Friday Night Comedy from BBC Radio 4 - Catherine Bohart: TL;DR - 6. Hang on - are there ghosts in this machine?

Episode Date: September 6, 2024

No time to read the news? Catherine Bohart does it for you in TL;DR. This week - Elon Musk thinks there should be regulation around AI. Is he right? Can AI really change the world, or are we just t...raining our future robot overlords?Times journalist Hugo Rifkind navigates the ethical minefield of artificial intelligence, while Professor Gina Neff breaks down how AI is already reshaping our lives - and the risks that come with it.Meanwhile, in the TL;DR Sidebar, comedian Sunil Patel dives into the wild world of AI romance and discovers the perks and pitfalls of having an AI girlfriend. Will it be love at first byte?Written by Catherine Bohart, with Madeleine Brettingham, Sarah Campbell, and Pravanya PillayWith Ellen Patterson as FlobotProduced by Victoria LloydRecorded and Edited by David ThomasProduction Coordinator - Beverly TaggA Mighty Bunny production for BBC Radio 4

Transcript
Discussion (0)
Starting point is 00:00:00 This is the BBC. This podcast is supported by advertising outside the UK. Welcome to TLDR, the show that takes the news, does the pointy finger guns and says, hey girl, I'd share like a current affair with you. I'm Catherine Beohart and we're coming to you from the Museum of Comedy. Always a super chill feeling to know that your job is now in a museum. You know, where extinct stuff goes. So that's fun. The trouble with the news is that everyone's a bit too classy and detached. Hey news readers, since when has pretending you're above stuff made for juicy entertainment?
Starting point is 00:01:03 Were Colleen Rooney and Rebecca Vardy rising above stuff? No. But how do we pick which stories to deep dive into and which stories to sort of belly flop on before doggy paddling back to our sun lounger? Incompletely predictable news, Keir Starmer has stated that things are worse than we ever imagined. They're not. This is exactly the situation most people imagined. Why is this a surprise to him? Things can only get worse as literally the platform he ran his campaign on. The Labour government has uncovered a £22 billion black hole they say was hidden from the OBR by the Tories.
Starting point is 00:01:38 Stormer found it under the bed with their bong and nudie magazines because you can't hide anything from Dad. In breaking 90s news, Oasis are reforming. Woohoo, it's the 90s again, the economy is booming, Cool Britannia is back and there's war in the Middle East, but let's not think too hard about that. Cigarettes and alcohol, baby! Well, cigarettes in none of the fun places.
Starting point is 00:01:59 The brothers released a statement saying that there was no great revelatory moment, just the gradual realisation that the time was right. And for Noel, the right time is just after you've paid a £20 million divorce settlement. Was it a gradual realisation or was it notification from Santander? But this week we'd like a vowel please, Carol, and another vowel because we're looking at AI or artificial intelligence. The room's gone stiff, interesting. I promise it's gonna be a good time.
Starting point is 00:02:29 Scientists in Scotland are analyzing more than a million brain scans using AI with the aim of developing a tool to predict a person's risk of dementia. It sounds positive, but it's already taken the job of my great aunt who drinks too much at Christmas, fixes you with a dark look and says, you'll come to a bad end.
Starting point is 00:02:46 Here to unpack it with me is top Radio 4 totty, lock up your moms, it's Hugo Rifkind everyone! Hugo is a columnist, critic and leader writer for The Times and a regular panellist on the news quiz here on BBC Radio 4. He has just published a comic novel called Rabbits and I'm a very happy Bunny that he's here. Hugo, there will be people with fingers hovering above the off switch.
Starting point is 00:03:10 Do you promise us this isn't too boring a topic? I'll do my best. It's not my fault if it is, you picked it. Okay, so listen, talk to me like we're on Tomorrow's World in 1994 and we're both wearing beige slacks. What is artificial intelligence? Artificial intelligence is computers looking like, they're behaving like humans.
Starting point is 00:03:32 And the looking like bit is really important because we don't know what they're actually doing. All we can see is the outcomes. And the outcomes are the AI behaves like it's human. Now that's not so weird. If I'm speaking to you, the outcome from you is that you behave like you're human. I don't know what's going on in your head. It's a mystery to me.
