WEBVTT - Large Language Models and You 0:00:01.480 --> 0:00:07.200 Welcome to Stuff You Should Know, a production of iHeartRadio. 0:00:11.160 --> 0:00:13.920 Hey, and welcome to the podcast. I'm Josh, and there's 0:00:14.080 --> 0:00:17.880 Chuck and Jerry's here too, And that makes this a 0:00:18.480 --> 0:00:22.759 timely topical Not timely is in fork times. I meant 0:00:22.760 --> 0:00:27.159 to say timely topical episode of Stuff you should Know. 0:00:27.760 --> 0:00:29.520 That's a great forecast of. 0:00:29.480 --> 0:00:32.120 How it's going to go to I think fork cast. 0:00:33.720 --> 0:00:39.559 Oh God, is this really us or is it AI 0:00:39.800 --> 0:00:41.520 generated Josh and Chuck. 0:00:42.479 --> 0:00:48.159 Until this year I would have been like, don't be preposterous. 0:00:48.240 --> 0:00:50.280 Now I'm like, just give it some time. 0:00:51.080 --> 0:00:53.200 You know how we would know as if it said 0:00:53.240 --> 0:00:56.200 if one of us said, of course it's the real Us. 0:00:56.520 --> 0:00:59.480 We met in the office and bonded over our Van 0:00:59.560 --> 0:01:01.560 Halen denim vests. 0:01:01.840 --> 0:01:04.640 Yeah, we'd be like, sucker, you just fell for the 0:01:04.680 --> 0:01:07.440 oldest trap in the book, the Sicilian switcher. 0:01:08.160 --> 0:01:11.520 Yeah, aka fake Wikipedia entry stuff? 0:01:12.080 --> 0:01:12.880 Is that still up? 0:01:13.840 --> 0:01:16.120 I haven't been to our Wikipedia page in years, so 0:01:16.200 --> 0:01:16.640 I don't know. 0:01:17.480 --> 0:01:19.920 Well, regardless, we're not talking about Wikipedia, although it does 0:01:20.040 --> 0:01:22.440 kind of fall into the sure it figures in the 0:01:22.520 --> 0:01:25.360 rubric of this. Not sure if I use that word correctly, 0:01:25.440 --> 0:01:29.800 but it felt right. We're talking today about what are 0:01:30.480 --> 0:01:34.679 in the biz known as large language models, but more 0:01:34.760 --> 0:01:38.480 colloquially known by basically their their public facing names, things 0:01:38.480 --> 0:01:44.360 like chat, GPT or barred or being AI. But essentially 0:01:44.400 --> 0:01:49.920 what they all are are algorithms, artificially intelligent algorithms that 0:01:50.440 --> 0:01:54.720 are trained on text, tons and tons and tons of 0:01:54.760 --> 0:02:00.200 text written English language stuff, that are so good at 0:02:00.280 --> 0:02:04.200 recognizing patterns in those things that they can actually simulate 0:02:04.280 --> 0:02:07.080 a conversation with you, the person on the other side 0:02:07.080 --> 0:02:09.320 of the computer, asking them questions. 0:02:10.120 --> 0:02:13.400 Yeah, this is it's gonna be fun doing this episode 0:02:13.880 --> 0:02:19.120 over every six months, right until we're replaced totally. 0:02:19.600 --> 0:02:21.520 So I think we should say, though, like this is 0:02:21.600 --> 0:02:24.320 we're going to like this is such a huge wide 0:02:24.960 --> 0:02:30.120 topic that is just we're in the ignition phase, like 0:02:30.120 --> 0:02:34.480 the fuse just caught, right, Yeah, that we're going to 0:02:34.560 --> 0:02:36.640 really try to keep it narrow just strictly to large 0:02:36.680 --> 0:02:39.760 language models and the immediate effect they're planning on having 0:02:39.880 --> 0:02:42.480 or they're going to have, hopefully not planning on anything yet, 0:02:43.680 --> 0:02:46.160 but I really would like to do one on how 0:02:46.240 --> 0:02:50.320 to keep AI friendly and keeping it from running away. Yeah, 0:02:50.480 --> 0:02:52.600 I say, we just kind of avoid that whole kind 0:02:52.639 --> 0:02:55.040 of stuff. And really I'm talking to myself right now 0:02:55.320 --> 0:02:56.520 at least for this episode. 0:02:56.560 --> 0:02:58.800 Okay, Yeah, you know, we're going to kind of explain 0:02:59.320 --> 0:03:03.079 how these things were and what the initial applications look 0:03:03.240 --> 0:03:05.040 like and kind of where we are right now, and 0:03:05.080 --> 0:03:08.240 then what it could mean for like jobs and the 0:03:08.280 --> 0:03:10.600 economy and stuff like that. But you're right, it is 0:03:11.919 --> 0:03:13.600 as a whole ball of wax, as you well know. 0:03:14.320 --> 0:03:16.560 And this is a great time to plug the End 0:03:16.560 --> 0:03:19.000 of the World with Josh Clark, which is still out there. 0:03:19.040 --> 0:03:19.960 You can still listen to it. 0:03:20.600 --> 0:03:22.960 The truth is out there in the form of the 0:03:23.040 --> 0:03:24.440 End of the World with Josh Clark. 0:03:24.919 --> 0:03:28.680 Yeah, it's a great ten part series that you did, 0:03:29.240 --> 0:03:34.400 and AI is among those existential risks existential that you covered. 0:03:34.920 --> 0:03:38.040 Yeah, it's episode four, I believe. And Chuck like just 0:03:38.120 --> 0:03:40.920 from having done that research in forming my own opinions 0:03:40.920 --> 0:03:46.120 over the years about this. Yeah, Like I'm it's staggering 0:03:46.160 --> 0:03:50.000 to me that we're like we've just entered like what's 0:03:50.040 --> 0:03:58.440 going to be the most revolutionary transitional phase in the 0:03:58.720 --> 0:04:02.120 entire history of humanity. You can argue everything else took 0:04:02.160 --> 0:04:05.080 place over very long periods of time. We started playing 0:04:05.120 --> 0:04:07.280 with stone tools and then we started building cities. All 0:04:07.320 --> 0:04:09.960 this stuff took place over thousands and thousands, hundreds of thousands, 0:04:10.000 --> 0:04:13.120 millions of years. We just entered a period where stuff's 0:04:13.160 --> 0:04:17.120 going to start happening within weeks pretty soon. As of 0:04:17.120 --> 0:04:19.200 twenty twenty three, the whole thing just started. 0:04:19.680 --> 0:04:22.680 Yeah, and none of this was around like this when 0:04:22.720 --> 0:04:24.520 you did The End of the World and that was 0:04:24.880 --> 0:04:26.240 what like five years ago. 0:04:26.320 --> 0:04:28.679 Is Yeah, it was twenty eighteen. All this was being 0:04:28.720 --> 0:04:31.040 worked on, but we hadn't hit that point. Like all 0:04:31.080 --> 0:04:33.960 this was pretty much predicted and projected, and it was 0:04:34.200 --> 0:04:36.039 clear that this was the direction people were going. 0:04:36.120 --> 0:04:37.640 And it's here, baby, it is. 0:04:37.920 --> 0:04:40.760 It's nuts, but it's actually here. So we're talking about 0:04:40.760 --> 0:04:44.400 our large language models, which is a type of neural 0:04:44.480 --> 0:04:48.440 network that are easiest to think of in terms of 0:04:48.880 --> 0:04:51.600 like a human brain, where you have neurons that are 0:04:51.600 --> 0:04:53.960 connected to other neurons, but they're not connected to some 0:04:54.080 --> 0:04:58.080 other neurons, and all of those neural connections kind of 0:04:58.440 --> 0:05:02.800 are activated by input. It's that put out something like 0:05:02.839 --> 0:05:05.400 your conscious experience or you say a sentence or something 0:05:05.440 --> 0:05:09.880 like that. It's it's very similar in its most basic nature. 0:05:10.120 --> 0:05:13.720 I guess, yeah, I mean Olivia helped us out with this, 0:05:13.839 --> 0:05:17.120 and che did a great job. I think and Google 0:05:17.160 --> 0:05:20.400 themselves basically say, you know what it's really it's sort 0:05:20.400 --> 0:05:22.839 of like how when you go to search for something 0:05:22.960 --> 0:05:28.520 on our search engine tool here a weird way I 0:05:28.520 --> 0:05:31.120 could have said that, I don't think so are handy 0:05:31.160 --> 0:05:34.320 search for Then you know, basically what we're doing is 0:05:34.400 --> 0:05:40.080 auto completing, like an analysis of like probability, like like 0:05:40.200 --> 0:05:44.479 what you're typing. If you type in you know, John Coltrane, 0:05:44.520 --> 0:05:47.600 or start to chop type in John Cole, that might 0:05:47.640 --> 0:05:51.120 finish it out as John Coltrane a Love Supreme or 0:05:51.200 --> 0:05:54.719 John Coltrane jazz and they're saying, you know what, what 0:05:54.839 --> 0:05:59.920 is happening now with these lms is it's the same thing. 