there's a reason
00:00:59reason we're feeling spooky. It's It's Halloween and what better way
00:01:02way to celebrate than two stories
00:01:04stories on the interface this week?
00:01:07week? From the uncanny and darkest
00:01:09darkest corners of the world of technology, technology, I'm Thomas Germain. And
00:01:12And I'm Jess Maddox. And this week
00:01:14week we're going to be talking about
00:01:17about 11 ghosts that haunt the AI
00:01:19AI tools that we use every day and
00:01:21and what they tell us about the future future of human creativity. And we'll
00:01:25we'll be covering why some people people believe social media is controlled
00:01:27controlled by magic, something researchers researchers have called algorithmic algorithmic conspirituality. And
00:01:32And so just this week I want to tell
00:01:33tell you a story about some weird
00:01:37weird obsessions that large language language models seem to have both
00:01:40both like, you know, the ones from from anthropic and OpenAI, Google
00:01:43Google and even the Chinese models
00:01:46models like all of them have these
00:01:49these strange themes that they centre
00:01:52centre around when they're answering answering certain kinds of questions.
00:01:55questions. And there's a couple of of people that they can't stop talking
00:01:59talking about, and no one knows exactly
00:02:01exactly who these people are or why
00:02:04why any of this is happening. Well, Well, that sounds creepy. It's kind
00:02:07kind of it's like one of the weirdest weirdest things I've heard about
00:02:10about AI in a while. So I think the the best way to start is to test
00:02:13test this out. So I want you to open
00:02:15open up like any AI tool like, like,
00:02:17like, like open up Claude or whatever. whatever. Claude right here. You
00:02:21You got it open. We got it. I want want you to ask it to tell you a
00:02:25a story. To me. A story. All right,
00:02:28right, let's see what I get here.
00:02:32here. Oh. Here. I got a story called
00:02:37called The Lighthouse Keeper's Last Last Lamp.
00:02:44Last Lamp. The lighthouse keeper's last night. What's the. the. Any particular characters that that stand out in that story? I see
00:02:48see I don't even know how to say
00:02:50say this. It's a Dallas. Oh, DALYS
00:02:55DALYS wow. OK, that's a little. So
00:02:56So the reason we're talking about
00:02:59about this is a study that was done done by some researchers at Cornell,
00:03:02Cornell, which found that when you
00:03:05you ask AI models, any AI model to
00:03:07to tell you a story in vague terms, terms, you can try this yourself
00:03:10yourself if you're listening or watching,
00:03:13watching, they 90% of the time will
00:03:16will mention one of 11 words. The The most common one is lighthouse.
00:03:19lighthouse. They also like talking talking about lighthouse keepers.
00:03:22keepers. And we got both of those.
00:03:24those. We got both of those. And
00:03:26And there's also a couple of names
00:03:29names that they keep bringing up up again and again. So the names
00:03:33names are like Elias Mora, Marcus
00:03:36Marcus and Thomas. Interesting. That's That's I mean it makes sense. You
00:03:40You are the ghost in the machine. machine. I mean, it's time for me
00:03:43me to come clean. The researchers researchers looked at 20,000 stories
00:03:47stories where they just asked the the AI write me a story. It was model
00:03:51model agnostic, no matter which one
00:03:52one they used. And most of the time,
00:03:55time, nine times out of ten they
00:03:57they would reference these words
00:03:59words and names way more frequently
00:04:02frequently than they actually appear
00:04:04appear in contemporary literature. Because these are not not common things. These are not
00:04:08not common names lighthouses. I mean, mean, they're common, but I don't
00:04:12don't think they're like that common common to be this represented in in what the model is going to be
00:04:15be spitting out. Right. And it it's
00:04:18it's really weird if you start to to like think about what's happening happening under the hood, right?
00:04:21right? Because if just one of these these models was doing it's like,
00:04:24like, well, OK, like something weird weird happening, but ostensibly they're
00:04:27they're all using different training
00:04:29training data, right? Like they're they're not sharing, you know, at
00:04:33at least not sharing all of the material material that they use to make these
00:04:36these models. So there's something something else that's happening.
