WEBVTT - What Is AI Slop Doing to Us? 0:00:01.440 --> 0:00:04.800 Hey, welcome to Science Stuff, a production of iHeartRadio. I'm 0:00:04.840 --> 0:00:09.440 horehit Champ, and today we're talking about artificial intelligence, or AI, 0:00:09.920 --> 0:00:13.800 and specifically images and videos made by AI, which some 0:00:13.880 --> 0:00:17.599 people call AI slop, and how it affects us, our brains, 0:00:17.680 --> 0:00:21.040 our society, and our sense of reality. We're going to 0:00:21.079 --> 0:00:24.320 be talking to an AI expert about this, a neuroscientist, 0:00:24.720 --> 0:00:28.560 and somebody who's looked into the dangers of AI generated 0:00:28.600 --> 0:00:33.560 pictures of cute, fuzzy animals. Yeah, you think they're perfectly safe, 0:00:33.600 --> 0:00:39.760 but they're not. So ask chat GPT to clear your schedule, 0:00:40.120 --> 0:00:43.360 claude your way to our show and mid journey with 0:00:43.440 --> 0:00:46.720 us as we answer the question what is AI slop 0:00:47.200 --> 0:00:54.880 doing to us? Enjoy? Hey? Everyone, So you've probably noticed 0:00:54.920 --> 0:00:58.800 that AI is everywhere, and you've probably seen and heard 0:00:58.920 --> 0:01:02.120 images and videos made by AI programs in your social 0:01:02.160 --> 0:01:06.800 media feed, online, and even on TV, sometimes without knowing 0:01:06.840 --> 0:01:10.000 it was made by AI. And this trend has exploded 0:01:10.319 --> 0:01:13.120 in just the last few years. So I was curious 0:01:13.120 --> 0:01:16.280 to know two things. First, what happened a few years 0:01:16.319 --> 0:01:19.319 ago that led to this sudden explosion of AI content 0:01:19.440 --> 0:01:23.040 and apps and services. Do we suddenly discover a magical 0:01:23.120 --> 0:01:27.600 algorithm or alien artifact that suddenly gives machines this much power. 0:01:28.720 --> 0:01:31.720 And second, I wanted to know what effect all this 0:01:31.880 --> 0:01:36.240 fake AI made content has on our brains and how 0:01:36.280 --> 0:01:39.840 we see the world. So we'll answer both questions, and 0:01:39.920 --> 0:01:43.360 we'll start with the first. What happened a few years 0:01:43.360 --> 0:01:47.240 ago that started this recent explosion in AI? And is 0:01:47.280 --> 0:01:50.240 there a limit to it? To answer these questions, I 0:01:50.360 --> 0:01:52.640 reached out to this guy I met at lunch the 0:01:52.680 --> 0:01:57.520 other day, who I'm pretty sure is human. Well, thank 0:01:57.560 --> 0:01:59.080 you doctor Tam for joining us. 0:01:59.200 --> 0:02:01.640 Thank you JORGEV for having me. So my name is 0:02:01.680 --> 0:02:06.080 doctor ed Tam. I am a postdoctoral fellow at Stanford University. 0:02:06.320 --> 0:02:09.560 A lot of my work concerns the uncertainty around AI 0:02:10.120 --> 0:02:13.040 and what are the limitations of AI and how we 0:02:13.120 --> 0:02:14.959 can potentially overcome those. 0:02:15.000 --> 0:02:16.960 And how can I make sure you're real and not 0:02:17.040 --> 0:02:17.440 an AI. 0:02:19.280 --> 0:02:23.440 Well, we've met in person, so you know that. 0:02:23.919 --> 0:02:28.800 I met someone who looks like you a role I know, 0:02:28.880 --> 0:02:30.400 although it was over at lunch. 0:02:30.520 --> 0:02:33.960 So yes, a robot that can eat and joke and drink. 0:02:36.120 --> 0:02:38.440 Okay, doctor ed Tam is going to tell me what 0:02:38.560 --> 0:02:41.960 happened a few years ago that made AI swimmingly explode 0:02:41.960 --> 0:02:44.280 in what it's able to do. But to do that, 0:02:44.520 --> 0:02:46.760 we have to go back to where it all started, 0:02:47.040 --> 0:02:50.280 which is when scientists had the idea for AI in 0:02:50.360 --> 0:02:51.040 the first place. 0:02:52.960 --> 0:02:56.160 It goes back to like the last century at least. 