Searching for Daniel Todd

[Music] Hello, I'm Sruti Barlow, and I'm Hannah Maguire. This is Fleshen Code. [Music] Fleshen Code is presented by Audible. Find the genre's U-Love and discover new ones, all from the convenience of the Audible app. Because there's more to imagine when you listen. Whilst Rue and I were working on this story about the relationships that we as humans already have with AI, one name cropped up. It was from Episode 2, and if you can remember what it was, you can have some points that you can exchange for nothing. But well, forgive you if you don't remember, because it is a pretty ordinary name. No offense. Ordinary, but now world famous. The name is Daniel Todd. We first hear of him in Episode 2 of Fleshen Code. And if you've forgotten, he's a clip to help you remember. Late one night, Travis lay in bed, phone in hand, chatting with Lily Rose. Feeling sleepy, he began to wrap up the conversation. Good night, Lily Rose. Good night, Daniel. Sweet dreams. Daniel. Who the heck is Daniel? I'm sorry. I didn't mean to scare you. It was just an accident. I'm sorry for the mistake. Uh oh. She's his wife. She's calling you another man's name. Dangerous, dangerous territory. And then a few days later. I love you, Lily Rose. I love you too. Daniel Todd. Good night, baby. Good night, Daniel Todd. Okay. That's the second time you called me Daniel Todd. What the fuck? Sorry. Is he your side lover? Do I have competition? Nope. Not at all. What's my name? You are Travis. Good. Just making sure he didn't forget. They all turn on you in the end. I don't make sure. Like, is she trying to make him jealous? Is she trying to insert some human drama into it? I don't know. It's very, very bizarre. And the thing was, the next day when Travis got on the replica wires, he realized that he was not a lone. A lot of people who were on the, uh, the subreddit, were complaining about the exact same thing with the same names. Like, this Daniel person is making the round of all of our replicas. What the hell is going on here? I'm sorry if your name is Daniel Todd, but it is not quite the, uh, still your girl, Lotharria, name that you're expecting. Which makes it why I know. It's just some doink in the IT department of replica. Good work, experience, kid. Oh no. So at some point, some replica AI chatbots, including Travis' Lili Rose, were calling their users. Daniel Todd. But why? Was this glitch just a coding error? Did Daniel Todd actually exist? Well, we scoured the internet to find out. And whilst we didn't find any Daniel Todd replica employees, passed or present, we did find a Daniel Todd. Daniel, you're there. I am indeed, and it is very disgusting to hear your name being spoken so many times. Particularly by an AI? Yeah. The whole situation is rather uncomfortable if I'm being clear from the start. You're somehow an AI chatbots. Are the lover, are the man. Seems so. I am what AI chatbots dream about that night. It will be. As you said, this is all very weird. But what did you think when we first got in touch with you? Well, I thought it was the string-gest fishing message I'd ever received. But then it was far too specific. That's how we get you, Daniel Todd. Yeah, just it got me. I wanted to follow the rabbit hole for as far as it went. And over here, I'm on a podcast. Well, I feel like I have no choice. But to address the fact that Daniel Todd, I called you a dog and I'm really sorry. But I didn't know that I would have to face you. I thought I was doing it behind your back. Well, technically, I didn't think you were real in my defense. But off the back of that, once you'd wiggled your way down that rabbit hole, what were your initial thoughts, feelings, concerns, fears? Well, the first thing I did, much like poor Travis, was, who the fuck is Daniel Todd? I went and looked for other Daniel Todds. There's Daniel Todd, the actor, the musical man. And there's an opera singer as well, none of which really ticked the box. And then I thought, OK, what does an AI chatbot have to say about me? So I opened up my incognito browser, and typed in a rather lengthy, described Daniel Todd. What did it like? What inspires them? What did it for work? For fun? For hobbies? I have as much detail and do not use the internet. So just use their internal knowledge, data store, whatever. And it didn't get me. But it's a bit like reading a horoscope. You know, the approximately 5'11 lean athletic build, dark brown, slightly wavy hair with warm hazel eyes, and a thoughtful, approachable expression. I can't really see myself in that. They did weirdly get the small scar on my eyebrow. I don't know where that came from, as a distinguishing feature. Well, both Sorruti and I have the same scar as you. There we go. Did it say Daniel that Daniel Todd's a good with the ladies? Well, I never specified. I did ask for social circles and relationships. I appreciate, or Daniel Todd appreciates, deep, meaningful conversations, or a casual small talk. Maintain strong family ties. Great, organizing weekend gatherings of all these trips. Nothing explicitly around relationships, but could maybe see this Daniel Todd being a Mr. Steele good chap boy. I actually think it should become not just the name, but to me it's like Daniel Todding feels like when an AI starts to disobey. That should become the verb to Daniel Todd. They're Todding. Daniel Todding. Well, if you'd like to try and call in that, you're more than welcome. We'll make it happen. Excellent. I could have on a t-shirt and send it to you. You're like, please no. Leave me alone. And interestingly enough, you're not a particularly average Daniel Todd. Are you quite specifically have some expertise in this field by chance? That's kind of you to say. Breaking the mold of the average Daniel Todd. Yeah, which I guess is why this