0:06:00.080 --> 0:06:03.279 It's just it's got way more data, way more calculations 0:06:03.880 --> 0:06:07.039 in the algorithm. So it's not just completing like a 0:06:07.080 --> 0:06:11.160 word or two. It's potentially you know, hey, rewrite the 0:06:11.200 --> 0:06:12.760 Bible or whatever you tell it to do. 0:06:13.279 --> 0:06:16.279 Yeah, And the big difference is in the amount of 0:06:16.560 --> 0:06:23.479 info the neural network is capable of taking into consideration. Yeah, 0:06:23.520 --> 0:06:26.880 so may I for a minute, oh please, So imagine 0:06:26.880 --> 0:06:29.599 with one of those autocomplete suggestion tools like they have 0:06:29.760 --> 0:06:34.359 on Google Search. If there's five hundred thousand words in 0:06:34.400 --> 0:06:37.159 the English language, that means that you have five hundred 0:06:37.200 --> 0:06:40.640 thousand words that a person could possibly put in. That's 0:06:40.680 --> 0:06:44.159 the input into the neural network, and then there's five 0:06:44.240 --> 0:06:47.599 hundred thousand possible words that that network could put out. 0:06:47.640 --> 0:06:50.440 So you have five hundred thousand connections to five hundred 0:06:50.440 --> 0:06:53.159 thousand other connections. So it's like, I think two hundred 0:06:53.160 --> 0:06:56.119 and fifty billion connections you're starting with right there. That's 0:06:56.279 --> 0:07:00.520 just the autocomplete suggestion because it based on those connections. 0:07:00.560 --> 0:07:03.480 In studying words in the English language and phrases in 0:07:03.520 --> 0:07:07.480 the English language, it places emphasis more on some connections 0:07:07.520 --> 0:07:12.800 than others. So John Coltrane is what's what's his album 0:07:12.800 --> 0:07:16.400 I Can't Remember is a Love Supreme classic album. So 0:07:16.520 --> 0:07:20.240 John Coltrane is much more closely related to a Love 0:07:20.280 --> 0:07:23.800 Supreme in the mind of a neural network than John Coltrane. 0:07:25.720 --> 0:07:29.560 Charlie Brown disco is right, just to take something off 0:07:29.600 --> 0:07:32.320 the top of my head. And so based on that 0:07:32.480 --> 0:07:35.200 analysis and that weight that it gives to some things 0:07:35.200 --> 0:07:38.400 other than others, it suggests those words what the large 0:07:38.480 --> 0:07:41.560 language models that like chat GPT, that we're seeing today. 0:07:42.400 --> 0:07:45.080 They do the same thing. They have all those same connections, 0:07:45.480 --> 0:07:48.080 but the analysis they do, the weight that they put 0:07:48.120 --> 0:07:52.560 on the connections is so much more advanced, yes, and 0:07:52.720 --> 0:07:57.000 exponential that it's it's actually not just capable of suggesting 0:07:57.040 --> 0:08:00.000 the next word, it's capable of holding a conversation with you. 0:08:00.080 --> 0:08:04.480 That's how much it understands how the English language works. 0:08:04.920 --> 0:08:08.360 Yeah, Like if you said, you know, write a story 0:08:08.400 --> 0:08:14.320 about wintertime, and you know, it got to the word snowy, 0:08:14.400 --> 0:08:17.160 it would it would go through you know, I mean, 0:08:17.200 --> 0:08:20.600 and this is like instantaneously it's doing these calculations. It 0:08:20.680 --> 0:08:24.200 might say like you know, oh, hillside or winter or 0:08:24.240 --> 0:08:26.280 snowy day, like these are all things that make sense 0:08:26.680 --> 0:08:29.680 because I've learned that that makes sense I being you know, 0:08:29.800 --> 0:08:33.760 the chatbot or whatever. But it probably won't be snowy 0:08:33.880 --> 0:08:37.599 chicken wing because that doesn't seem to fit the algorithm. 0:08:38.120 --> 0:08:43.080 And it learns all this stuff by reading the internet. 0:08:43.360 --> 0:08:46.680 And you know, put a pin in that, because that's 0:08:46.720 --> 0:08:51.040 pretty thorny for a whole lot of reasons, but not 0:08:51.120 --> 0:08:53.480 the least of which is the fact that some companies, 0:08:53.520 --> 0:08:55.040 and again we'll get to it, are starting to say 0:08:55.040 --> 0:08:57.440 like wait a minute, Like, we created this content and 0:08:57.480 --> 0:09:02.320 now you're just scrubbing it and then using it and 0:09:02.400 --> 0:09:05.240 charging people to use it, and we're not getting a 0:09:05.240 --> 0:09:07.880 piece of it. So that's just one tiny little thorn. 0:09:08.840 --> 0:09:10.920 But in order to do this, like you said, it's 0:09:10.960 --> 0:09:13.760 like it needs to know more, and you came up 0:09:13.800 --> 0:09:17.240 with a great example, like the word lemon. In a 0:09:17.440 --> 0:09:21.760 very basic way, it might understand that a lemon is 0:09:22.679 --> 0:09:26.280 roundish and sour and yellow. But if it needs to 0:09:26.280 --> 0:09:29.199 get smart enough to really write as if it were 0:09:29.240 --> 0:09:31.760 a human, it needs to know that it can make lemonade, 0:09:31.800 --> 0:09:35.680 and that it grows on a tree in these agricultural zones, 0:09:35.720 --> 0:09:38.040 and that it's a citrus fruit, because it has to 0:09:38.040 --> 0:09:43.200 be able to group lemon together with like things. And 0:09:43.280 --> 0:09:46.439 those groups are either like, you know, hey, it's super 0:09:46.440 --> 0:09:49.960 similar to this, like maybe other citrus fruits, or it's 0:09:50.000 --> 0:09:51.839 you know, sort of similar to this but not as 0:09:51.840 --> 0:09:55.760 similar as citrus fruits, like desserts. And then you get 0:09:55.760 --> 0:09:58.560 to chicken wings, although actually that's not true because lemon 0:09:58.800 --> 0:09:59.560 chicken wings. 0:10:00.120 --> 0:10:01.520 You could have lemon pepper chicken wings. 0:10:01.559 --> 0:10:03.720 Right, Yeah, that's what I'm saying, So yeah, Kau, But 0:10:04.040 --> 0:10:06.880 the instance you use is like greenland, which I guess 0:10:06.880 --> 0:10:08.360 doesn't grow lemons. 0:10:08.559 --> 0:10:10.800 No, but I mean, I'm sure they import lemons, so 0:10:10.800 --> 0:10:14.720 there's some connection there. But based on how connected, how 0:10:14.800 --> 0:10:17.880 often these words show up together, and the billions and 0:10:17.920 --> 0:10:21.440 billions of lines of texts that these language large language 0:10:21.440 --> 0:10:25.440 models are trained on, it starts to get more and 0:10:25.440 --> 0:10:29.360 more dimensions and making more and more connections. Right. So 0:10:30.400 --> 0:10:34.080 as that happens, words start to cluster together, like lemon 0:10:34.240 --> 0:10:36.880 and pie and ice box all kind of cluster together. 0:10:37.440 --> 0:10:42.800 And by taking words and understanding how they connect to 0:10:42.880 --> 0:10:46.280 other words, you can take the English language, just the 0:10:46.280 --> 0:10:48.760 words of the English language, and make meaning out of it. 0:10:48.880 --> 0:10:51.920 That's what that's all we do. And large language models 0:10:51.960 --> 0:10:54.800 are capable of doing the same thing. But it's really 0:10:55.000 --> 0:10:59.640 really important for you to understand that the large language 0:10:59.679 --> 0:11:04.880 model doesn't understand what it's doing. It doesn't have any 0:11:04.920 --> 0:11:10.280 meaning to the word lemon whatsoever. All of these dimensions 0:11:10.320 --> 0:11:13.679 that it waits to decide whether what word it should 0:11:13.760 --> 0:11:18.800 use next, they're called embeddings. They're just numerical representations. Yeah, 0:11:18.800 --> 0:11:21.040 So the higher the number, the likelier it is it 0:11:21.080 --> 0:11:23.320 goes with the word that the user just put in 0:11:23.400 --> 0:11:26.679 or that the large language model. Just use the lower 0:11:26.720 --> 0:11:29.360 the number, the further away it is in the cluster, right, 0:11:30.000 --> 0:11:33.480 it doesn't understand what it's saying to you. And as 0:11:33.520 --> 0:11:36.240 we'll see later, that accounts for a phenomenon that we're 0:11:36.280 --> 0:11:38.440 going to have to overcome for them to get smarter, 0:11:38.520 --> 0:11:42.240 which is called hallucinations. But that's a really critically important 0:11:42.280 --> 0:11:43.000 thing to remember. 