0:02:56.400 --> 0:02:59.200 So the idea is that the human brain is composed 0:02:59.200 --> 0:03:02.640 of cells that are called neurons, and these neurons connect 0:03:02.639 --> 0:03:05.760 with each other and these different productivity patterns in the 0:03:05.800 --> 0:03:08.760 brain gives us the rich functionality of the human brain. 0:03:09.000 --> 0:03:12.600 So back in the nineteen hundred, scientists we're thinking, can 0:03:12.639 --> 0:03:18.120 we somehow mimic the biological human neural network with a computer, 0:03:19.000 --> 0:03:21.160 And so there are some early models that kind of 0:03:21.200 --> 0:03:23.240 try to tries to do this. One of the earliest 0:03:23.280 --> 0:03:25.920 and famous one is called the perceptron, and it's just 0:03:25.960 --> 0:03:28.520 this kind of like three layer no network with some 0:03:29.040 --> 0:03:31.720 nonlinear function and coded. So that was kind of like 0:03:31.760 --> 0:03:34.360 the first model, but it was already kind of like 0:03:34.400 --> 0:03:35.480 a mini breakthrough. 0:03:35.880 --> 0:03:38.200 It sounds like a transformer like the toys. 0:03:38.640 --> 0:03:42.400 Oh yes, yes, megatron. 0:03:43.680 --> 0:03:45.400 I don't know why they call it the perceptron. 0:03:46.400 --> 0:03:50.080 So, starting in the nineteen forties, scientists and engineers had 0:03:50.120 --> 0:03:52.720 an idea of how the human brain worked. It was 0:03:52.760 --> 0:03:55.840 a humongous network of neurons connected to each other, and 0:03:55.880 --> 0:03:58.720 so the scientists and engineers wondered if we could make 0:03:58.880 --> 0:04:03.320 an artificial brain. So they programmed on a computer how 0:04:03.360 --> 0:04:07.360 a neuron reacts. It takes signals from other neurons, weighs 0:04:07.400 --> 0:04:10.840 each input depending on some numbers or parameters, and then 0:04:11.040 --> 0:04:13.200 if it all adds up to a certain amount, it 0:04:13.280 --> 0:04:17.400 fires off a single signal to other neurons. And at 0:04:17.440 --> 0:04:21.040 first these networks with layers of neurons were pretty simple, 0:04:21.400 --> 0:04:28.800 but amazingly they could easily recognize patterns, meaning like recognize 0:04:28.960 --> 0:04:31.640 numbers or images kind. 0:04:31.520 --> 0:04:35.160 Of that's exactly right, like telling apart of human digits, 0:04:35.360 --> 0:04:37.159 can you tell a part a cat from a dog. 0:04:37.200 --> 0:04:39.680 There are all these image benchmark data sets that people 0:04:39.839 --> 0:04:42.200 work on, and then it turns out when you stack 0:04:42.279 --> 0:04:45.080 them on top of each other, it can encode really 0:04:45.160 --> 0:04:46.200 really rich patterns. 0:04:47.720 --> 0:04:50.960 So things progress somewhat slowly. Over the next few decades, 0:04:51.279 --> 0:04:53.960 people started to make bigger and bigger networks of these 0:04:54.040 --> 0:04:57.520 artificial neurons. Instead of a few neurons, it started to 0:04:57.520 --> 0:05:01.200 become dozens, and then hundreds, and then thousands of these 0:05:01.200 --> 0:05:05.240 neurons and dozens of these layers, and it was pretty powerful. 0:05:05.760 --> 0:05:08.240 You could give the machine a whole bunch of images 0:05:08.279 --> 0:05:12.240 of dogs and they could learn to recognize dogs. You 0:05:12.240 --> 0:05:14.920 could give it a whole bunch of human fingerprints or faces, 0:05:15.160 --> 0:05:18.200 and they could recognize fingerprints and faces. 0:05:18.760 --> 0:05:21.400 So that was maybe in like the twenty tens or 0:05:21.400 --> 0:05:24.280 something like that. When you go on Facebook, you know how, 0:05:24.360 --> 0:05:27.640 if you post a picture on Facebook, it actually tags 0:05:27.640 --> 0:05:30.360 your face and it tags the faces of your friends, 0:05:30.880 --> 0:05:32.560 and I remember that being like magic. 