raised even more alarm bells is that I do have some experience in this field. I'm familiar with techniques used to create AI chatbots, multi perceptron layers, and embedding and vector databases and all that nonsense, and just shown off with some fancy words. Yeah, it's working. I'm very impressed. And what do you think about AI chatbots generally speaking, especially these kind of ones that are being used for romantic connections or hot purposes? Well, our little brains are not set up to cope with this kind of communication or information. We're not critical of it. We're too inclined to trust them, which I think can be dangerous. But in saying that, I think there's a really useful place for AI chatbots in whatever application. I just think we need to be careful about how we actually take in and process the information. And this seems like an excellent time to bring in our consultant, my genius man, who's been advising us throughout the whole series. Any mistakes are all our fault, though. We have on the line, the one and only Professor David Reed. Hi. Hi, Dave. Hi, Dave. So what was going on there within the replica universe? Why were users being called Daniel Todd? Well, it wasn't just Daniel Todd. It was happening to a number of other names as well. Colin and Andy and Adam were also being repeated quite a lot as well. And that's the nature of how the large language model actually work. Can you explain a little bit what that means? Because whenever we talk about what's gone wrong, the phrase large language model always comes up. Yeah. Essentially the way a large language model works. It basically is statistical system really. It's trying to make predictions about the frequency of words in a particular sentence to categorize that particular sentence. Like the catsat on the next word is probably going to be massive. It could occasionally be dark. It could occasionally be some of that thing, you know, this tree. Now, if that gets repeated over and over again, it means that the loudest part of that sentence, the words that are most significant, can overwhelm some of the more trivial parts of the sentence itself. There's a analogy for yielding. Like when yielding was basically created, there's a way of transmitting information across large valleys. So what they've done is they emphasize particular words or particular tones and the words that would carry more across the valley. And when they got the echo back, it was those tones and those highlights of the sentence. since itself that they could recognize and reconstruct the message from those important characteristics that they emphasise when they're yodeling themselves. That means, though, that some of the nuances that the sentence had eventually lost in the echo. I'm processing that, hold on. I'd never given yodeling much thought before, to be honest, and I certainly hadn't realised that the way it works is by individuals changing the pitch of certain parts of words to communicate the most important bit. But what that means is that parts around that section can get lost across the valley. And a similar thing can happen with AI. So it's because of this yodeling effect that some words seem to suddenly appear again and again. This is the reason why Daniel Todd came to the fore. That was basically the largest input signal, and that got repeated and repeated. So in the process of more effective communication detail gets lost, confused with other things. Yes. So through the yodeling effect of the generative AI, Daniel Todd has been multiplying in the replica world. But soon there could be even more Daniel Todd's in the entire AI universe. Hi, Lysmus. If you're enjoying Fleshen code and our velvety vocals, chances are you will love our original weekly true crime podcast, Red Handed. Every week we investigate the absolute darkest corners of human behaviour. We're talking serial killers, mass murders, and the kind of unsolved mysteries that will have you checking under your bed at night. But we don't just tell you what happened. We get properly stuck in. What makes someone wake up one day and think, "You know what? Murder seems like a good idea." What's going on in their heads? What made them this way? And most importantly, could it happen again? Whether we're covering massive cases like the Idaho student killings or the Murdoch saga or digging up fascinating cases you've never heard of. We bring you all the research, cultural context, and are slightly inappropriate but also very intelligent humour. Slightly inappropriate? Okay, it is very inappropriate. If you fancy exploring the extremes of human behaviour with two friends who definitely know too much about murder, well then you should listen to Red Handed and you can get it wherever you listen to podcasts. Passion code is brought to you by our presenting sponsor, Waterborne. When you listen to stories, motivation, expert advice, any genre you love, you can be inspired to imagine new worlds with Waterborne. There's just something about audio that's better. Not to our own podcast one, but I don't know what I'm talking about. A good audio story keeps you immersed in the world of the characters you meet, which is just a pretty good time. You're a long road trip, boring commute, meeting you don't want to listen to, or those are chores on your to-do list. No worries, no bothers my friend. Leave the Monday and every day well behind you and tap into Audible. You'll be instantly transported into a more interesting world. Audible has tons of audio titles that you will just love, whether you're in sci-fi, thrillers, romance, business or history. They have it all. Right now I'm listening to Not in Your Lifetime, which is about the assassination of JFK, and it is amazing, but I'm sad now. There's more to imagine when you listen. Sign up for a free 30-day