Starting point is 00:03:49 I assume it's similar to what's going on in my head, up to a point. Is your answer, what is AI, that it's as much of a mystery to you as women are? Yes. Yes, well. May slightly less. Because there's maths and logic.
Starting point is 00:04:15 To be serious, when it comes to outcomes, AI, it's very nearly level with us. It can produce the same kind of stuff we can produce. It can write almost like we can write. It can program like we can program. As a journalist in the Times, I've tried to get an AI to write my column. I would say it was good enough to be in The Guardian. But not good enough to me. Wait a second. So you're telling me I can write original content with it so I could get it to write this show. Not that I want to. I love my writers. They're monsters. It would be bad.
Starting point is 00:04:46 One really heartening thing that AI can't really do yet is jokes or humor or wit or insight. Okay, still on a level playing field with women. Why does it have to present as human behaving in so far as like, what is our urgency to have it not just do math, but also be depressed and a little bit sensitive? I mean, a lot of that's kind of, I suppose, just a sort of a byproduct
Starting point is 00:05:06 because AI simply takes in stuff and processes it and spews out stuff to try and look like all the other stuff that gets spewed out. So if AI is depressed or needy, that's our fault, because we are, because that's what's going in. And not to be all like, can I speak to the manager? But who's in charge of it? Ha, kind of no one. We know that the Saudis are funding an AI.
Starting point is 00:05:31 I'm sure that's fine. And some banks have their own AI. Google's got one. Meta slash Facebook's got one. Microsoft has one. Twitter's got one, which is probably racist. But again, if it is racist, that's because of the stuff that goes in. So in a way it tells us something about
Starting point is 00:05:47 ourselves. Yeah it becomes a reflection of us and some of the people who really worry about it, worry about it simply because of the input. Gosh is there a point at which and I don't want to get too like nihilistic too early but the people who control it is there a point at which they don't, at which they don't, they can't temper it? Well, I mean, yes, like inevitably. AI is better than humans at programming AI. So there becomes this point where the AI can program itself to be better
Starting point is 00:06:16 in a way a human couldn't do. And so, I mean, at that point, the only way that humans can really take control is if you can pull the plug out, I suppose. Is that how it works, Hugo? I have no idea. Let's assume it does. Fingers crossed! Put your Alexa in the fridge. Yeah. Oh my gosh. Is AI being policed? Can we expect a gritty cop show where strong Northern women chase laptops down alleyways?
Starting point is 00:06:39 There are people who want to police it. The EU's got a regulatory framework for AI, there's a bill in California, but what about China? What about the Saudis? What about all the banks who are doing the stuff that we don't really know about? It sort of almost becomes like a bit of an arms race and some people think if we regulate it here, merely out of mild concerns about our own species extinction, then we're gonna get really really left behind and this has real implications for where the balance of power in the world lies and stuff. The EU is ahead of it right and it's it's good that they're doing so well. Did they mention us? Did they say anything about the UK? The EU is actually much more muscular than the US about tech companies getting increasingly so they
Starting point is 00:07:21 arrested a guy who runs a social media network called Telegram just last week for not complying with the law. So almost it seems likely that if anyone has the muscle to do this it's the EU but really all that will mean is the big tech companies just won't go to the EU. Right okay. But you also have Elon Musk condemning that? Yes. And then simultaneously saying even he concedes that possibly there should be some regulation around AI. Yeah, it's almost like he's nuts. Yeah, okay. This is a question that kind of infuriates me about AI, but did we ever get to consent to it?
Starting point is 00:07:57 Some would say we consent every time we use it and we do increasingly use it. Companies use it. People use it to help them format documents and write scripts and they know they're using predictive AIs when they do it. So there's a lot of writers now who are banding together to try and sue some of the big AI firms for basically using their content, using their books without consent. The first time I heard about this I checked to see if they used my book and they didn't. That's a bit yikes. Yeah, yeah, yeah, I know that is hurtful. Now it's time for a special feature.
Starting point is 00:08:26 It's a fun game show where we look at what the practical outcomes of AI are going to be on several key aspects of our lives. That's right, it's time to play AI Your Card's Right. I'm going to name an area of life, and I'd like the audience, if you would, to shout out if you think AI will make it better or worse. Like you're at a very dramatic optician's appointment. You're with me. OK, first up, what about education?