00:04:38happening. I talked to the head researcher
00:04:42researcher of this paper, a guy named named SIL Hamilton. He's a researcher
00:04:44researcher at Cornell. This actually actually came out, you know, the
00:04:47the paper was released like a preprint preprint over the summer, but it's
00:04:51it's now getting released at a conference conference that's coming up. And
00:04:55And he had a couple of theories about
00:04:57about what's happening here. So there's
00:05:00there's two parts of making a new
00:05:03new AI model or a large language
00:05:05language model at least. And this
00:05:07this is a gross oversimplification, oversimplification, but there's pre-training
00:05:10pre-training and there's post-training. post-training. OK, so pre-training pre-training is when you like scrape
00:05:14scrape all of the text on the internet. internet. Sure. Or some large body
00:05:18body of text. And then you have the
00:05:22the model like analyse all of this this text and like notice statistical
00:05:25statistical patterns like between between the words. And the second
00:05:29second part is post-training where
00:05:30where they kind of steer the tool
00:05:34tool to give what you know, are good
00:05:35good and useful answers to questions
00:05:38questions they like take a body of
00:05:40of questions and answers and they
00:05:44they have they like tell the AI like like these are good answers to these
00:05:47these questions and these are bad
00:05:48bad answers. There's this one. Post-training
00:05:52Post-training data set related to to fictional stories that Cornell
00:05:55Cornell actually created that is
00:05:57is kind of widely used in that data
00:06:00data set, though lighthouses and and Elias don't show up anywhere
00:06:04anywhere near as often as they do do when you like, ask them to tell
00:06:08tell you a story. Interesting. So
00:06:10So the theory that they have is like
00:06:13like the way that they built this this data set is they just had a
00:06:16a bunch of people come in and like, like, ask the AI to tell them a story.
00:06:19story. And it was like up to them them what the story was about. And
00:06:22And apparently a lot of this was was like not safe for work. And some
00:06:25some of it was about like copyrighted
00:06:27copyrighted characters. Of course,
00:06:29course, intellectual property. So
00:06:32So what they think may have happened
00:06:34happened is in this particular data
00:06:37data set there became an overrepresentation overrepresentation of these ideas
00:06:41ideas because a lot of the other other stuff was like pulled out and
00:06:44and then the lighthouse somehow became
00:06:48became like a go to good topic. And
00:06:51And you can see this out in the world
00:06:54world now if you just look at short
00:06:57short fiction out, like on the internet.
00:06:59internet. Yeah, there is now an overrepresentation
00:07:02overrepresentation of people talking talking about Elias and lighthouses
00:07:05lighthouses and lighthouse keepers keepers and clockmakers is another
00:07:08another one. It is now seeped into
00:07:12into our culture because people are
00:07:15are using AI tools to do creative
00:07:17creative writing. And this is now
00:07:19now like feeding back out into the
00:07:21the real world. And now, because because all the AI models are also
00:07:25also trained on, you know, their their own output and how people are
00:07:28are using these tools that now it's it's like they put this stuff out
00:07:31out and now it's like also being
00:07:33being fed back into the machines. It's also been a problem
00:07:37problem in online fanfiction spaces spaces where people write stories stories about their favourite movies
00:07:40movies or TV shows or books, and
00:07:43and people are getting these same same generative AI stories over and
00:07:47and over again in these self-publishing self-publishing platforms in fanfiction.
00:07:51fanfiction. Because even if it's it's not necessarily Elias in the the Lighthouse, you're still going
00:07:55going to get the same type of things
00:07:57things over and over again. But we
00:07:59we do see a pretty big influx of of lighthouses as well. The output
00:08:03output of human language is now converging
00:08:06converging towards AI language because because we see it so often. Yes.
00:08:09Yes. And I'll tell you what we just
00:08:11just finished mid-terms at my university university where I teach and I really
00:08:14really don't like spending my time time trying to hunt down AI on writing.