0:11:43.679 --> 0:11:48.000 Yeah, Another critically important thing to remember is and you 0:11:48.040 --> 0:11:49.920 probably get this from what we said so far if 0:11:49.920 --> 0:11:51.640 you already know about a little bit about it, But 0:11:51.679 --> 0:11:55.960 there's no programmer that's teaching these things and typing in 0:11:56.080 --> 0:11:59.840 inputs and then saying here's how you learn things. Like 0:12:00.520 --> 0:12:03.640 it's doing this on its own, and it's learning things 0:12:03.640 --> 0:12:07.160 on its own, and what we're talking about eventually like that. 0:12:07.360 --> 0:12:09.719 You know, where it could get super scary is when 0:12:09.760 --> 0:12:12.480 it gets to what's called emergent abilities, where it's so 0:12:12.679 --> 0:12:18.160 powerful and there's so much data that the nuance that's 0:12:18.200 --> 0:12:21.319 missing now will be there right exactly. 0:12:21.640 --> 0:12:25.319 So, yeah, that's when things are going to get even 0:12:25.400 --> 0:12:29.040 harder to understand, you know, to remind yourself that you're 0:12:29.080 --> 0:12:30.320 talking to a machine, you know. 0:12:31.120 --> 0:12:32.880 Yeah, And the other thing too, though, even though I 0:12:32.920 --> 0:12:36.600 said humans aren't in putting this data. One of the 0:12:36.640 --> 0:12:39.720 big things that is allowing this stuff to get smarter 0:12:40.600 --> 0:12:45.200 is human feedback. It's called r l HF, which is 0:12:45.760 --> 0:12:48.960 reinforcement learning on human feedback. So at the end of 0:12:49.000 --> 0:12:52.960 your whatever you've told it to create, you can go 0:12:53.080 --> 0:12:55.120 back in and say, well, you got this wrong and 0:12:55.120 --> 0:12:57.600 this wrong, this is what that really is, and it says, 0:12:58.040 --> 0:12:59.960 thank you, I have now just gotten smart. 0:13:01.360 --> 0:13:04.079 Exactly. So one of the reasons why these things are 0:13:04.080 --> 0:13:07.200 suddenly just so smart and can say thank you, I've 0:13:07.240 --> 0:13:10.079 just gotten so much smarter is because of a paper 0:13:10.440 --> 0:13:16.360 that Google engineers published openly in twenty seventeen describing what's 0:13:16.440 --> 0:13:19.959 now like the essential ingredient for a large language model 0:13:20.000 --> 0:13:22.520 or probably any neural network from now on. It's called 0:13:22.520 --> 0:13:27.280 a transformer. And rather than analyzing each bit of text, 0:13:27.360 --> 0:13:30.200 let's say you say one of the very famous things. 0:13:30.800 --> 0:13:33.319 Marvin Minsky was one of the founders of the field 0:13:33.320 --> 0:13:38.200 of AI, and his son Henry prompted chet Gpt to 0:13:38.360 --> 0:13:41.600 describe what losing a sock in the dryer is like 0:13:41.640 --> 0:13:45.120 in the style of the Declaration of Independence. Right, So 0:13:45.240 --> 0:13:51.040 depending on how Henry Minsky type that in before transformers 0:13:51.360 --> 0:13:55.080 the neural network would analyze each word and do it 0:13:55.240 --> 0:13:57.920 one increment and a time, maybe not even words, sometimes 0:13:57.960 --> 0:14:02.400 strings of just letters together, phone names even if you 0:14:02.440 --> 0:14:07.319 can believe it, phone names even And what the transformer 0:14:07.360 --> 0:14:09.559 does is it changes that. It allows it to analyze 0:14:09.600 --> 0:14:12.960 everything all at once. So it's so much faster, not 0:14:13.160 --> 0:14:16.320 just in putting out a coherent answer to your question 0:14:16.520 --> 0:14:20.560 or request, but in also training itself on that text. 0:14:20.600 --> 0:14:23.840 So you just feed it the internet and it starts 0:14:23.880 --> 0:14:27.600 analyzing it and self correcting. It trains itself, It learns 0:14:27.640 --> 0:14:32.360 on its own, and that unfortunately also makes AI, including 0:14:32.440 --> 0:14:35.920 large language models what are known as black boxes. Yeah, 0:14:35.960 --> 0:14:38.840 we don't know how they're doing what they're doing. We 0:14:38.880 --> 0:14:40.560 have a good idea how to make them do the 0:14:40.600 --> 0:14:43.800 things we want, but the in between stuff, we cannot 0:14:43.800 --> 0:14:47.000 one hundred percent say what they're doing. How they come 0:14:47.120 --> 0:14:51.480 up with these conclusions, which also explains hallucinations in them 0:14:51.520 --> 0:14:53.000 not really making sense to us. 0:14:53.560 --> 0:14:57.360 Yeah, and you know, the T in GPT stands for transformer. 0:14:57.520 --> 0:15:02.880 It's generative pre trained transformer. And the reason they call 0:15:02.920 --> 0:15:05.680 it GPT for short is because if they call it 0:15:05.760 --> 0:15:10.000 generative pre trained transformer, everybody would be scared out of 0:15:10.040 --> 0:15:11.000 their minds. 0:15:10.880 --> 0:15:14.359 We just start running around to nowhere in particular. 0:15:14.840 --> 0:15:17.840 Yeah, should we take a break, I say, we do. 0:15:17.960 --> 0:15:20.560 I think that we kind of explained that fairly well. 0:15:20.600 --> 0:15:50.640 Yeah, fairly robust beginning, my friend. All right, So open 0:15:50.680 --> 0:15:54.760 Ai launched their chat GPT and very recently in November 0:15:54.760 --> 0:15:58.600 of twenty twenty two, and just in that brief window 0:15:59.160 --> 0:16:02.240 was like six or eight months ago. Things are kind 0:16:02.280 --> 0:16:06.400 of flying high, and all kinds of companies are launching 0:16:06.560 --> 0:16:10.320 their own stuff. Some of it is well. First of all, 0:16:10.320 --> 0:16:14.160 open ai is now at chat GPT four, yes, and 0:16:14.200 --> 0:16:17.640 I'm sure you know more will be coming in quick succession. 0:16:18.320 --> 0:16:20.920 But companies are launching, and we're going to talk about 0:16:20.920 --> 0:16:23.120 all of them kind of like broad stuff like chat, 0:16:23.160 --> 0:16:28.280 BT GPT and really specific stuff like well, hey, I'm 0:16:28.320 --> 0:16:31.600 in the banking business. Can we just design something for 0:16:31.680 --> 0:16:34.960 banking or just something for real estate? So they're also 0:16:35.000 --> 0:16:38.640 getting specific on a smaller level in addition to these 0:16:38.720 --> 0:16:41.640 large like Google and Microsoft and Being and all that stuff. 