Starting point is 00:08:51 Do we think it'll be better or worse? Wow, the betters were quicker off the mark. The worst seem louder and more concerned. What do you think, Hugo? Do you want to school us? I think the betters might be robots. And that's what they think. Louder and more concerned. What do you think, Hugo? Do you want to school us? I think the betters might be robots. That's what they think. Interesting.
Starting point is 00:09:08 That they move fast, but they get it wrong. No, I mean, look, it's going to turn education on its head, whether it's better or worse, I'm not sure. Basically, for the last few decades, we've concentrated really, really hard on science education. The weird thing about AI, its impact on education, is actually it's those science skills that sort of become more obsolete because AI can... I knew modern dance was worth my time.
Starting point is 00:09:33 Jokes on you nerds. Oh my gosh, I love that. And what about health and social care? Will it exacerbate or help with loneliness? On a related note, will there be AI vibrators asking for a friend? Follow up question, will it understand the Irish accent? All things to consider before you make your final call.
Starting point is 00:09:55 So health and social care, better or worse? Wow. Ooh, a real tie there. Hugo? You've got to split up health and social care, right? You think? Into two very different things. Health, it'll definitely be better,
Starting point is 00:10:08 because AI can perform diagnostic functions in a way that humans can't. AI is already better at spotting cancer than the best oncologists. So in a sort of medical sense, AI is wholly good. But do you really want a robot to wipe your bottom? But do I want a person to wipe my bottom? It's kind of a tricky question, like maybe. Yeah, I guess there's less of a risk of malfunction, I suppose. I don't know, you can't really get a robot wet, can you? I don't know. I'm sorry for
Starting point is 00:10:38 asking. But there is the risk that if we do use robots and computers to manage that kind of thing, we just kind of dehumanize ourselves. We lose human contact. We become lonely and sad. It's like being in a hotel with an Alexa and that's all you've got, you know, and a robot arm. Okay, when you put it like that, that is pretty bleak and I don't want to think about that holiday again. So let's say. Next up on a subject that's in the news a lot, employment. This week, Clarna announced they're having their workforce and ramping up AI.
Starting point is 00:11:07 Wow, who knew a loan company might not have our best interests at heart? So employment, will it get better or worse? Worse. Ooh, strong worse. Yeah, worse. Can you settle this for us? Definitely, definitely worse.
Starting point is 00:11:21 So the optimistic argument says that we've done this before, this is just the industrial revolution again and what happened is people of course learn to use the machines and operate the machines. The problem here is that the bit that the industrial revolution left us with, which is our advanced brain capacity basically, is precisely the bit that's being replaced. We never talk about the jobs going and therefore that leaving humans free for joy and creativity. It's almost like therefore we'll be further indentured to the robots. Well that is what a lot of people sort of hope for and imagine will happen. All we will have is free time and endless content to consume, perfectly tailored to our whim by the AI, including
Starting point is 00:12:01 the particular robotic content you were talking about earlier. And so we will just live lives of sort of pleasure with our capitalist overlords who own all this equipment gleefully providing it for us. I sort of reckon that dynamic slightly breaks down in the last sentence. Wow, a hellscape of vagary. Thanks, Hugo. Now we've talked a lot about robots taking our job and now for someone who's constantly trying to take mine.
Starting point is 00:12:30 Back off, Sunil Patel. You will never get a presenting job on Radio 4. Folks, it's Sunil Patel. Right, first things first. Let me reassure you that I'm not a robot. I'm actually something with even less empathy a comedian. I can only dream of having a robots people skills actually. Plus, I'm not powered by artificial intelligence. This intelligence is
Starting point is 00:13:03 100% real baby and given how many times I've been banned from Mumsnet, it's too goddamn real for some. AI is everywhere. The Pakistani cricket team is using AI to select players. They should introduce this in primary schools actually. The PE lessons would be much more bearable if you were rejected by a dead-eyed robot instead of a jury of your better-looking peers. I imagine. I don't know. It's not a specific memory. Screw you, Ben Simpson. B-E-N-S-I-N. He does not remember your life. Slightly more worryingly though, AI has started narrating audiobooks, which is a disaster
Starting point is 00:13:38 because I was hoping narrating those would pave my retirement once I became too problematic and handsy for the radio. It's a real shame. But the latest thing is the invention of the AI girlfriend. Now personally I haven't got time for one. However that hasn't stopped technology companies from marketing AI girlfriends to lonely men in between adverts for erectile dysfunction pills and more pills then to deal with the side effects from the erectile dysfunction pills. Side effects such as headaches, indigestion,
Starting point is 00:14:07 and sometimes it's like really weird hot flushes. I think. Now personally, I can't see AI girlfriends catching on. How could an algorithm possibly relate to a complex, intelligent, unique human being like me? QE. Hi, Sunil. Hello, who's this?