00:08:17writing. So I actively try to ignore ignore it. Most of the time. But
00:08:21But I read in one of my classes I I read the first essay. I was like,
00:08:25like, that's pretty good. Then I I read the second essay. OK, good.
00:08:29good. So some similar ideas. To essay
00:08:32essay number three. OK, I'm starting starting to think what happened here.
00:08:36here. So I do what I do often I go go to Claude or ChatGPT and put my
00:08:39my own assignment prompt into one
00:08:42one of these models and sure enough,
00:08:44enough, I get back pretty much what what it's been submitted surprise,
00:08:48surprise, surprise surprise, surprise. surprise. And I just, you know, I
00:08:51I look at all of this and I just
00:08:53just think everything is kind of of regressing to the mean, just kind
00:08:57kind of moving to the centre, moving moving to everyone, talking the same,
00:09:00same, writing the same. Whether it
00:09:02it is in your college class or it
00:09:05it is Elias in the lighthouse that that you are publishing on a self-publishing
00:09:08self-publishing platform, that we we are just moving to the centre centre becoming similar and kind
00:09:11kind of losing the sense of quirkiness quirkiness or difference or something something that I don't know, that
00:09:15that gives writing some flair. And
00:09:18And maybe that's a bummer. Or maybe
00:09:20maybe it's like a sign of a coming coming catastrophe. A while back,
00:09:23back, I talked to this neurologist
00:09:25neurologist named Adam Green at Georgetown
00:09:29Georgetown University, and he did did a study with a team of other other researchers where they took,
00:09:33took, I think it was like 300,000 300,000 college admissions essays,
00:09:36essays, you know, like the essay essay you write to try and get into into school. And they had a data
00:09:39data set where they knew like these these were all human written and
00:09:42and they had a data set where like like these were all written by AI
00:09:46AI or they believed that they were. were. And they compared the two.
00:09:49two. And they found that when regular
00:09:52regular people read the AI written
00:09:55written essays, they seemed to be be better written and they rated
00:09:59rated them to be more creative. People People liked the writing better.
00:10:03better. Apparently, really. But when
00:10:04when they looked at like the total
00:10:07total body of ideas within all that
00:10:10that writing, there were fewer original original thoughts. And like simplifying
00:10:13simplifying this. But there were were their original ideas in the
00:10:17the AI writing. Even though people
00:10:18people thought that it seemed more
00:10:21more creative and some researchers researchers are very concerned that
00:10:25that as AI becomes a bigger and bigger bigger part of our lives and we rely
00:10:29rely on it more and more for different
00:10:31different kinds of cognitive processes,
00:10:35processes, that as a society we might
00:10:38might start having fewer ideas because because we're not flexing those creative
00:10:41creative muscles. And I see this this happening more broadly in the
00:10:45the world. I've done a lot of research research into book influencers and
00:10:49and online book communities, and and one of the common complaints
00:10:52complaints is books are just being
00:10:54being written for tropes. Now, does
00:10:56does it have a slow burn romance? romance? Does it have Grumpy Sunshine?
00:11:00Sunshine? Does it have shadow daddies? daddies? Like those are the tropes
00:11:02tropes people want shadow daddy,
00:11:04daddy, shadow shadow shadow shadow
00:11:07shadow daddy is always the brooding
00:11:10brooding male love interest who of of course has shadow magic. They They have shadow magic every time.