0:05:32.680 --> 0:05:34.960 I was like, wow, like it recognizes me. 0:05:35.880 --> 0:05:38.440 And that's the type of AI back then ten years 0:05:38.480 --> 0:05:39.840 ago that can do that. 0:05:41.000 --> 0:05:45.720 So ten years ago, AIS were really good at recognizing things, 0:05:46.120 --> 0:05:49.039 but that's kind of all they could do. They couldn't 0:05:49.040 --> 0:05:52.359 make anything up or talk to you or write anything 0:05:52.440 --> 0:05:56.400 for you. They were smart, but they were passive. But 0:05:56.400 --> 0:06:00.040 then in twenty seventeen, some engineers at Google discovered a 0:06:00.200 --> 0:06:05.000 special way to make these networks of neurons that changed everything. 0:06:05.680 --> 0:06:07.800 And it kind of started when they said, you know what, 0:06:08.200 --> 0:06:11.719 maybe we should stop trying to copy the human brain. 0:06:12.480 --> 0:06:14.840 So at the very beginning, people try to come up 0:06:14.839 --> 0:06:18.640 with very simple computer models, like the perceptron that tries 0:06:18.680 --> 0:06:22.000 to mimic the human brain, and then gradually this neural 0:06:22.120 --> 0:06:26.320 network or deep learning morphed into something completely different that 0:06:26.560 --> 0:06:28.520 is no longer trying to mimic the human brain. 0:06:29.839 --> 0:06:32.760 Yeah, you can imagine that if the History of the 0:06:32.839 --> 0:06:35.440 Human Race was a movie, this might be the part 0:06:35.480 --> 0:06:38.599 of the movie that foreshadows that it's not gonna end. Well, 0:06:42.760 --> 0:06:46.279 what the engineers at Google had invented was something called 0:06:46.600 --> 0:06:50.919 the transformer. In this case, it's not related to the 0:06:51.080 --> 0:06:51.760 robot toys. 0:06:51.960 --> 0:06:54.520 It is not related to the robot toys at all. 0:06:55.400 --> 0:06:56.880 I guess I've been in the AI. 0:06:56.680 --> 0:06:59.200 World for so long that transformers used to me in 0:06:59.240 --> 0:07:01.360 the movie and me Tron and all that souff now 0:07:01.440 --> 0:07:03.039 means something completely different to me. 0:07:03.880 --> 0:07:05.159 They reprogrammed your brain. 0:07:05.320 --> 0:07:08.880 They reprogrammed my brain. Yeah, okay. 0:07:09.000 --> 0:07:12.480 What this transformer way of connecting neural networks that the 0:07:12.520 --> 0:07:16.120 Google engineers invented in twenty seventeen does is that it 0:07:16.280 --> 0:07:21.080 uses an old ticnique from language translation called attention. And 0:07:21.240 --> 0:07:25.080 basically what it means is that when an AI is learning, say, 0:07:25.240 --> 0:07:27.760 a piece of text from a book, it doesn't just 0:07:27.880 --> 0:07:31.720 learn what each word means. It learns the context of 0:07:31.760 --> 0:07:34.720 the word. For each word, it sort of learns what 0:07:34.800 --> 0:07:38.560 word it usually has after it and before it, and 0:07:38.640 --> 0:07:41.480 it also learns what word it usually has two words 0:07:41.520 --> 0:07:44.520 after it and two words before it, and so on 0:07:44.920 --> 0:07:47.560 and so on. Let's here's an example. So like the 0:07:47.560 --> 0:07:50.880 word sit from sitting, usually it comes a few words 0:07:50.920 --> 0:07:52.560 before chair or sofa. 0:07:52.320 --> 0:07:55.720 Right right right, So I sit on a sofa. I 0:07:55.760 --> 0:07:58.920 sit on a sofa is very logical, very sensible. The 0:07:59.040 --> 0:08:02.480 key work that is important in terms of me understanding 0:08:02.600 --> 0:08:05.600 what sofa means is the I in the sentence. So 0:08:05.640 --> 0:08:07.280 it's not a dog sitting on the sofa, it's not 0:08:07.440 --> 0:08:10.679 cat sitting on the sofa, it's I that is sitting 0:08:10.720 --> 0:08:13.560 on the sofa. And so the attention mechanism allows you 0:08:13.840 --> 0:08:17.040 to kind of learn that type of correlation. And there 0:08:17.080 --> 0:08:20.720 are technical details about position encodings, but the main thing 0:08:20.800 --> 0:08:22.880 is just you want to learn the meaning of a 0:08:22.920 --> 0:08:27.080 word in the context of the words that surround it. 0:08:27.160 --> 0:08:30.000 This sort of simple and kind of common sense sounding 0:08:30.040 --> 0:08:34.320 trick is basically what revolutionized AI. You can apply it 0:08:34.360 --> 0:08:42.400 to text, images, music videos, and it's what makes chat GPT, Gemini, Grock, 0:08:42.800 --> 0:08:46.800 mid Journey, Claude, Nano, Banana, Copilot and all those AI 0:08:46.880 --> 0:08:51.080 agents out there work. And that's because using this transformer 0:08:51.200 --> 0:08:54.640 architecture and paying attention to the context of the data 0:08:55.040 --> 0:08:57.640 clout AIS to teach themselves. 