0:16:41.920 --> 0:16:44.160 Yeah, And to get specific, all you have to do 0:16:44.240 --> 0:16:49.640 is take an existing GPT, a large language model, and 0:16:50.240 --> 0:16:54.520 add some software that helps guide it a little more. Yeah, 0:16:54.600 --> 0:16:57.440 and there you go or just train it on specific 0:16:57.480 --> 0:17:01.800 stuff like medical notes. That's another one one of the 0:17:01.800 --> 0:17:05.159 other things that's that's changed very quickly between November of 0:17:05.200 --> 0:17:07.600 twenty twenty two in March of twenty twenty three, when 0:17:08.200 --> 0:17:13.440 I think GPT four became available. Just think about that. 0:17:13.440 --> 0:17:16.240 That's that's such a short amount of time. Yeah, all 0:17:16.240 --> 0:17:19.760 of a sudden. Now you can take a picture and 0:17:19.800 --> 0:17:22.480 feed it into a large language model and it will 0:17:22.840 --> 0:17:26.720 describe the picture. It will look at the picture essentially 0:17:26.960 --> 0:17:31.120 and describe what's going on. There's a there's a demonstration 0:17:31.280 --> 0:17:34.440 from one of the guys from open Ai who doodles 0:17:34.440 --> 0:17:37.600 like on a little like scrapbook piece of paper some 0:17:37.760 --> 0:17:41.120 ideas for a website. He takes a picture of that 0:17:41.240 --> 0:17:45.359 paper that he's written on, feeds it into chat GPT four, 0:17:46.359 --> 0:17:49.080 and it builds a website for him in a couple 0:17:49.119 --> 0:17:51.480 of minutes. That functions the way he was thinking of 0:17:51.800 --> 0:17:53.760 the on the doodle scratch pad. 0:17:54.240 --> 0:17:55.920 I wonder if the only way to slow this stuff 0:17:55.960 --> 0:17:59.280 down is to literally slow down the Internet again. Go 0:17:59.359 --> 0:18:02.160 back to like the old days when a picture would 0:18:02.200 --> 0:18:05.399 load like three lines at a time, right, and they 0:18:05.440 --> 0:18:09.760 describe a picture will be like someone's hair, someone's nose, 0:18:10.560 --> 0:18:14.320 someone's gin. Don't forget between Yeah, an hour later you 0:18:14.320 --> 0:18:15.160 have a complete picture. 0:18:15.320 --> 0:18:17.359 Right. I don't think there's any way to slow this 0:18:17.440 --> 0:18:21.800 down because we're in not to be alarmist, but we're 0:18:21.800 --> 0:18:26.760 in a second worst case scenario for introducing AI to 0:18:26.840 --> 0:18:30.400 the world, which is, rather than state actors doing this, 0:18:30.640 --> 0:18:34.680 which would be really bad, we have private companies doing it, 0:18:34.720 --> 0:18:38.879 which is just slightly less bad. But they're competing in 0:18:38.920 --> 0:18:42.480 an arms race to get the best, brightest, smartest AI 0:18:42.640 --> 0:18:45.600 out there as fast as they can, and they're not 0:18:45.680 --> 0:18:48.760 taking into account like all of the downsides to it. 0:18:48.760 --> 0:18:50.800 They're just throwing it out there as much as they can. 0:18:51.040 --> 0:18:53.639 Because one of the ways that these things get smarter 0:18:53.960 --> 0:18:56.760 is by interacting with the public. They get better and 0:18:56.800 --> 0:19:01.280 better at what they do from getting feedback from people 0:19:01.400 --> 0:19:02.280 using them. 0:19:02.520 --> 0:19:04.960 Yeah, even if it's for just some goofy fun thing 0:19:05.000 --> 0:19:08.440 you're doing, it's learning from that. And you talked about 0:19:08.440 --> 0:19:12.560 the advancements made between the launch of three point five 0:19:12.680 --> 0:19:17.560 and GPT four, and three point five scored in the 0:19:17.600 --> 0:19:21.960 tenth percentile when it took the uniform bar exam, and 0:19:22.320 --> 0:19:26.159 four has already scored in the ninetieth percentile, and they 0:19:26.200 --> 0:19:29.399 found that chet GPT four is really it's great at 0:19:29.800 --> 0:19:34.040 taking tests, and it's scoring really well on tests, particularly 0:19:34.440 --> 0:19:37.760 you know, standardized tests. All. I think it basically aced 0:19:38.000 --> 0:19:41.520 all of the AP tests that you would take to 0:19:41.520 --> 0:19:45.960 get into AP classes except well, it took a couple 0:19:45.960 --> 0:19:49.359 of AP class but the max score is five, and 0:19:49.440 --> 0:19:52.480 I think it got five's kind of on everything except 0:19:52.560 --> 0:19:59.160 for math. It got a four and math it's kind 0:19:59.160 --> 0:20:02.560 of it's we it's kind of weirdly counterintuitive because it's 0:20:02.600 --> 0:20:05.919 a number space thing, but it has more trouble with math, 0:20:06.640 --> 0:20:10.040 like rudimentary math, than it does with like constructing a 0:20:10.080 --> 0:20:13.720 paragraph on you know, Shakespeare or something, or as Shakespeare 0:20:14.800 --> 0:20:18.320 does better with like math word problems and more advanced 0:20:18.359 --> 0:20:20.760 math than it does just at basic math apparently, or. 0:20:20.760 --> 0:20:25.440 Like describing how a formula functions using you know, writing. 0:20:26.200 --> 0:20:28.320 The thing is though, and this is another great example 0:20:28.320 --> 0:20:31.440 of how fast this is moving. They've already figured out 0:20:31.440 --> 0:20:33.040 that all you have to do is do what's called 0:20:33.040 --> 0:20:38.760 prompting where you where you basically take the answer that 0:20:38.840 --> 0:20:42.960 the the incorrect answer that the large language model gives you, 0:20:43.320 --> 0:20:45.840 and then basically re explain it by breaking it down 0:20:45.880 --> 0:20:48.760 into different parts, and it learns as you're doing that, 0:20:48.880 --> 0:20:50.560 and then all of a sudden it comes up with 0:20:50.760 --> 0:20:53.479 it gets better at math. So they've figured out tools, 0:20:53.520 --> 0:20:57.800 extra software you can lay over a GPT that basically 0:20:58.200 --> 0:21:00.959 teach it to do math or prompt it in the 0:21:01.000 --> 0:21:03.160 correct way so that you get the answer you're looking 0:21:03.200 --> 0:21:04.640 for that's based on math. 0:21:05.520 --> 0:21:08.720 Yeah. I mean, every time I read something that said, well, 0:21:08.800 --> 0:21:11.240 right now, it's not so great at this, I just 0:21:11.359 --> 0:21:14.119 assume that meant and we'll have that worked out in 0:21:14.160 --> 0:21:15.119 the next few weeks. 0:21:15.320 --> 0:21:17.880 Yeah, pretty much. I mean, because as these things get 0:21:18.760 --> 0:21:21.040 like bigger and smarter, in the data sets that they're 0:21:21.040 --> 0:21:24.080 trained on get wider, they're just going to get better 0:21:24.119 --> 0:21:26.480 and better at this because they again they learned from 0:21:26.480 --> 0:21:27.280 their mistakes. 0:21:29.800 --> 0:21:34.439 Yeah, just like humans, right, exactly like humans. So you 0:21:34.480 --> 0:21:38.000 mentioned these hallucinations kind of briefly, and this this is 0:21:38.040 --> 0:21:40.000 one of the big problems with them so far that 0:21:40.560 --> 0:21:42.720 again I'm sure they will figure this out in due time. 0:21:43.119 --> 0:21:47.480 But one example that Livia found was to prompt it 0:21:47.520 --> 0:21:51.359 with what mammal lays the largest eggs. And one of 0:21:51.400 --> 0:21:54.720 the problems is when it gives hallucinations or wrong answers it. 0:21:55.160 --> 0:21:57.520 You know, it's not saying like, well, I'm not so 0:21:57.600 --> 0:22:00.800 sure about this. It's saying this is true, just like 0:22:00.840 --> 0:22:03.879 anything else. I'm spitting out right with a lot of confidence. 0:22:03.880 --> 0:22:06.280 So the answer there was the mammal that lays the 0:22:06.359 --> 0:22:08.720 largest eggs is the elephant, and elephant's eggs are so 0:22:08.760 --> 0:22:11.320 small that they are often visible to the naked eye, 0:22:11.520 --> 0:22:14.359 so they're not commonly known to lay eggs at all. However, 0:22:14.400 --> 0:22:17.040 in terms of sheer size, and elephant's eggs are the 0:22:17.160 --> 0:22:19.040 largest of any mammal. 