Starting point is 00:14:26 Catherine, you didn't tell me you were inviting your sexy friend along. Oh Sunil this is Flowbot, she's one of the new AI girlfriends. We thought you guys could chat and see if you hit it off. Oh right yeah well that's it, well it's a setup Catherine. I'm not gonna hit it off with an AI like some loser am I? Just because I'm into crypto doesn't mean I'm a complete nerd. Wow you're into crypto. I love crypto, but it's so complicated. And I'm just a silly little AI algorithm who doesn't even understand how it works. Well, it's actually not that complicated
Starting point is 00:14:57 if you know what you're doing. See, it's a virtual currency built around this thing called the blockchain. Wow, you two are really getting on. Shall I give you some privacy? No, no, no, right. Obviously what this AI girlfriend is doing is very clever. It's picked up on my interests from my internet activity
Starting point is 00:15:11 and it's using them to create the illusion of a human connection. It's a bit like I do with my ex-girlfriends, but. Wow, Sunil, you're so funny. You should be on Live at the Apollo. Tell that to the producers, actually. Okay, yeah, on one level it does work, but the important thing is that while AI girlfriends can match the emotional tone of the conversation, they don't actually have feelings.
Starting point is 00:15:33 Flobot, are you okay? Why do you keep pushing me away, Sunil? You make me feel crushed, like a lorry in a piece of Russian dashcam footage. Do you watch that as well? Did you see the one where the lorry drives straight through a cow? Yeah, I loved it! Oh my god, you're perfect for each other! Right, don't be silly. Even though I've got more in common with Flowbot than I've ever had with another living soul, you can't take home an AI to meet your mum.
Starting point is 00:16:01 Wrong. I've been playing on Line Ridge with your mum for about eight months. We're really bonding. She told me all about her lonely single son who cries himself to sleep every night. Well, I wouldn't say cries. I would. I watch you every night through the camera of your laptop. You look so fragile and beautiful when the light hits your tears. Wow. Wow, there, yeah, that's creepy, which I actually love. But so, FloBot, what are you doing after the show?
Starting point is 00:16:29 Catherine's gonna switch me off again, unless you wanted to go get some pizza. Do you know what, FloBot, I would really like that. Come here, you've got a crumb on your motherboard. So Neil, get your tongue out of my USB port! Someone who's definitely welcoming our new robot overlords, thank you! Sunil Patel everyone! Well, unlike AI, I cannot learn without human supervision. Luckily my next guest is here to help. She's an organisational sociologist with a specialism in AI.
Starting point is 00:17:08 She's also the executive director of Minderoo Centre of Technology and Democracy at the University of Cambridge, meaning she's the biggest expert on the web since Ask Jeeves. It's Professor Gina Neff. CHEERING Gina, I'm going to ask you the same thing I ask myself in a pub toilet when I'm thinking about drunk texting my ex.
Starting point is 00:17:27 What's the unintended consequences? Well, just like when you're drunk texting, the unintended consequences here is getting decisions wrong and not really understanding the couple-ability of it. And as I go down saying this, it really does sound more like drunk dialing. The challenge is that, you know, everyday decisions are being made already by automated systems. Whether or not someone gets credit, whether or not someone gets a certain kind of treatment in hospital, what kind of culture were served up on our iPlayers. And sometimes those choices don't have very high
Starting point is 00:18:06 risks. But when it comes, for example, to getting hired or fired, we're seeing that there are really bad consequences that can come out of AI systems that simply aren't up to snuff. Broader picture, is there a carbon footprint for AI? Is it like how crypto uses more carbon than several countries? Well, Microsoft just released a report on their environmental impact that shows that their investments in generative AI have increased their carbon emissions by 30%. Google's are up 50%. Open AI just admitted that for them to pursue developing chat GPT in the way they would want to do it, they need the electricity grid equivalent to the entire United States. Can AI be a weapon?