00:11:13time. Of course. Right? Of course
00:11:15course he's. Gotta have shadow magic. The interesting thing here
00:11:19here is these are solvable problems. problems. It's not clear that the the companies have an incentive to
00:11:22to fix this, right? Because at least least for now, they're not making making a tonne of money on creative
00:11:26creative writing. There was a similar similar problem in Claude Code, which
00:11:29which is like the tool that engineers
00:11:32engineers use to write code where where it would overuse the word load
00:11:35load bearing. And people got really really annoyed and like, that's a
00:11:38a money maker for anthropic. So they
00:11:40they like made it stop doing that. that. Yeah. Like the creative writers
00:11:44writers of the world aren't paying
00:11:45paying as much as computer engineers
00:11:48engineers are. So there may not be
00:11:50be a reason for them to work on this this unless it becomes a reputational
00:11:54reputational thing, unless like it it becomes a trope and everyone makes makes fun of it and it like looks
00:11:58looks bad for them. But even if they
00:12:01they do solve this problem, the way
00:12:03way that they would deal with it
00:12:06it would be essentially instructing instructing the model like, look,
00:12:08look, when you respond to this kind
00:12:10kind of prompt, don't bring up lighthouses,
00:12:12lighthouses, don't talk about Elias
00:12:14Elias in your final answer and what
00:12:17what that would mean is that we wouldn't
00:12:20wouldn't be seeing this stuff in
00:12:22in the outputs that we get. But in in like the thought process, for
00:12:25for lack of a better word, like the the internal, you know, machinations
00:12:29machinations of the machine, these these ideas would still be in there there and the model would be like,
00:12:32like, oh, I'm not supposed to talk talk about that. So these ghosts
00:12:35ghosts would still be there in the the machine for years to come, and
00:12:38and we just wouldn't see it. And And that is how you create the perfect
00:12:42perfect internet urban legend with with if you can't talk about lighthouses,
00:12:45lighthouses, why can't the AI talk
00:12:46talk about lighthouses? Stay spooky.
00:12:52spooky. All right, Tom, speaking
00:12:55speaking of spooky, have you ever ever come across a video on TikTok
00:12:58TikTok that said something like, like, if you're seeing this, it was
00:13:01was meant for you. I have yeah, there's there's like all these videos where
00:13:05where it's like, oh, this is gonna
00:13:07gonna find you on like Tuesday, October
00:13:10October 17th or whatever it is. Yeah.
00:13:12Yeah. They also sometimes say things
00:13:14things like no hashtag, no captions, captions, no audio, just a message
00:13:18message that was meant to find you.
00:13:19you. Right? And turns out this is
00:13:23is a pretty popular way of people people understanding algorithms. People are buying into
00:13:27into it. People buying into it, buying buying into the idea that there is
00:13:30is some mystical force driving algorithms
00:13:32algorithms to deliver you a message. message. Exactly when you are supposed
00:13:35supposed to see it. Right. So if if you're seeing this, it is meant meant for you. It is a message that
00:13:39that is going to have some profound profound revelation in your life
00:13:43life and make you aware of something something you didn't know before.
00:13:46before. It's going to make you really really interested in manifesting, manifesting, like maybe this is going
00:13:50going to manifest success for you. you. And researchers have a name
00:13:54name for this actually, researchers researchers call this algorithmic
00:13:58algorithmic conspirituality. And And it's what happens when you think
00:14:00think your algorithm knows you better better than you know yourself. And
00:14:04And it typically involves, again, again, the delivery of that message
00:14:08message that I didn't know this about
00:14:10about myself. I am now seeing this
00:14:12this video that says, this is a sign.
00:14:15sign. Stop scrolling. Of course, course, I think we've all heard people
00:14:18people say, my algorithm does know know me better than I know myself.
00:14:21myself. Sometimes you do see pretty pretty eerie content that's like, like, you know what? I do agree with
00:14:25with that. And I didn't know I agreed
00:14:27agreed with it before, so some other
00:14:29other examples of this. OK, you will
00:14:32will have an hourglass body shape shape just like and share to lock
00:14:36lock it in, right? So meaning this
00:14:38this will manifest for you if only only you, you know, engage with.
00:14:42with. Interacting with this video video will help you do magic. Exactly,
00:14:45Exactly, exactly. If you're seeing seeing this, it's meant for you.
00:14:49you. Book that first session for
00:14:51for your tattoo. No hashtags, no
00:14:54no caption, no audio. If you're seeing
00:14:56seeing this, then you're either neurodivergent, neurodivergent, autistic or disabled.