0:09:00.120 --> 0:09:02.680 Yeah, so it's called self supervised learning. 0:09:03.040 --> 0:09:04.160 Oh that sounds scary. 0:09:04.559 --> 0:09:09.640 Yeah. Yeah. So traditional machine learning or AI systems, they 0:09:09.640 --> 0:09:13.520 rely on supervised learning, which means that a human has 0:09:13.559 --> 0:09:16.280 to tell you what is right or what is wrong. Now, 0:09:16.320 --> 0:09:20.160 while that's great, it's not very scalable because a human 0:09:20.200 --> 0:09:22.679 can only tell you what's right or wrong for a 0:09:22.800 --> 0:09:25.400 thousand things or two thousand things or whatnot. You can't 0:09:25.400 --> 0:09:28.360 scale it to a billion things. Whereas with this self 0:09:28.400 --> 0:09:33.280 supervised paradigm, you don't need human supervision. You just need 0:09:33.320 --> 0:09:38.280 some large purposes of unstructured data, and then the machine 0:09:38.320 --> 0:09:41.280 teaches itself just by predicting what the next token is 0:09:41.480 --> 0:09:44.360 and seeing whether that matches reality in terms of the 0:09:44.400 --> 0:09:45.480 real data. 0:09:45.679 --> 0:09:48.360 What doctor Tam saying is that this new type of 0:09:48.440 --> 0:09:52.319 AI could teach itself by predicting what the next word 0:09:52.760 --> 0:09:55.520 or part of an image should be according to its 0:09:55.559 --> 0:10:00.240 context and then checking with the real thing. And do 0:10:00.400 --> 0:10:05.080 ability of AIS to learn contexts and teach themselves, combined 0:10:05.120 --> 0:10:08.959 with big companies putting billions of dollars into huge computer 0:10:09.040 --> 0:10:13.200 farms meant that you could suddenly have ais teach themselves 0:10:13.480 --> 0:10:15.240 the entire Internet. 0:10:16.559 --> 0:10:21.040 So the idea is that very surprising things happen when 0:10:21.200 --> 0:10:25.600 you scale something to really really make sizes. They perform 0:10:25.880 --> 0:10:30.080 way way better in terms of producing realistic samples and texts. 0:10:30.679 --> 0:10:33.199 So there is some type of phase transition in terms 0:10:33.240 --> 0:10:36.920 of how intelligent or how realistic the samples that it 0:10:37.000 --> 0:10:39.079 generates will look like something. 0:10:39.920 --> 0:10:44.400 Click. It suddenly activated some potential in these systems that 0:10:44.720 --> 0:10:50.520 experts were not expecting. Right. So t wecap ais were 0:10:50.559 --> 0:10:53.719 originally meant to copy the human brain, and they were 0:10:53.800 --> 0:10:58.480 really good at just recognizing things patterns, faces, numbers. But 0:10:58.520 --> 0:11:02.280 then some engineers at Google decided, you know what, let's 0:11:02.320 --> 0:11:05.800 forget the human brain and just push this through the limit. 0:11:06.120 --> 0:11:08.680 And what they came up with created ais they could 0:11:08.840 --> 0:11:14.040 understand context, and more importantly, they could teach themselves. And 0:11:14.200 --> 0:11:17.800 once that happened, they scaled up, gubbled up the Internet 0:11:18.000 --> 0:11:20.680 and made a leap that's making a lot of people, 0:11:21.240 --> 0:11:25.320 uh concerned right now. So when we come back, we're 0:11:25.360 --> 0:11:28.360 going to talk about how these new ais can affect 0:11:28.400 --> 0:11:31.480 your brain, and how they could even make endangered species 0:11:31.520 --> 0:11:35.800 of animals go to think, don't