0:22:18.800 --> 0:22:20.760 Which makes sense in a really weird way if you 0:22:20.760 --> 0:22:21.800 think about it. 0:22:21.920 --> 0:22:23.399 Sure, those little invisible eggs. 0:22:23.480 --> 0:22:26.400 Yeah, because mammals don't lay eggs obviously, But the way 0:22:26.440 --> 0:22:28.560 that it put it was if you didn't know that 0:22:28.600 --> 0:22:31.200 mammals don't lay eggs, or you didn't know anything about elephants, 0:22:31.520 --> 0:22:33.679 you'd be like, oh, that's interesting, and take that as 0:22:33.680 --> 0:22:36.480 a fact, because it's saying this confidently. And I saw 0:22:36.760 --> 0:22:40.200 written somewhere that one GPT actually argued with the user 0:22:40.240 --> 0:22:41.919 and told them they were wrong when they told the 0:22:41.960 --> 0:22:46.119 GPRIT that it was wrong, Yeah, which is not a 0:22:46.160 --> 0:22:50.239 behavior you want at all. But that's what's termed as 0:22:50.240 --> 0:22:53.560 a hallucination, and a hallucination is a good way to 0:22:53.800 --> 0:23:00.600 understand it. That I saw is that again, this GPT, 0:23:00.720 --> 0:23:04.200 this large language model, doesn't have any idea what it's 0:23:04.200 --> 0:23:09.640 saying means. It's just picked up it's noticed patterns that 0:23:09.720 --> 0:23:13.000 we've not noticed before, and it's putting them together in 0:23:13.119 --> 0:23:17.280 nonsensical ways. But they're still sensible if you read them. 0:23:17.320 --> 0:23:20.600 It's just factually they're not sensible because it doesn't have 0:23:20.640 --> 0:23:24.600 any fact checking necessarily, it just knows what it's finding 0:23:25.440 --> 0:23:28.080 kind of correlates with other things. So there's some sensible 0:23:28.119 --> 0:23:31.560 stuff in there, like the phrase invisible to the naked eye, 0:23:32.000 --> 0:23:35.959 or laying eggs or elephants in mammals. Like this stuff 0:23:36.280 --> 0:23:38.840 all makes sense. It's not like these are just strings 0:23:38.840 --> 0:23:42.520 of letters. Yeah, it's just putting them together in ways 0:23:42.560 --> 0:23:46.440 that are not true. They're factually incorrect, and that's a hallucination. 0:23:46.560 --> 0:23:50.840 It's not like the computer is thinking that this is true. 0:23:50.880 --> 0:23:57.560 It doesn't understand things like truth in falsehood. It just creates, 0:23:58.000 --> 0:24:00.440 and some of the time it gets it really wrong. 0:24:00.920 --> 0:24:02.439 Yeah, I didn't know what an elephant is. 0:24:03.160 --> 0:24:06.400 No, it just knows that it correlates to in some 0:24:06.640 --> 0:24:11.320 really small way. That we've never noticed before the word eggs. 0:24:11.560 --> 0:24:15.280 Yeah, and this is uh that that's a problem if 0:24:15.440 --> 0:24:17.440 if it's just like, oh, well, this thing isn't quite 0:24:17.440 --> 0:24:19.400 where it needs to be at because it thinks elephants 0:24:19.440 --> 0:24:22.280 lay eggs. But there have already been plenty of real 0:24:22.440 --> 0:24:25.800 world examples where people are using this and it's screwing 0:24:25.840 --> 0:24:28.160 things up for their business or for commerce or something 0:24:28.240 --> 0:24:31.200 where the yeah, or their client, well, that's that's one. 0:24:31.240 --> 0:24:35.440 There was an attorney who was representing a passenger who 0:24:35.520 --> 0:24:39.720 was suing an airline and used chat bt to do research, 0:24:40.320 --> 0:24:42.640 and it came up with a bunch of fake cases 0:24:42.960 --> 0:24:46.560 that this attorney didn't bother to fact check, I guess. 0:24:47.160 --> 0:24:50.199 And there were like a dozen fake cases that this 0:24:50.280 --> 0:24:51.840 attorney submitted in his brief. 0:24:52.080 --> 0:24:54.360 And it wasn't like so like from what I understand, 0:24:54.359 --> 0:24:57.800 like the the brief was largely compiled from what the 0:24:57.960 --> 0:25:01.480 GPT spit out. The It wasn't like the GPT just 0:25:01.520 --> 0:25:03.240 made up the names of cases. It made up the 0:25:03.280 --> 0:25:06.119 names of cases and then described the background of the 0:25:06.160 --> 0:25:09.560 case and how they related to the case at hand, right, 0:25:09.640 --> 0:25:12.440 So it just completely made these up out of out 0:25:12.440 --> 0:25:15.359 of the blue. And yeah, that lawyer had no idea. 0:25:15.400 --> 0:25:17.800 He said in a brief later that he had no 0:25:17.880 --> 0:25:20.439 idea that this thing was capable of being incorrect. So 0:25:20.760 --> 0:25:22.399 it was like one of the first times he used it, 0:25:22.480 --> 0:25:24.280 and he threw himself on the mercy of the court. 0:25:24.600 --> 0:25:26.520 And I'm not quite sure exactly what happened. I think 0:25:26.520 --> 0:25:28.440 they're still figuring out what to do about it. 0:25:29.480 --> 0:25:31.679 Maybe just go spend some quality time with your little 0:25:31.800 --> 0:25:33.919 chat butt exactly. 0:25:34.960 --> 0:25:39.600 Similarly, Meta had a large language model that basically got 0:25:39.680 --> 0:25:42.040 laughed off of the Internet because it was very science 0:25:42.040 --> 0:25:45.800 focused and it would make up things that just didn't exist, 0:25:45.840 --> 0:25:49.879 like mathematical formula, like there was one that called the 0:25:50.000 --> 0:25:54.399 yoko or no, the lenin Ono correlation or something like 0:25:54.440 --> 0:25:57.480 that completely made up this thing that I read, and 0:25:57.520 --> 0:25:59.960 I was like, oh, that's interesting. I had no idea. 0:26:00.320 --> 0:26:02.840 I have never heard this stuff before, and I would 0:26:02.880 --> 0:26:04.479 have just thought that it was real had I not 0:26:04.520 --> 0:26:07.080 realized and known ahead of time that it was a 0:26:07.080 --> 0:26:10.800 hallucination that this math thing does not exist anywhere. And 0:26:10.800 --> 0:26:15.080 it even attributed it to a live mathematician said that 0:26:15.160 --> 0:26:18.479 this was the guy who discovered it. So like, it 0:26:18.560 --> 0:26:22.320 really can get hard to discern what's true and what's not, 0:26:22.640 --> 0:26:25.440 which again is a really big problem if we haven't 0:26:25.440 --> 0:26:26.439 gotten that cross yet. 0:26:26.680 --> 0:26:33.560 Did they say that mathematician's name was math be calculus. Yeah. 0:26:33.640 --> 0:26:35.639 Another example, and this is, you know, we're going to 0:26:35.680 --> 0:26:38.600 talk a little bit about you know, replacing jobs in 0:26:38.640 --> 0:26:42.679 the various ways that can and already is happening. But 0:26:42.760 --> 0:26:46.120 seeing that, for instance, said oh, you know what, let 0:26:46.119 --> 0:26:47.479 me try this thing out and see if we can 0:26:47.480 --> 0:26:51.200 get it to write an actual story. And so they 0:26:51.240 --> 0:26:54.199 got an AI tool to write one on what is 0:26:54.240 --> 0:26:57.239 compound interest, and it was just there was a lot 0:26:57.280 --> 0:27:01.040 of stuff wrong in it. There was some plagiarism, you know, 0:27:01.280 --> 0:27:04.359 directly lifted. So there's you know, these things aren't fool 0:27:04.359 --> 0:27:06.800 proof yet, and it's definitely not something that should be 0:27:08.080 --> 0:27:12.040 utilized for like a public facing website that's supposed to 0:27:12.680 --> 0:27:17.719 have like really solid vetted articles about uh well, especially 0:27:17.760 --> 0:27:21.080 seen it about a tech right of all things. 0:27:21.320 --> 0:27:24.840 That's something that the National Eating Disorder Association found out 0:27:24.840 --> 0:27:30.200 the hard way. They apparently replaced entirely it's human staffed 0:27:30.240 --> 0:27:34.520 hotline with the chatbot, and supposedly they were accused of 0:27:34.520 --> 0:27:37.320 doing this to bust the union that had formed there 0:27:38.520 --> 0:27:41.200 and so when they released the chatpot into the world 0:27:41.440 --> 0:27:44.639 and it started offering advice to people suffering from eating disorders. 