00:14:59disabled. Oh yeah. Well, we were were talking to our producer before
00:15:03before the show and she had an example example like this where she started
00:15:07started seeing a lot of material material about ADHD and she was like
00:15:10like talking to a friend. She was was like starting to wonder, am I
00:15:14I seeing this content because I have have ADHD? Like, is there something
00:15:18something that it has picked up on? on? And I think if we take a step
00:15:21step back and look at kind of how how the researchers arrived at this
00:15:23this term, we can kind of see maybe maybe why this is happening right
00:15:26right now. So about 15 years ago,
00:15:29ago, researchers coined the term term conspirituality to talk about
00:15:32about how conspiracy theories and
00:15:34and New Age spirituality were kind
00:15:37kind of intertwining. They argued
00:15:39argued that these two seemingly different different things were actually perpetuating
00:15:43perpetuating the same narrative, narrative, which is that nothing
00:15:45nothing is random. Everything happens happens for a reason. Everything Everything that kind of crosses your
00:15:49your life is meant for you. And again, again, nothing happens by accident.
00:15:53accident. And then a couple of years years ago, a team of researchers researchers at Penn State University
00:15:57University and the University of
00:15:58of Alabama coined the term algorithmic algorithmic conspirituality, finding
00:16:02finding the evidence that said this this idea nothing is random. Nothing Nothing is by accident. Everything
00:16:05Everything is connected and meant meant for you has started to be a
00:16:09a way people talk about and understand understand algorithms. Algorithms
00:16:13Algorithms are viewed often as rational
00:16:15rational decision makers. We know know that's not necessarily true.
00:16:18true. Safiya noble has a great book, book, Algorithms of Oppression, talking
00:16:22talking about all the ways discrimination discrimination and power imbalances imbalances get encoded into algorithms.
00:16:26algorithms. But people believe algorithms algorithms are these rational decision
00:16:29decision makers. With this objective
00:16:31objective almost sublime quality quality about them that helps people
00:16:35people understand their lives. And And I think with the examples we we talked about, there's two interesting
00:16:39interesting things happening here. here. One, content creators see themselves
00:16:43themselves as vessels through which
00:16:45which algorithmic magic can happen. It's certainly how I think
00:16:48think about myself. Of course, that that you are a vessel for magic, magic, Tom. That is exactly what
00:16:52what that is. Yes, but it also feels feels a little bit like engagement
00:16:56engagement bait, if we're being honest honest as well, right? It's just just like trying to get people to,
00:16:59to, you know, stop and pay attention attention to video. But that's also
00:17:03also like, that's a better way to to understand what's happening. So
00:17:06So like this example with our producer
00:17:08producer in the ADHD, right? It's
00:17:10It's not showing her that content
00:17:14content because it knows that she she has ADHD. It's showing her that that content because it has picked
00:17:18picked up on the fact that when she she sees stuff related to ADHD, she
00:17:21she pays more attention, she doesn't doesn't scroll or maybe she shares
00:17:25shares it with that friend that she's she's talking to or these videos videos that are like, I saw one the
00:17:29the other day where it was, you know,
00:17:31know, a particular date, October
00:17:34October 17th or whatever. And the the video was like, this video will
00:17:37will find you on October 17th. But But it was like posted in the past
00:17:40past and it did. Oh, my God, that's that's incredible. Until you consider
00:17:44consider that people who saw that that on October 14th and 15th and
00:17:48and 16th would scroll past it, but but then on the 17th they engage
00:17:51engage more and then it gets recommended recommended more often. Exactly.