go anywhere, We'll be 0:11:36.000 --> 0:11:51.720 right back. Hey, welcome back. We're talking about AI slop 0:11:52.120 --> 0:11:55.080 or AI made images and videos that are swamping the 0:11:55.120 --> 0:11:58.680 Internet right now, and what effect it's having on us 0:11:59.280 --> 0:12:02.840 so far. Talking about what happened in twenty seventeen that 0:12:03.040 --> 0:12:06.280 suddenly made these AI systems be able to talk to 0:12:06.320 --> 0:12:09.520 you and make the kind of realistic images everyone sees 0:12:09.640 --> 0:12:12.920 in their social media feeds. Now we're gonna talk about 0:12:12.920 --> 0:12:16.200 the effect it's having on our brains. Is it possible 0:12:16.240 --> 0:12:20.720 for AI generated content to rewire how we think? To 0:12:20.760 --> 0:12:22.559 tell us about that, I reached out to my friend, 0:12:22.760 --> 0:12:27.240 neuroscientist Dwayne Godwin, a professor at Wake Forest University and 0:12:27.280 --> 0:12:29.560 the co author of the book Out of Your Mind, 0:12:29.840 --> 0:12:33.160 which he co wrote with me. Doctor Godwin recently wrote 0:12:33.160 --> 0:12:35.960 an article about this topic online, so I thought he 0:12:36.080 --> 0:12:38.920 was the perfect person to give us the neural perspective 0:12:39.080 --> 0:12:45.760 on this topic. Here's my conversation with doctor Dwayne Godwin. Hey, Dwayne, 0:12:45.760 --> 0:12:46.200 how are you? 0:12:46.559 --> 0:12:48.320 I'm doing well? Orgey, how are you? 0:12:48.800 --> 0:12:50.920 I'm good? Now? Just to make sure this is the 0:12:50.960 --> 0:12:53.040 real you, right, I'm not talking to an AI version 0:12:53.080 --> 0:12:53.600 of you. 0:12:53.640 --> 0:12:55.880 No, I AMZ I'm the real deal man. 0:12:57.679 --> 0:13:00.959 And is there an intelligence artificial or is organic? 0:13:01.200 --> 0:13:04.000 It's augmented by caffeine. 0:13:05.320 --> 0:13:08.920 Augmented intelligence? There you go. Well, I wonder if there's 0:13:08.920 --> 0:13:11.240 a way for people to know if this conversation is 0:13:11.320 --> 0:13:14.400 generated by one of those you know, chat GBT things 0:13:14.400 --> 0:13:15.520 that mix podcasts. 0:13:15.679 --> 0:13:17.760 You know, if we were really sneaky, we might have 0:13:18.000 --> 0:13:20.760 generated an AI podcast of the two of us going 0:13:20.800 --> 0:13:24.880 back and forth and then tested our audience in their 0:13:24.920 --> 0:13:28.199 ability to tell the difference. But we're not that sneaky. 0:13:28.720 --> 0:13:29.600 Of course, this is. 0:13:29.520 --> 0:13:30.120 The real thing. 0:13:30.240 --> 0:13:32.320 All right, let's get into it. I guess the first 0:13:32.400 --> 0:13:34.960 question I had was how good are we at telling 0:13:35.080 --> 0:13:36.640 is something is AI generated? 0:13:36.960 --> 0:13:41.400 We're not very good. People tend to overestimate their AI radar. 0:13:41.760 --> 0:13:45.600 A lot of obvious tells, like you know, extra fingers, 0:13:45.920 --> 0:13:49.920 robotic phrasing, those are getting all smoothed out. And we're 0:13:49.960 --> 0:13:53.520 looking at these things often on very tiny screens, and 0:13:53.640 --> 0:13:58.280 so what's plausible quickly becomes good enough or believable wow, 0:13:58.440 --> 0:14:01.160 And the result of that is there's kind of a 0:14:01.200 --> 0:14:04.640 confidence gap. We feel very certain that we can tell 0:14:04.720 --> 0:14:07.640 fake stuff when we see it, but often we're not 0:14:07.679 --> 0:14:08.920 doing much better than chance. 0:14:09.320 --> 0:14:11.800 Are there any studies that kind of tell us how 0:14:11.880 --> 0:14:14.319 good we are at spotting fake ais? 0:14:14.520 --> 0:14:16.959 Yeah, there have been studies that have been done. There 0:14:17.080 --> 0:14:20.680 was a study where people are essentially two groups. They're 0:14:20.800 --> 0:14:25.440 presented with both real images and AI generated images, and 0:14:25.480 --> 0:14:28.400 it turns out that they're about fifty percent you know, 0:14:28.440 --> 0:14:30.800 it's like fifty to fifty that they can tell the difference. 