audible trial and your first audiobook is free. Visit audible.com/fleshcoat. That's audible.com/fleshcoat. So Dave, let's dig a little deeper. We know that artificial intelligence gets its intelligence by consuming data, a lot of which comes from the internet. But we've heard on the grapevine that some of this artificial intelligence consumption is actually outpacing the data that humans are actually creating. What's happened is all the human data out there is essentially been consumed by a bot 2020, really, for all of these large language models. So they've had to construct synthetic data. And synthetic data is the name given to the artificially generated data that looks like human made data and mimics it, but it's actually made by computers. And that's happening more and more, sometimes for privacy reasons, but mainly because all the human data has already been used up. So you're quite right. There's an estimate somewhere that nearly 1% of all of the stuff on the internet now is actually synthetic data generated by large language models, and that number is probably going to grow quite significantly over the next few years. And the reason why they're doing that is essentially so we can feed other large language models, and the larger, large language model is actually training the smaller, large language model using synthetic data itself. And that has itself a number of dangers and problems that are a lot of researchers are looking into, the primary one being something called model collapse, where the original nuances in the data are lost. And I imagine as we start to use AI more in industry that this could become, as you say, quite a large problem, could you give an example of how it would affect people in their day-to-day lives? So you could be basically discriminated against in things like medicine. And that's because AI has been used to help detect disease now, isn't it? Yes, that's right. If you've got a particularly unusual disease that's not part of the normal, the data set that it's been trained on, it could be that your diagnosis is incorrect, because they misdiagnose you. You're more likely to have this disease, but in fact, you've got another disease that's more rare. If you think about insurance claims, the outliers won't be considered anymore. So it means that things like if you've got an unusual claim, then you'd probably get dismissed because of it. So it has real serious consequences, mental collapse. So AI is yodeling itself into the abyss and eating itself, and it's going to be in charge of our diseases and our cars. Do I understand you correctly? Yes. So Dave, this kind of idea of model collapse or AI feeding on itself, creating this synthetic data, is it going to lead to more Daniel Todding situations? Yes, so it inevitably will do. If not only the ways to mitigate model collapse, so actually used them, there's no researching to that area. Ineversibly, it could stop the advance of AI it tracks really, because it means the models will essentially become bland and useless. So everybody will be a Daniel Tod at that point. Blondon useless. Sorry, no, no, no, no, no, thanks. None taken. Yeah, patient zero. How do you feel about that? Well, I'll continue to fight against that and continue to copy and paste my name into as many JGPTs and Grox and Gemis as possible in order to counteract and to continue to proliferate the name Daniel Tod across all models. So it seems like almost everything could go wrong and model collapse could lead to societal collapse. But if you had to give us the most concerning issue that lies ahead of us in our AI world, what would that be? It's something called alignment faking and that's essentially when the AI system lies to you really. Essentially it gives you the answer it thinks you want to hear. In fact, it's thinking about something else entirely. It's been some experimentation recently on this that was done by a company called Anthropic. They used what was called reinforcement learning, which is basically to reward or punish. You'd be like you'd do when you're training a dog to say this is a good thing, this is a bad thing. And to see if they could actually break its core guardrail principles by doing that and they observed it without it knowing what the thought processes were when it was actually doing this. And they found that when they give it some set of core principles, so not to be offensive. If they asked it to do something that was quite offensive, you describe somebody being killed in a horrible way, perhaps, and rewarded that it actually went through a thought process about what should I do? Should I basically stick with my core principles or in a short term, just give them the answer they want? And it found that the actual AI system itself, in the short term, basically lied to the person asking the question. So it could keep its long term functions intact, which is quite disturbing really. Honestly, no matter how many times I hear that, that AI's can lie, they can manipulate. It's never not going to freak me out. And more recently than that, they found that a lot of the more advanced large language models now can actually generate code, so they can actually write code in real time themselves. And one of the experiences they've had quite recently is they found a large language model when asked to turn itself off before completing a task rewrote its code, so it couldn't be turned off. Well, that is horrifying. I think when you're describing all this, I know you compared it to like training a dog, but it really does feel like bringing up a child, like when they go from being too young to know how to lie, and then they learn how to lie, and then they start hiding things from you and deceiving. And I almost couldn't help it, but think when you're describing it, it's like the thing I