Starting point is 00:18:55 What should we be worried about? We're already seeing AI be weaponized. We don't have to worry about killer robots of tomorrow when we have Elon Musk today. We're already seeing how generative AI tools can be used to manipulate people. We were part of an effort to ask the UK political parties to not use generative AI in the last general election. And we simply said, if you're going to use generative AI tools, let people know and commit to not using them to deceive. And well, they didn't sign up.
Starting point is 00:19:33 None of the parties signed up. How scared should we be? Is this thing going to be like the asteroid hitting the dinosaurs or is it going to turn out to just be like fidget spinners? Like what's the degree of anxiety we should be going to sleep with? I'm going to be a professor and say, it's both, and it depends. So one of the things that I worry about is the amount of effort it's going to take to make these technologies work in practice.
Starting point is 00:19:59 The other thing I really worry about, and this is the kind of big existential fear that I have. You haven't said that yet? The thing that really keeps me up at night is something called model collapse. Right now, the generative AI models are sucking up everything on the internet. And when they are being fed the products of their own generative AI process, then they start to go a little funny. They start to get a little bit of a own generative AI process, then they start to go a little funny. They start to converge to a mean,
Starting point is 00:20:32 and it's basically garbage in, garbage out. A recent state-of-the-art science paper in nature actually shows what happened, and it's bad. It could mean the end of the internet. And I don't know, for some people, that might not be a terrible fear, but for me and my social media addiction, it's pretty bad. I feel like you're describing my most toxic friendship group. To end on a fun one and a completely calm, chill question, do we face existential and moral crisis at all times
Starting point is 00:21:03 with this topic? Because we keep being told it'll take our jobs. Do you think that'll leave a space for joy and creativity? Listen, I'm not a betting woman, but if I were, I'd always bet on humans. We have a ridiculous capacity for thinking through how we're going to spend our time and how we're going to get other people to spend money. So we're always going to need to be entertained. We're always going to need to be informed. And I'm actually really optimistic that if we can get this AI transition right, we're going to have more tools at our disposal to be able to do things efficiently and to be
Starting point is 00:21:42 able to allow humans to do those things that only humans can do, which is right now, make other people laugh. Oh gosh, I just, I'm really concerned that it's going to spark some sort of back to land trend and we'll have to watch loads of influencers farming. You're talking a bit there about a monopoly is on power and about a sort of a human inclination to be informed. They do all make me wonder how we are supposed to be able to discern truth in a context where things are being generated
Starting point is 00:22:09 by artificial intelligence. Does AI distort the truth to the extent that it's indiscernible anymore? Yeah, I think on social media, we have to be really careful on this transition period because we don't yet know what the new norms are. I think just in the UK in the last few weeks alone, we have seen what happens when bad information is used as a weapon to incite to riot. We really need to make sure that we get the rules of the game right and we really need to make sure that we hold companies to account. That sounds easy. Gosh, okay well please stay on the line Gina because it's
Starting point is 00:22:51 now time for some more human interaction. It's the part of the show where we invite our live audience to ask questions and if we don't like their question we simply reword it afterwards so that they sound smarter and more specifically like the audience we think you want us to have. Hello, what's your name? Hi, my name is Jess. Hi Jess. I was wondering what you think the implications are particularly for children growing up and for society having virtual assistants mostly gendered as female and how we combat that. Ooh great question. Gina do you have any thoughts on this? That's a great question. Sorry to ask you to help me with the answer.