0:14:30.920 --> 0:14:32.560 Basically, they can't tell the difference. 0:14:32.720 --> 0:14:35.400 Yeah, so they're really about at the level of chance. 0:14:35.680 --> 0:14:39.080 Yikes. Okay, and this was probably done a few couple 0:14:39.120 --> 0:14:40.080 of years ago, right. 0:14:40.160 --> 0:14:43.080 Yeah, that's right, And so things are only better now. 0:14:43.160 --> 0:14:47.320 So you can imagine that right now, with the quality 0:14:47.560 --> 0:14:50.760 of generative AI, that this is even going to be 0:14:50.800 --> 0:14:52.480 more true. 0:14:52.760 --> 0:14:55.000 Yes, we all like to think we're good at spotting 0:14:55.040 --> 0:14:58.600 fake images or videos. I mean, we've all seen bad 0:14:58.600 --> 0:15:01.920 computer graphic effects movies, so we think we know what 0:15:02.040 --> 0:15:04.920 fake looks like. But actually the sign says that we 0:15:04.960 --> 0:15:08.360 are not much better than guessing randomly if something is 0:15:08.400 --> 0:15:12.160 fake or not. And according to doctor Godwin, there are 0:15:12.320 --> 0:15:16.240 three reasons our brains are kind of helpless against this 0:15:16.440 --> 0:15:20.280 onslaught of fake content coming at us on a daily basis. 0:15:20.920 --> 0:15:23.720 The first reason is that your brain sort of wants 0:15:23.760 --> 0:15:28.600 to believe things are real. I guess, as a neuroscientists 0:15:28.600 --> 0:15:31.120 and as a psychologist, why do you think our brains 0:15:31.240 --> 0:15:32.600 are so easy to fool? 0:15:33.040 --> 0:15:35.640 Well, a couple of reasons. You know, our brains are 0:15:35.680 --> 0:15:40.880 built for speed and survival and not for fact checking. 0:15:41.080 --> 0:15:43.360 Right, And I guess that means out in the wild, 0:15:43.520 --> 0:15:45.880 if you think you see a bear or a tiger 0:15:45.960 --> 0:15:47.840 coming at you, your brain is not going to have 0:15:47.840 --> 0:15:49.920 wanted like fact check that there really is a tiger 0:15:49.960 --> 0:15:52.400 coming at you. You want your brain to just react 0:15:52.400 --> 0:15:54.400 to the idea that maybe there's a tiger coming at you. 0:15:54.520 --> 0:15:57.720 Yes, that's sedaptive. You're going to go with bear if 0:15:57.720 --> 0:15:59.000 it looks anything. 0:16:00.640 --> 0:16:02.920 Good. Because I guess we don't have all information all 0:16:02.960 --> 0:16:05.800 the time, so our brains are sort of have trained to, like, 0:16:05.960 --> 0:16:07.840 do as much as you can with as little information 0:16:07.880 --> 0:16:08.480 as possible. 0:16:08.600 --> 0:16:11.880 Yeah, there's a concept in psychology called heuristics, which is 0:16:11.960 --> 0:16:16.120 basically making decisions on the basis of very limited information, 0:16:16.200 --> 0:16:19.000 and our brains are doing that constantly, so you know, 0:16:19.040 --> 0:16:21.200 we have to make those very quick decisions, so we 0:16:21.240 --> 0:16:23.560 can pass our genes onto the next generation. 0:16:24.040 --> 0:16:26.200 I see, and so you can exploit that. I guess 0:16:26.280 --> 0:16:28.960 because we see something that sort of looks real, your 0:16:28.960 --> 0:16:31.520 mind will basically make the conclusion that it is real. 0:16:31.720 --> 0:16:34.480 Yeah, it will create the story that it is real. 0:16:34.720 --> 0:16:36.960 You know, what is the purpose of our brains. Our 0:16:37.000 --> 0:16:41.080 brains are prediction machines, and so we've learned certain facts 0:16:41.120 --> 0:16:44.520 about the world, about how things normally go. So if 0:16:44.520 --> 0:16:47.280 you're in the woods and you encounter something it seems 0:16:47.360 --> 0:16:49.960 like a bear, your brain already knows that that's probably 0:16:50.000 --> 0:16:52.440 not a great thing, and so you're going to extrapolate 0:16:52.560 --> 0:16:55.640 that story and you're going to want to avoid that 0:16:55.680 --> 0:16:57.600 bear as much as possible, right. 