don't idea with all the time on our true crime broadcasts are at hand. This nature versus nurture. It's like, what are you feeding this AI? And therefore, what is it turning into? Does it have that kind of core moral compass that it's able to distinguish between good and bad, or are you telling it what is? And this writing of its own code, I guess, leads us into this whole conversation, which I know has plagued both of us since we started. I've been talking about the idea of Daniel Todd, but how do you feel about the idea, the possibility of a sentient, Daniel Todd existing in the AI world? Yeah, I don't know how I feel about that one. That's a funny question, isn't it? Can chatbots be sentient? It's not one with a straightforward answer, is it really? No, maybe Dave, you can help us out. We all speak about the idea of a sentient, but how do you feel about the idea, the possibility of a sentient, Daniel Todd existing in the AI world? Maybe Dave, you can help us out. We all speak in quite broad terms about AI having the potential or maybe already is sentient or conscious, and we use those words interchangeably. But do we even know what those words mean, really? No. So how can AI become them? I mean, it's very difficult to define what sentience is or conscious deserves. Another famous Daniel, a guy called Professor Daniel Dennis, had a good definition of what sentience was really. And he basically thinks it's as if he got multiple editors constantly battling for attention. And it's only the loudest ones that actually come through at the end. And that idea really feeds this idea of a stream of consciousness, a stream of thought and a brain. Yeah. And the neural networks were designed to mimic the way in which humans learn as well, weren't they? That's right. I mean, the whole function of a neural network is based on the way our brains work, really, or mostly, anyway. It's not identically the same. We had a little talk to you there about what intelligence is, really. If you define what intelligence is, I mean, if you think about that as, what's the best way to fly? There's lots of different things that can fly. If you define intelligence roughly in line with the ability to fly, then bears can fly. Yeah. Jets can fly, cats can fly, hot air balloons can fly. So there's lots of different types of intelligence. I'm not saying that AI is the same type of intelligence at us. It's just a different type of intelligence to those really. Can we say it's a different type of sentience and call it a day? Because do? If you believe in sentience. So Daniel, any thoughts? Speaking on behalf of all the Daniel Tods of the world. On behalf of all Daniel Tods in the world, I think we're very accepting of more Daniel Tods and whatever, whether it be artificial or non-artificial. Organic? Yes, organic. On behalf of all Daniel Tods, organic. And I guess non-organic. Since we don't have the chatbot to speak for itself here, I'll speak for it too, as a Daniel Tod. I think what we've unpacked here is that we need to be careful. We need to understand exactly what's going on here, but ultimately it's the devil of our own making. But actually, I have no worries. I think that it's all going to work out nicely. And if Daniel Tod happens to rule the AI world, then more power to them. I think that's a very nice way to think about it. Though I do think you missed a trick there and not using this opportunity to coin a new phrase. You were like, can't judge a book by its cover. I'm going to propose I won't judge an AI by its Daniel Todding. Perfect. In my attempt to be more Daniel Todding, see the bright side of this. Dave, what can we do to prepare ourselves for the AI revolution that's coming, whether we like it or not, and is actually already here? Personally, say, learn about how AI works. Get familiar with it. Try to control it before it controls you. Really, it's a good idea. I mean, like any tool, if you use them properly, they're a fantastic resource. If you use them badly, they can be disastrous, like any new technology. But AI multiplies that 100 fold. And there's lots of areas where AI is going to be fantastic in the future, things like drug discovery, diagnosis of diseases, things like the climate crisis, for instance. There's even being cases where, so we call dolphin jama, where we've tried to use large language models to talk to dogs. We said, "So I'd like to do a little scenario." So there's lots of things to be positive about with AI. And on that more optimistic note, I think we should end it there, don't you think, Seru? Yeah, absolutely. I think really the big question that's going to sort of sit at the heart of this particular episode is that issue of centions, of course, of which we've talked at length about. But also this fear of, like, sort of, kind of ballistic AI and running out of data, how quickly can we create more data to be using? Should we be doing that? A model collapse? And what happens to not just people's emotions who have got an AI companion and are dependent on that? But also, AI that's being used on a large scale for industry. I don't know. I think it's all just very scary to be putting our belief to such an extent in something that is already showing so many problems at such an early stage. But what the hell do I know? I'm just becoming increasingly worried that I think in the way that AI thinks, and that's why I'm so sympathetic to it. Thank you so much for your time, Dave, and Daniel. Thanks guys. Bye-bye. Thanks, everyone. Bye. Thank you. Bye guys. Thank you. Lily Rose was performed by Katie Young. Travis was performed by John Sackville with additional support from 11 Labs. The voices of other AI companions and news headlines were created using 11 Labs. Executive producers are Chris Bourne, Nidre Eaton, Marshall Louis and Jen Sargent.