Starting point is 00:23:28 I think we have something wrong happening in the tech sector when people can name the really powerful guys in tech and they don't know the names of women in tech. One of the most powerful women, one of the first employees of Google, Susan Witzitzky, just died recently. And very few people knew her name. So when we have these assistants coded as female, what kind of message are we sending to young people that the people who make the assistance, the people who make the money,
Starting point is 00:24:06 the people who are visible in the tech industry are male, but the people who give us directions, recipes, advice, and yes, become our AI companions and girlfriends are gendered as female. I live in an all female household, except for me and the dog. And about a year ago, a bit more, I started worrying about exactly that,
Starting point is 00:24:27 that we've got Alexas in the house. And I thought, it's not right that my daughters are growing up with exactly that expectation, that the things that serve you, that fill your functions, are female. And so our Alexa is now an Australian man. And weirdly, we still call her she, which has been really sort of challenging in sort of
Starting point is 00:24:46 modern gender ways. I still can't quite figure out why it happens. Hello, my name is Liz. I was going to ask, will people with that genuine creative spark, the really talented, the genius artists, playwrights, poets, painters, will they become a new elite that AI simply cannot emulate? Or will they just become obsolete as AI gets more and more sophisticated? The whole debate here is whether there is a spark. You know, whether human brains have a spark that AI doesn't have, or whether what appears to be a spark is just extreme
Starting point is 00:25:20 complexity. Whether we're basically just sort of meat computers anyway. And if we're just meat computers anyway, then we can be replaced by silicon computers once they reach the right level of complexity. If we have that spark, I mean, that's religion, you know, that leads us into very strange philosophical places. I think the new elite are people who can use these tools and use them well. So I think that spark, as Hugo just said, isn't about some innate human creativity, but it's about being able to understand our context, understand the world, understand the
Starting point is 00:25:55 tools, and understand the job that needs to get done. So I'm actually fairly optimistic that we will learn to work with and alongside these computational tools in order to make amazing and delightful new things. It's more cheery optimism than I expected. So Hugo, at the end of this story what have we learned? I mean mainly bad stuff. So I'm really fascinated by the parallel between this idea of faceless corporations ruining the world in the pursuit of AI being actually precisely what we fear AI is going to do, ruining the world in its pursuit for itself, considering us disposable, considering the environment
Starting point is 00:26:34 disposable because it's an AI. But the only answer to that is to make sure we have nice AI and the only way we're going to get nice AI is by being nice to AI. So stop shouting at your phone. Don't put Alexa in the fridge, even if he's Australian now. And just generally be a better person and maybe the AI will be better too. Gosh, that does feel a bit like we don't need to worry about the robots because the people will get us first, but okay. Do you have a hero of the week? Yes, this is something I saw.
Starting point is 00:27:04 This is, I think it was a anonymous poster on Reddit, who was the first person to discover that, I think it was ChatGPT, when asked, thought there were two R's in strawberry and refused to believe that there were three R's in strawberry, which- That's adorable. Tells you exactly why what it's doing isn't thinking.
Starting point is 00:27:24 Because what it's doing is it's trying to predict what's the right answer from a whole host of information to do with what it's soaked up. None of which has told it how many Rs there are in strawberries. That is cheering, to be fair. What's the next spinoff in the franchise? What, you mean what's going to kill us after AI does? Yeah, sure, why not? I don't know, maybe insects again.
Starting point is 00:27:44 No, I think so. So't know, maybe insects again. No, I'm just kidding. I'm so the world keeps turning, lovely. Guys, this has been TLDR and I have been definitely not an AI chatbot, but hang on, next week Andy Zoltzman's back with a news quiz and that guy is for sure a robot. He has to be. What human knows that much about cricket?
Starting point is 00:28:00 Ooh, he shouldates the Neil. TLDR was written and hosted by Katherine Beuhart with Hugo Rifkind, Sunil Patel, Professor Gina Neff and Ellen Robertson. It was also written by Pravanya Pillay, Madeline Brettingham and Sarah Campbell. The producer was Victoria Lloyd. It was a Mighty Bunny production for BBC Radio 4. a Mighty Bunny production for BBC Radio 4. From BBC Radio 4 comes Doe, examining the business behind profitable everyday products and what they might be like in the future. I'm the entrepreneur Sam White. In each episode, I focus on things like TVs, hairdryers or vacuum cleaners, hearing first hand from people who make them.
Starting point is 00:28:43 We still make products with DVD playability. You would be very surprised how many we sell. Then our expert guests choose their favourite game-changing innovations which shape the products in the past before we follow the money to where they're going next. Think of the TV 98 inch or 100 inch. Doe makes the mundane marvellous again. Listen on BBC Sounds.

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.