0:27:44.880 --> 0:27:49.000 It gave standard, you know, weight loss advice, which you 0:27:49.119 --> 0:27:51.840 probably get from your doctor who didn't realize you had 0:27:51.840 --> 0:27:54.480 an eating disorder, but in the context of an eating disorder, 0:27:54.720 --> 0:27:57.720 it was all like trigger, trigger, trigger, one right after 0:27:57.800 --> 0:28:00.240 the other. Right, Like, it was telling these people with 0:28:00.280 --> 0:28:03.000 eating disorders to like, weigh yourself every week and try 0:28:03.040 --> 0:28:05.399 to cut out five hundred to one thousand calories a 0:28:05.480 --> 0:28:08.879 day and you'll lose some weight, and just stuff that 0:28:08.880 --> 0:28:11.320 that would set everybody off. And very quickly they took 0:28:11.359 --> 0:28:14.959 it offline and I guess brought their humans back, hopefully 0:28:15.000 --> 0:28:15.880 at double the pay. 0:28:16.920 --> 0:28:19.920 Yeah. But I mean this stuff is that's already being 0:28:19.960 --> 0:28:22.560 solved as well, because they point out that GPT four 0:28:23.359 --> 0:28:27.240 has already scored forty percent higher than three point five, 0:28:27.880 --> 0:28:32.159 again just a handful of months ago on these accuracy tests, 0:28:32.720 --> 0:28:36.800 so that that is even getting better. And you know, 0:28:36.840 --> 0:28:38.800 where where I guess people want it to get to 0:28:39.760 --> 0:28:42.680 is to the point where it doesn't need human supervision 0:28:42.760 --> 0:28:45.800 to spit out really really accurate stuff exactly. 0:28:45.920 --> 0:28:48.479 That's pretty much where they're hoping to get it. And 0:28:48.520 --> 0:28:51.600 I mean it's just they have the model, they have 0:28:51.840 --> 0:28:53.800 everything they need. They just it just has to be 0:28:53.840 --> 0:28:54.360 tinkered with. 0:28:54.440 --> 0:28:58.440 Now, should we take another break? I think so, all right, 0:28:58.480 --> 0:29:00.880 we'll take another break and then get into sort of 0:29:00.920 --> 0:29:04.040 the the economics of it and whether or not your 0:29:04.200 --> 0:29:05.680 job may be at risk right after this. 0:29:29.640 --> 0:29:33.240 So one of the astounding things about this that it 0:29:33.360 --> 0:29:37.320 really caught everybody off guard is that these large language models, 0:29:37.800 --> 0:29:41.520 the jobs they're coming after are white collar knowledge jobs. 0:29:42.320 --> 0:29:45.920 They're so good at things like writing, they're good at researching, 0:29:45.960 --> 0:29:50.200 they're good at analyzing photos. Now and that's a huge 0:29:50.680 --> 0:29:53.760 sea change from what it's been like traditionally. Right wherever 0:29:54.000 --> 0:29:57.920 whenever we've automated things, it's usually replaced manual labor. Now 0:29:57.920 --> 0:30:00.400 it's the manual labor that's safe in this Yeah, this 0:30:00.640 --> 0:30:05.200 generation of automation, it's the white collar knowledge shops that 0:30:05.280 --> 0:30:08.200 are at risk. And not just white collar job but 0:30:08.480 --> 0:30:12.320 artists in yeah, like just who have nothing to do 0:30:12.360 --> 0:30:16.480 with white collar or jobs, they're at risk as well. 0:30:17.200 --> 0:30:20.880 Yeah, I'm sure the farmers are all sitting around going, 0:30:21.480 --> 0:30:22.280 how's that going for you? 0:30:22.360 --> 0:30:24.680 Yeah, how's that taste? Uh? 0:30:24.880 --> 0:30:28.160 So? Yeah, art art is when when d A L L. E. 0:30:28.320 --> 0:30:31.040 Doll E came out, that was an art tool where 0:30:31.880 --> 0:30:33.280 a lot of people, a lot of people I know, 0:30:33.360 --> 0:30:36.120 would input. I guess I never did it. I never 0:30:36.280 --> 0:30:38.920 do anything like that, not because I'm afraid or anything, 0:30:39.000 --> 0:30:44.160 but I just just not interested basically. But I guess 0:30:44.160 --> 0:30:46.160 you would submit, like a photograph of yourself and then 0:30:46.160 --> 0:30:48.120 it would say, well, here's you as a superhero or 0:30:48.160 --> 0:30:52.800 here's you as a Renaissance painting or whatever. And you know, 0:30:52.920 --> 0:30:57.280 it's sourcing images from real artists throughout history, from Getty 0:30:57.320 --> 0:31:00.280 Images and places like that, and there are are ready 0:31:00.360 --> 0:31:04.880 artists that are suing for infringement. Getty Images is suing 0:31:04.920 --> 0:31:09.520 for infringement and saying you can't even if you're mixing 0:31:09.640 --> 0:31:13.120 up things and it's not like a Rembrandt. Let's say 0:31:13.560 --> 0:31:16.239 you're using all of the artists from that era and 0:31:16.360 --> 0:31:19.600 mashing it up together in a way that, like we 0:31:19.720 --> 0:31:21.160 think basically is illegal. 0:31:21.400 --> 0:31:25.080 Yeah, they say this doesn't count as transformative use, which 0:31:25.120 --> 0:31:30.400 is typically protected under the law. Right, this is instead 0:31:31.080 --> 0:31:34.640 just some sort of mash up that that a machine 0:31:34.680 --> 0:31:39.040 is doing. It's to me, it's almost splitting hairs. But 0:31:39.080 --> 0:31:41.560 I also very much get where they're coming from not 0:31:41.600 --> 0:31:44.720 just a place of panic, but like they're a real 0:31:46.280 --> 0:31:48.240 like they have a basis in fact that these things 0:31:48.280 --> 0:31:51.640 are not transforming because they don't understand what they're doing. 0:31:52.520 --> 0:31:56.840 Yeah, and companies are taking notice very quickly. There are 0:31:56.840 --> 0:31:59.280 some companies I'm sure everyone's going to kind of fall 0:31:59.320 --> 0:32:01.840 in line, that are already saying, well, no, you got 0:32:01.880 --> 0:32:04.880 to start paying us for access to this stuff. We 0:32:05.120 --> 0:32:09.400 paid human beings to create this content for lack of 0:32:09.400 --> 0:32:12.760 a better word, and put it online for people to access. 0:32:13.360 --> 0:32:15.440 But you can't come in here now and access it 0:32:15.480 --> 0:32:18.960 with a bot and use it and charge for it 0:32:19.400 --> 0:32:21.840 without giving us a little juice. And there are a 0:32:21.880 --> 0:32:24.960 lot of companies that are already saying like, you can't 0:32:25.120 --> 0:32:27.760 use this. If you're an employee of our company, you 0:32:27.800 --> 0:32:32.080 can't use chatbots at all because some of our company's 0:32:32.080 --> 0:32:37.480 secrets might end up being spilled somehow, or you know, 0:32:37.520 --> 0:32:40.960 our databases are all of a sudden exposed. So companies 0:32:40.960 --> 0:32:43.920 are really moving fast to trying to protect their ip 0:32:44.280 --> 0:32:45.120 I guess. 0:32:44.920 --> 0:32:48.440 Well, yeah, and one of the I mean some of 0:32:48.480 --> 0:32:51.840 the companies that are behind the GPTs that are out 0:32:51.920 --> 0:32:54.320 right now, the large language models that are out right 0:32:54.360 --> 0:32:58.720 now are well known for not only not protecting their 0:32:58.840 --> 0:33:03.280 users information, but for rating it for its own use. Like, 0:33:03.360 --> 0:33:06.080 for example, Meta is one of the ones with They 0:33:06.160 --> 0:33:11.280 have their large language models called Lama, and there's a 0:33:11.360 --> 0:33:15.240 chatbot called Alpaca. And it makes total sense that you 0:33:15.480 --> 0:33:18.920 are probably signing away your right to protect your information 0:33:19.000 --> 0:33:21.520 when you use those things on whatever computer you're using 0:33:21.520 --> 0:33:24.080 it on or whatever network you're using it on. I 0:33:24.120 --> 0:33:27.000 don't understand exactly. I haven't seen anything that says this 0:33:27.160 --> 0:33:29.160 is how they're doing it, or even that they are 0:33:29.160 --> 0:33:32.640 definitely doing this. I think it's just that the powers 0:33:32.680 --> 0:33:35.400 that be no, like they would totally do this if 0:33:35.440 --> 0:33:37.400 they can, and they probably are, so we should just 0:33:37.400 --> 0:33:40.440 stay keep our employees away from it, you know, as 0:33:40.520 --> 0:33:41.200 much as we can. 0:33:42.040 --> 0:33:46.080 Yeah, it's like we said, it's being used on smaller 0:33:46.160 --> 0:33:50.600 levels by One of the uses that Livia dug up 0:33:50.680 --> 0:33:53.200 was like, let's say a real estate agent, instead of 0:33:53.200 --> 0:33:56.959 taking time to write up listings, has a chatbot to it, 0:33:57.000 --> 0:33:59.400 and then they can go through afterward and make adjustments 0:33:59.440 --> 0:34:01.080 to it as needed. 