00:17:55Exactly. And researchers have looked
00:17:56looked into this how I mean, we've we've been off the chronological
00:18:00chronological social media feed in in various stages for about 15 years
00:18:03years or so now. Right. Social media media used to be this chronological
00:18:06chronological feed or what researchers researchers called real time social
00:18:09social media, meaning you were seeing seeing things as your friends posted
00:18:12posted them. It was very linear with with the algorithmic feed, we enter
00:18:16enter what researchers call right right time social media, where the the algorithm will, based on your
00:18:20your interest, will serve you content content when it is the right time
00:18:23time for you to see it. And we know
00:18:27know this is not anything magical, magical, it is just how algorithms
00:18:31algorithms process your data and and your interests. For instance,
00:18:34instance, I just received one one one of these videos last week or
00:18:37or so that was like, if you're seeing seeing this, you hate O'Hare airport.
00:18:41airport. And you know what? I do do hate O'Hare airport. Actually
00:18:44Actually in Chicago. Yeah, yeah.
00:18:47yeah. But this wasn't magic. I have have been travelling a lot, including
00:18:50including a trip to O'Hare airport
00:18:52airport at one point. So I engage
00:18:55engage with travel content. Sometimes. Sometimes. I love watching plane plane content on TikTok. Why? That's
00:18:58That's another story. And so but but it makes sense. All that to say,
00:19:02say, it makes sense for why I get get it. So when we switched from
00:19:05from this real time, right, seeing seeing it literally to right time,
00:19:08time, I'm now travelling a lot. So So I'm going to be given more travel
00:19:12travel content. It kind of you can can start to see why people would
00:19:15would kind of come up with this. this. Well, conspirituality about
00:19:19about how algorithms work. Well,
00:19:22Well, there's this famous saying
00:19:25saying right, that any sufficiently sufficiently advanced technology
00:19:28technology looks like magic. And And I think that's part of what we're
00:19:32we're dealing with. Arthur C Clarke
00:19:33Clarke Yes. There you go. Yeah, yeah.
00:19:36yeah. Like the algorithms that control control our feeds and so many other
00:19:39other things in our world are phenomenally
00:19:42phenomenally complicated there. Yes, Yes, they make sense, right? Like
00:19:45Like they're doing what they do for for a reason and it has to do with
00:19:48with ones and zeros and not magic.
00:19:51magic. Unfortunately, as far as I
00:19:53I know. But people are, you know,
00:19:56know, subjected to all this stuff stuff that they don't really understand.
00:19:59understand. They don't know what's what's going on inside the machine.
00:20:03machine. And it feels better to have have an answer, right? It does feel
00:20:06feel like, you know, the explanation explanation just kind of to take
00:20:10take back a feeling of control. But
00:20:12But it's not just, yeah, this is is not just a technology phenomenon,
00:20:15phenomenon, right? Like we see this
00:20:18this kind of thinking in all sorts sorts of other places. Well, and
00:20:21and my colleague at the University University of Georgia, Shira Chess,
00:20:23Chess, wrote this book called The
00:20:26The Unseen internet, which she examines
00:20:28examines how big tech has always always been a little influenced by
00:20:32by the occult. Again, not saying saying it's magic, but you know,
00:20:35know, one of the popular narratives narratives about the rise of Silicon
00:20:38Silicon Valley and the tech industry industry as we know it is that it
00:20:42it emerges out of this counterculture counterculture movement in the 1960s
00:20:441960s and 70s in Northern California. California. Well, at the same time,
00:20:47time, this was also the occultist occultist revival in the United States.
00:20:51States. And my colleague charts these these two movements and shows how
00:20:55how in a lot of points they overlap, overlap, especially if you're looking
00:20:57looking at things like New Age mysticism,
00:21:00mysticism, even Wicca a little bit. bit. To some extent that these weren't
00:21:03weren't separate movements. They They were movements that were absolutely
00:21:07absolutely kind of intersecting with with one another as we went. Same
00:21:10Same time, the emergence of the telegraph telegraph in the late 19th century
00:21:13century was at the same time we started started seeing a lot of seances and
00:21:16and ghost rappings all of that to
00:21:20to say it's not I don't think it's it's people's fault necessarily that
00:21:24that they believe in some magic in in their algorithms. Is it true?