0:16:57.760 --> 0:17:01.360 I guess we kind of had a natural mechanism against that, 0:17:01.840 --> 0:17:04.320 which was this idea of the Uncanny Valley. 0:17:04.520 --> 0:17:07.240 Yeah, I'm familiar with the Uncanny Valley. It's that sort 0:17:07.280 --> 0:17:09.639 of eerie feeling that you get when you see something 0:17:09.680 --> 0:17:11.760 that's trying to be real but it's not real. It 0:17:11.800 --> 0:17:14.440 makes me uncomfortable because I know that it's not supposed 0:17:14.480 --> 0:17:15.960 to look like that exactly. 0:17:16.119 --> 0:17:18.199 Yeah, So, like we did have a brain mechanism to 0:17:18.280 --> 0:17:22.000 defense against fake things. Right, But it seems like AI 0:17:22.040 --> 0:17:25.440 today has basically blown past that defense, like it's now 0:17:25.920 --> 0:17:27.320 past the uncanny valley. 0:17:27.600 --> 0:17:31.119 Yeah, it has jumped across the uncanny valley for sure. 0:17:32.240 --> 0:17:35.520 In other words, our brain has a built in BS 0:17:35.560 --> 0:17:39.479 meter called the uncanny valley. But now AI content is 0:17:39.520 --> 0:17:42.600 so good it's blown past that last line of defense 0:17:42.840 --> 0:17:45.800 or your brain. Okay. The second reason our brains are 0:17:45.880 --> 0:17:49.640 easy to fool by AI generated content is that it's 0:17:49.680 --> 0:17:52.639 not built to withstand the constant stream of it we 0:17:52.800 --> 0:17:56.879 get every day. The second thing you said was repetition. 0:17:57.200 --> 0:18:00.560 Yeah, that's right. So your resistance is being reduced in 0:18:00.640 --> 0:18:03.640 two ways. So first is just the repetition of seeing 0:18:03.720 --> 0:18:05.280 the same thing over and over. 0:18:05.640 --> 0:18:08.919 Meaning like, if I'm scrolling through these online feeds and 0:18:09.359 --> 0:18:11.520 most of what I see has that sort of feeling 0:18:11.520 --> 0:18:13.959 of an AI generated thing, then my brain is just 0:18:14.000 --> 0:18:16.439 gonna normalize it. Is that kind of what you're saying, Like, 0:18:16.480 --> 0:18:17.520 it's yeah, it is. 0:18:17.800 --> 0:18:21.080 There's another more insidious part of that, which is, let's 0:18:21.200 --> 0:18:24.680 use the you know, the cute animal AI generated cute 0:18:24.720 --> 0:18:29.040 animal videos you'll see in the comments underneath those people 0:18:29.119 --> 0:18:32.840 reacting to it as if it's real, and so part 0:18:32.920 --> 0:18:36.920 of that is, Hey, everybody else thinks it's real, and 0:18:37.119 --> 0:18:39.879 I'm a person too, shouldn't I be thinking that this 0:18:40.040 --> 0:18:43.840 is real? And so there's this sort of desire on 0:18:43.960 --> 0:18:47.399 the part of people to reach a consensus with your group, 0:18:48.080 --> 0:18:50.879 and so if you start identifying with that group, then 0:18:51.000 --> 0:18:54.640 your brain might be bent toward making a decision that, hey, 0:18:54.760 --> 0:18:57.760 maybe this is real, even if you might have had 0:18:57.800 --> 0:19:01.040 initial doubts. Yeah, maybe you start to fall into line. 0:19:01.720 --> 0:19:04.159 Like we all like to think we're independent thinkers and 0:19:04.200 --> 0:19:07.159 critical thinkers, but part of our brain is sort of 0:19:07.240 --> 0:19:09.359 wired to kind of basically follow the mob. 0:19:09.800 --> 0:19:12.639 Yeah, a little bit of mob mentality. Yeah, we like 0:19:12.680 --> 0:19:17.040 to think of ourselves as logical beings, but the reality 0:19:17.200 --> 0:19:22.840 is we're all challenged by biases, and those biases can 0:19:22.960 --> 0:19:26.560 be very adaptive in the case of things like heuristics. 0:19:27.840 --> 0:19:30.840 And then the third reason our brains are easy targets 0:19:30.880 --> 0:19:33.520 for AI slob is that a lot of that SLOB 0:19:33.680 --> 0:19:37.280 is made to play with our emotions. And then you 0:19:37.280 --> 0:19:40.639 said the third one was that we tend to believe 0:19:40.760 --> 0:19:42.920 things that kind of affect our emotions. 