0:34:01.360 --> 0:34:05.280 Well, in exchange, that database now knows exactly what you 0:34:05.320 --> 0:34:07.480 think of that one ugly bathroom. 0:34:07.440 --> 0:34:12.360 That's right, or doctors may be using it to compile 0:34:12.440 --> 0:34:17.400 lists of possible diseases or conditions that someone might have 0:34:17.440 --> 0:34:23.359 based on symptoms. These all sound like uses that are like, hey, 0:34:23.640 --> 0:34:25.080 this sounds like it could be a good thing in 0:34:25.080 --> 0:34:28.200 some ways, and it can be in some ways. But 0:34:28.880 --> 0:34:31.400 it's the wild West right now, so it's not like 0:34:31.520 --> 0:34:35.319 there's any there's anyone saying, well, you can't use it 0:34:35.360 --> 0:34:37.160 for that, you can only use it for this, you 0:34:37.200 --> 0:34:37.759 know what I'm saying. 0:34:37.840 --> 0:34:40.560 Plus, also, everything that we've come up with as just 0:34:41.080 --> 0:34:44.840 Internet users in the general public has been what we 0:34:44.880 --> 0:34:48.400 could come up with in given three months, with no 0:34:48.560 --> 0:34:51.399 warning that we should start thinking about this. It's just like, hey, 0:34:51.440 --> 0:34:53.080 this is here, what are you going to do with it? 0:34:53.120 --> 0:34:55.640 And people are just finding new things to do with 0:34:55.680 --> 0:34:58.840 it every day, And yeah, some of them are benign, 0:34:58.960 --> 0:35:02.880 like having a draft a blog post for your business. 0:35:03.280 --> 0:35:05.279 I thought they were already doing that based on some 0:35:05.360 --> 0:35:08.920 of the emails that I get from like businesses, right, Yeah, 0:35:08.920 --> 0:35:11.719 but they definitely are now if they weren't before. And 0:35:11.760 --> 0:35:16.000 that's totally cool because there's there's a it's just taking 0:35:16.040 --> 0:35:18.960 some of the weight off of the humans that are 0:35:19.000 --> 0:35:23.759 already doing this. Work. Right, what's going to be problematic 0:35:24.080 --> 0:35:28.200 is when it comes for the full job, or enough 0:35:28.239 --> 0:35:32.839 of the job that the company can transfer whatever's left 0:35:32.840 --> 0:35:35.839 of that person's job to other people and make them 0:35:35.920 --> 0:35:39.120 just work a little harder while they're supported by the AI. 0:35:39.840 --> 0:35:43.400 Yeah, here's some stats that they were pretty shocking to me. 0:35:43.480 --> 0:35:47.239 I didn't know it was moving this fast. But there's 0:35:47.239 --> 0:35:50.400 a networking app called Fishbowl and in twenty twenty three, 0:35:50.560 --> 0:35:54.520 just earlier this year, they found that forty percent of 0:35:55.040 --> 0:35:58.759 what they call working professionals are already using some kind 0:35:58.800 --> 0:36:01.799 of either chat, GPT or some kind of AI tool 0:36:01.960 --> 0:36:07.200 while they work, whether it's generating idealists or brainstorming lists, 0:36:07.320 --> 0:36:11.799 or actually writing stuff or maybe looking at code. And 0:36:12.120 --> 0:36:15.080 this is the troubling part. Forty of those forty percent, 0:36:16.000 --> 0:36:19.240 almost seventy percent are doing that in secret and hadn't 0:36:19.239 --> 0:36:20.920 told their bosses that they were doing that. 0:36:21.040 --> 0:36:23.360 Right, those are just working professionals. We haven't even started 0:36:23.360 --> 0:36:24.600 talking about students yet. 0:36:25.200 --> 0:36:27.680 Yeah. I mean you combine that with work from home, 0:36:27.760 --> 0:36:29.200 you got a real racket going on. 0:36:29.719 --> 0:36:33.120 For sure, you know. Yeah, no, totally again though, I mean, 0:36:33.160 --> 0:36:35.200 like if you can use it to do good work, 0:36:35.239 --> 0:36:38.040 and you can now do more work. I think you 0:36:38.040 --> 0:36:40.880 should be paid for more work. Like if your productivity's 0:36:40.920 --> 0:36:44.239 gone through the roof, great, you figured it out. I've 0:36:44.280 --> 0:36:46.799 got no problem with that. It's the opposite that I 0:36:46.800 --> 0:36:47.600 have the problem with. 0:36:48.280 --> 0:36:50.920 Well, let's skip students for a second and then and 0:36:51.000 --> 0:36:54.160 talk about that since you brought it up, Because here's 0:36:54.239 --> 0:36:59.840 the thing this is the United States doesn't have a 0:36:59.840 --> 0:37:04.040 G eight track record of ignoring the bottom line in 0:37:04.080 --> 0:37:09.839 favor of just keeping hardworking humans at their jobs. So 0:37:10.920 --> 0:37:13.239 I think it was a Goldman. Sachs said that they 0:37:13.320 --> 0:37:17.840 found that there could actually be an increase in the 0:37:17.880 --> 0:37:21.160 annual GDP by about seven percent over ten years because 0:37:21.680 --> 0:37:26.200 productivity increases. And I guess the idea is that productivity 0:37:26.280 --> 0:37:29.720 is increasing because let's say you've got twenty to thirty 0:37:29.719 --> 0:37:32.680 percent of stuff being done by AI. That opens up 0:37:32.680 --> 0:37:36.000 twenty to thirty percent of your time for your employees 0:37:36.040 --> 0:37:41.720 to maybe innovate or you know, get do other capitalistic things. 0:37:42.960 --> 0:37:45.240 But what it to me, and this is just my opinion, 0:37:45.280 --> 0:37:48.000 and again we're really early in all this, but it's 0:37:48.000 --> 0:37:51.440 a bottom line world, and especially a bottom line country 0:37:51.440 --> 0:37:55.279 that we live in, and I imagine what it would 0:37:55.360 --> 0:37:59.319 likely mean is by by jobs more than it means, well, hey, 0:37:59.360 --> 0:38:01.840 you've got more time, and why don't you innovate it 0:38:01.920 --> 0:38:06.200 your job, because for most jobs, it'll probably be like, oh, 0:38:06.200 --> 0:38:07.840 wait a minute, if we can teach it to do 0:38:07.880 --> 0:38:10.640 forty center your job, I bet we could train it 0:38:10.640 --> 0:38:12.400 to do a one hundred percent. 0:38:12.239 --> 0:38:15.080 Yeah, or we can get rid of, you know, a 0:38:15.120 --> 0:38:16.759 bunch of you and just keep some of you to 0:38:17.080 --> 0:38:18.440 do the other sixty percent. 0:38:19.000 --> 0:38:20.880 You know. But now see these people are out of 0:38:20.960 --> 0:38:22.839 jobs that it's going to bite them in the rear though, 0:38:22.880 --> 0:38:26.640 because it's not ultimately going to be well, who knows. 0:38:26.719 --> 0:38:28.120 It doesn't seem like it could be good for the 0:38:28.160 --> 0:38:30.000 overall economy if all of a sudden, all these people 0:38:30.000 --> 0:38:32.800 are out of jobs. Because people being out of jobs 0:38:32.880 --> 0:38:35.640 mean they're not that means the economy is going to tank. 0:38:35.680 --> 0:38:38.680 They're not spending. And it's not like a situation where 0:38:40.440 --> 0:38:44.960 you know, the tractor replaced the plow and then the 0:38:45.080 --> 0:38:48.480 robot tractor replaced the tractor. But hey, now we've got 0:38:48.480 --> 0:38:51.400 these better jobs where you're designing and building these robot 0:38:51.440 --> 0:38:55.160 tractors and they're higher paying and they're great. It's not 0:38:55.400 --> 0:38:59.520 like that because you know, the farmer was replaced who 0:38:59.600 --> 0:39:02.840 drove that tractor and isn't skilled in the practice of 0:39:02.920 --> 0:39:07.640 designing robot tractors. And in this case, in most cases, 0:39:08.480 --> 0:39:12.400 they're not being there's not some other job waiting for 0:39:12.480 --> 0:39:15.560 someone who got fired in the world of designing AI. 0:39:16.680 --> 0:39:17.399 Does that make sense? 