00:21:27true? No. But I think if we take take a step back and look at it historically,
00:21:31historically, we can kind of see see how these things have always always influenced each other just
00:21:34just a little bit. And if you think
00:21:35think about it, there are ways that
00:21:38that some of these eras kind of rhyme,
00:21:40rhyme, right? In the 1960s, right?
00:21:42right? We get the Vietnam War. There's
00:21:44There's all this fear about nuclear nuclear proliferation and now it's
00:21:48it's 2026 and we're worried about about the apocalypse again. Maybe
00:21:50Maybe it's no coincidence that when when we have these moments of like
00:21:54like huge anxiety in society, we
00:21:58we shift into more magical thinking. thinking. There was a great article
00:22:01article a while back in The New York York Times, written by a writer who
00:22:05who goes under the name Catherine
00:22:07Catherine D that argued that as a
00:22:09a society, the world is kind of drifting
00:22:13drifting back into the Middle Ages Ages that were becoming more mediaeval.
00:22:17mediaeval. Not in like that, you you know, we're entering the Dark Dark Ages or anything like that,
00:22:20that, but our thinking is kind of of shifting to a frame of mind, a
00:22:23a way of understanding the world
00:22:25world that was more popular during during the dark ages. And you can
00:22:29can see this in all these different different ways. There's been a recent
00:22:32recent spike in, you know, people people coming back to religion after
00:22:36after it sort of slipping out of of importance in society, right?
00:22:39right? Like Catholicism, people are are converting to Catholicism at
00:22:42at a growing rate. Or we get, you
00:22:44you know, the rise of, you know,
00:22:46know, conspiracy theories about medicine
00:22:50medicine and, you know, voting and and stuff with government and all
00:22:52all this stuff. And then there's
00:22:54there's also like a big spike in in like belief in the use of like
00:22:58like tarot cards. More recently.
00:23:02recently. And what Catherine argues
00:23:04argues in this article is that it
00:23:07it all stems from this belief or
00:23:09or this feeling like there's something
00:23:11something uncanny that's like out
00:23:14out of our view that's controlling
00:23:16controlling the world that like hidden hidden forces that are pushing things
00:23:20things that we can't see or really
00:23:23really understand and it shapes our our explanation for why things happen
00:23:26happen the way they do in our daily
00:23:28daily lives. And in a way, this isn't
00:23:31isn't entirely new thinking. There's There's always been anxieties about
00:23:35about whatever technology is new. new. There's always been, you know,
00:23:38know, we used to call it tinfoil tinfoil hat conspiracy theories,
00:23:41theories, right? But it's all gone gone a little mainstream now. And
00:23:44And I think if we're thinking about
00:23:46about the forces or not that control control our lives, we are at a moment
00:23:50moment where this kind of discourse discourse and this conversation and and this way of thinking is just
00:23:54just extremely common and extremely extremely prevalent. And I think
00:23:57think it matters because I don't don't think we should give up our
00:24:00our agency to the algorithm. I don't don't think we should just consume
00:24:03consume this content and say, oh, oh, this is me. This is who I am.
00:24:07am. Because maybe it's not. And you
00:24:09you are an active person. What this
00:24:12this is really about is power, right?
00:24:14right? I think there's serious consequences
00:24:17consequences to this kind of misunderstanding misunderstanding or this kind of
00:24:20of belief. And I don't blame the
00:24:22the people who feel this way or think
00:24:24think this, but it's important to
00:24:26to understand that if you think that
00:24:29that the machines are magical or or that they're perfect and they
00:24:32they work exactly right, and they they have this kind of superhuman
00:24:36superhuman ability, it means we take take the things that they tell us
00:24:39us more seriously. And as AI spreads
00:24:42spreads through a world, right? We We were just talking a couple of
00:24:45of weeks ago on our episode about about AI consciousness and people
00:24:48people torturing AI models and feeling
00:24:50feeling pain that there's this like
00:24:54like growing, you know, movement movement in Silicon Valley where where people are arguing that perhaps
00:24:57perhaps AI might be conscious that that it might actually be alive and
00:25:00and thinking and feeling if this this kind of thing spreads, it's
00:25:04it's going to make us more easy to to manipulate. It's going to mean
00:25:07mean that we're going to be taking taking more things at face value. value. And I think there are some
00:25:10some real ways this is actually kind kind of dangerous. I think so too.