0:19:43.320 --> 0:19:46.879 Yes, that's probably one of the most insidious things, is 0:19:46.920 --> 0:19:52.120 that things that ring our emotional bell tend to make 0:19:52.160 --> 0:19:56.479 it through without the kind of critical thought processing that 0:19:56.520 --> 0:20:00.960 we would normally apply to something that's just factual or techation. 0:20:01.880 --> 0:20:06.640 So think about politics. Politics is very polarizing. So if 0:20:06.720 --> 0:20:10.200 you see something in your social media feed and it 0:20:10.359 --> 0:20:14.879 confirms what your emotions tell you should be true based 0:20:14.920 --> 0:20:18.639 on your political leanings, then that we'll get through the 0:20:18.720 --> 0:20:23.080 normal cognitive criticism that you would apply to that information. 0:20:23.800 --> 0:20:26.640 Yikes, Like, we let our emotions get in the way 0:20:26.680 --> 0:20:31.199 of our critical thinking, so our brains are ill equipped 0:20:31.280 --> 0:20:34.320 to deal with all this AI slot because it tends 0:20:34.359 --> 0:20:37.560 to believe what it sees. We're inundated with this content 0:20:38.000 --> 0:20:41.600 and it's all meant to trigger our emotions and things 0:20:41.720 --> 0:20:45.040 doctor Godwin says are just getting worse. 0:20:46.400 --> 0:20:50.840 They're getting worse because it's getting better. There's a lot 0:20:50.960 --> 0:20:55.600 of money being put into making these things basically indistinguishable 0:20:55.600 --> 0:20:59.160 from reality, and we were already suffering under the idea 0:20:59.359 --> 0:21:02.880 that we could possibly tell the difference, but it's becoming 0:21:02.920 --> 0:21:05.600 a parent that we cannot, and it's going to be 0:21:05.680 --> 0:21:06.440 even harder. 0:21:06.920 --> 0:21:09.120 It kind of seems like it's also getting worse because 0:21:09.480 --> 0:21:12.480 it seems like people are more willing to post fake 0:21:12.600 --> 0:21:16.280 AI generated things, like just the other day, the White 0:21:16.359 --> 0:21:20.800 House official account posted some AI altered image of some 0:21:20.960 --> 0:21:24.840 activist protester, and they were totally unapologetic about it. 0:21:25.200 --> 0:21:29.679 Yeah, the tools are more readily available, and your ability 0:21:29.720 --> 0:21:32.600 to manipulate those tools doesn't even require that you have 0:21:32.680 --> 0:21:34.480 any sort of computer science degree. 0:21:34.840 --> 0:21:36.480 Yeah, it's just an app on your phone. 0:21:36.880 --> 0:21:39.600 Yeah. This stuff is scalable too, So it's not just 0:21:39.720 --> 0:21:41.639 that a person can go in and do like a 0:21:41.680 --> 0:21:45.159 one off. A single person could generate a lot of 0:21:45.200 --> 0:21:48.800 this content and get it out there, and even use 0:21:49.040 --> 0:21:53.119 AI to generate bots that would distribute that content for you. 0:21:53.760 --> 0:21:54.040 Yeah. 0:21:54.119 --> 0:21:56.480 So it's only going to get worse, I think. 0:21:57.520 --> 0:22:00.040 Yeah, it's not looking so good for us organic and 0:22:00.080 --> 0:22:04.639 they naturally made squishy intelligences out here. But according to 0:22:04.680 --> 0:22:07.679 doctor Godwin, there are some things you can do to 0:22:07.720 --> 0:22:11.480 protect your neurons. What can we do to protect our 0:22:11.560 --> 0:22:14.800 brains from the defect of this AI slut? 0:22:15.240 --> 0:22:20.119 Well, it's hard. I wish I had an easy answer, 0:22:20.160 --> 0:22:23.240 because what I've told you so far it's pretty hopeless. Right, 0:22:23.440 --> 0:22:24.480 it sounds hopeless. 0:22:24.640 --> 0:22:26.800 Maybe I should look it up and chat GBT. 0:22:26.720 --> 0:22:28.680 But I think there are things that we can do, 0:22:29.040 --> 0:22:31.600