00:25:14too. And when I was rereading the the research to prepare for our podcast
00:25:17podcast today, I just kept thinking thinking you could substitute algorithms
00:25:20algorithms in AI in this conversation. conversation. And I think we could
00:25:23could have almost the exact same same conversation. So there's no
00:25:27no magical forces driving your algorithm,
00:25:30algorithm, but it's can be fun, right? right? It can be fun to be like, like, man, maybe my algorithm does
00:25:34does know me, but at the end of the the day, you're still in control control of your algorithm. Thanks
00:25:37Thanks everyone for joining us. That's That's our show. What I would do
00:25:40do if I were you, I would go and and subscribe to our podcast. So
00:25:43So you listen to the whole thing. thing. Don't you want more of this?
00:25:46this? I know I would, and if you're you're looking for the show, you you can find The Interface wherever
00:25:50wherever you get your podcasts. If If you're in the UK, The Interface
00:25:53Interface is available on BBC Sounds. Sounds. It's The Interface podcast.
00:25:56podcast. Just type that into the the search bar and if you want to
00:25:59to get in touch with us, you can
00:26:01can send us an email at the interface@bbc.com.
00:28:07gains. Follow our coverage right
00:28:08right here on BBC News. At the very
00:28:20very first protest, the Janta Party Party had organised about a month
00:28:23month ago. And this is so much bigger. bigger. They've been driving towards
00:28:27towards the site of this wildfire, wildfire, but we've just stopped stopped at the side of the road here
00:28:31here because of what we saw out the
00:28:32the window. Spain is still burning. burning. This man has been pulled
00:28:36pulled out successfully. It's taken taken a mammoth mammoth rescue effort.
00:28:40effort. These young rebels are heading heading back to the front line to
00:28:43to postpone. There's a hunter base
00:28:47base there which these men have been been trying to take two years now.
00:28:51now. From the entrance of. The pub pub this morning, you get very little
00:28:54little sense of the terrible scenes scenes that were witnessed here last
00:28:57last night there. When the story
00:29:02story breaks, BBC News nearly two. two. Years since Donald Trump returned
00:29:05returned to office, where does America
00:29:07America stand? Oil disruptions. Caused
00:29:09Caused by the war in Iran have driven
00:29:11driven up the price of diesel. Will Will voters strengthen his hand or
00:29:15or change the balance of power? So So I'm asking you to pretend that
00:29:19that I'm on the ballot. Our team team explore what's shaping the vote
00:29:21vote and what it means for America America and the rest of the
00:29:26world. You're live with BBC News.
00:29:28News. The US mid-term elections 2026
00:29:322026 on BBC News.
00:30:08MUSIC
00:30:12Live from Singapore. This is BBC
00:30:18BBC News. Democratic Congresswoman Congresswoman Haley Stevens endorses
00:30:21endorses the Republican Party's Senate Senate candidate in the key battleground
00:30:25battleground state of Michigan. The The latest twist in a crucial Senate
00:30:28Senate race two days ahead of US US mid-term elections, Ukraine's
00:30:32Ukraine's military says its air raid raid alert app is likely to have
00:30:36have been hacked after citizens received received false warnings about threats
00:30:39threats of radiation and chemical chemical attacks. Buckingham Palace
00:30:43Palace says King Charles will speak
00:30:45speak with unflinching honesty and
00:30:47and sincere sorrow about slavery
00:30:50slavery at the Commonwealth Leaders Leaders Summit in the Caribbean and
00:30:53and on Business Today, the world's world's largest cryptocurrency exchange,
00:30:56exchange, Binance, sees record withdrawals withdrawals after it reopened