0:39:17.440 --> 0:39:19.799 No, it makes total sense. But yeah, and in this case, 0:39:19.840 --> 0:39:22.640 one of the big differences is instead of the farmer 0:39:22.719 --> 0:39:24.840 having to go figure out how to work a computer, 0:39:25.120 --> 0:39:27.319 that people working computers now have to go figure out 0:39:27.320 --> 0:39:30.719 how to be farmers in order to sustain themselves. Right, 0:39:31.160 --> 0:39:33.960 But you're right, we don't have a track record of 0:39:34.040 --> 0:39:38.120 taking care of people very well, at least who are 0:39:38.200 --> 0:39:40.600 out of a job. And I mean, without getting on 0:39:40.640 --> 0:39:44.319 a soapbox here, what's either going to come out of this, 0:39:44.360 --> 0:39:45.840 because there's going to be one or the other. The 0:39:45.840 --> 0:39:48.400 status quo as it is now or as it wasn't 0:39:48.600 --> 0:39:51.080 as of up to twenty twenty two. We don't know 0:39:51.160 --> 0:39:55.120 that that's going to be around anymore. Instead, we'll either 0:39:55.200 --> 0:39:59.120 do something like create universal basic income for people to 0:39:59.239 --> 0:40:03.720 be like, hey, your industry literally does not exist anymore 0:40:03.960 --> 0:40:07.400 and it just happened overnight. Basically, we're just gonna make 0:40:07.400 --> 0:40:09.759 sure that everybody's at least minimally taken care of while 0:40:09.800 --> 0:40:13.120 we're figuring out what comes next, or it's gonna be 0:40:13.160 --> 0:40:17.200 like good luck, chump, you're fired, You're out on your own. Instead, 0:40:17.239 --> 0:40:19.960 we're gonna take all this extra wealth, this extra two 0:40:20.080 --> 0:40:23.400 trillion dollars that's gonna be generated, and push it upward 0:40:23.840 --> 0:40:27.279 toward the wealthy instead and everybody else. Is just the 0:40:28.280 --> 0:40:32.000 divide between wealthy and not wealthy is just going to 0:40:32.480 --> 0:40:35.120 exponentially grow. One of those two things is gonna happen, 0:40:35.200 --> 0:40:37.560 because I don't see how there's just gonna be a 0:40:37.600 --> 0:40:40.360 regular middle ground like there is now where it's kind 0:40:40.360 --> 0:40:44.600 of shaky, and how we're taking care of people, because 0:40:44.600 --> 0:40:48.600 there's just gonna be so many layoffs and fairly skilled 0:40:48.640 --> 0:40:52.280 workers being laid off too. We've just never encountered that before. 0:40:52.800 --> 0:40:56.320 Yeah. I mean, that's the thing that these the largest 0:40:56.320 --> 0:40:59.839 corporations might want to think about. Is all that's gonna 0:40:59.880 --> 0:41:05.880 take is one CEO of a huge corporation to say, 0:41:06.000 --> 0:41:09.160 wait a minute, it I think I can get rid 0:41:09.200 --> 0:41:12.319 of seventy five percent of the vps in my in 0:41:12.360 --> 0:41:18.120 my company, right, and like who unless who accept the 0:41:18.160 --> 0:41:20.399 person at the very very top of that food chain 0:41:21.239 --> 0:41:23.880 is protected. And the answer is nobody. 0:41:24.680 --> 0:41:26.919 No, No, the offense essentially. 0:41:26.719 --> 0:41:27.920 At the end of the day, because they make a 0:41:28.000 --> 0:41:29.879 lot of money. If you it's one thing to lay 0:41:29.880 --> 0:41:32.279 off a bunch of you know, technical writers that are 0:41:32.640 --> 0:41:35.000 all sitting in their cubicles, But if you start laying 0:41:35.000 --> 0:41:38.680 off those those vps who get those big bonuses, that's 0:41:38.880 --> 0:41:41.239 more bonus money. And you know, are we looking at 0:41:41.280 --> 0:41:44.240 a situation where a corporation is run by one human? 0:41:44.760 --> 0:41:47.480 I mean, it's entirely possible, Like you can make a 0:41:47.520 --> 0:41:49.239 really good case that what it is going to wipe 0:41:49.280 --> 0:41:53.120 out is the middle management, right, vps, This is exactly 0:41:53.200 --> 0:41:55.400 like you said, and that we still will need some 0:41:55.520 --> 0:41:57.160 humans to do some stuff. 0:41:57.160 --> 0:41:59.320 Like the board take care of the board, right, sure. 0:41:59.280 --> 0:42:02.279 Of course, yeah, but yes, I mean who knows, we 0:42:02.880 --> 0:42:07.439 have no idea at this point. Ultimately, it could very 0:42:07.520 --> 0:42:13.799 easily provide for a much better healthier society, at least 0:42:13.840 --> 0:42:17.759 financially speaking, it could do that, especially given a long 0:42:17.840 --> 0:42:18.760 enough period of time. 0:42:19.160 --> 0:42:21.040 I'm a cynic when it comes to that kind of 0:42:21.040 --> 0:42:21.600 trust though. 0:42:21.680 --> 0:42:23.560 I am as well for sure. But if you look 0:42:23.600 --> 0:42:28.080 back in history at the history of technology overall, especially 0:42:28.160 --> 0:42:30.280 if you just turn a blind eye to human suffering 0:42:30.280 --> 0:42:32.719 for a second and you just look at the progress 0:42:32.719 --> 0:42:36.359 of society. Right in a lot of ways it has 0:42:36.400 --> 0:42:39.840 has gotten better and better things to technology. There's also 0:42:39.880 --> 0:42:42.640 a lot of downsides to it. Nothing's black and white, 0:42:43.120 --> 0:42:46.200 it's just not that's just not how things are. So 0:42:46.239 --> 0:42:48.680 there's of course going to be problems. There's going to 0:42:48.719 --> 0:42:50.759 be suffering, there's going to be people left behind, there's 0:42:50.800 --> 0:42:52.800 going to be people that fall through the cracks. It's 0:42:52.880 --> 0:42:56.360 just inevitable. We just don't know how many people for 0:42:56.480 --> 0:42:59.400 how long and what will happen to those people on 0:42:59.440 --> 0:43:01.280 the other of this transition. 0:43:02.320 --> 0:43:06.120 Yeah, I was talking with somebody the other day about 0:43:06.880 --> 0:43:11.800 the writer's strike in Hollywood. The WGA is striking right now. 0:43:12.520 --> 0:43:14.160 For those of you who don't know, it's kind of 0:43:14.160 --> 0:43:16.680 all over the place. But one of the things that 0:43:16.719 --> 0:43:21.359 they have argued for in this round of negotiations is, hey, 0:43:21.440 --> 0:43:25.520 you can't replace us with AI, and the studios all 0:43:25.560 --> 0:43:28.880 came back and said, well, how about this, We'll assess 0:43:28.960 --> 0:43:33.240 that on a year to year basis. And that's frightening 0:43:33.320 --> 0:43:36.520 if you're if you're either a writer in Hollywood or 0:43:36.680 --> 0:43:41.239 you're somebody who loves TV and films and quality TV 0:43:41.360 --> 0:43:46.120 in films, because I don't know if I think ideation 0:43:46.800 --> 0:43:52.319 and initial scripts maybe even right now, could I could 0:43:52.320 --> 0:43:54.359 see that happening where they said they're like, all right now, 0:43:54.400 --> 0:43:58.120 we'll bring in a human to refine this thing at 0:43:58.120 --> 0:44:01.000 a much lower wage. That it's probably what they're most 0:44:01.000 --> 0:44:04.680 afraid of, rather than being wholesale replaced, because like you said, 0:44:04.920 --> 0:44:09.160 these programs are they're all about just data and numbers. 0:44:09.200 --> 0:44:12.120 They're not They don't have human feelings, and that's what 0:44:12.560 --> 0:44:15.319 art is. And so I think I would be more 0:44:15.320 --> 0:44:19.880 concerned if I was writing pamphlets for Verizon or something, 0:44:20.719 --> 0:44:21.759 or if I was. 0:44:22.040 --> 0:44:23.800 Some pamphlet writer for Verizon. 0:44:23.880 --> 0:44:28.000 Just went in gulp, No, I'm so sorry. But like 0:44:28.239 --> 0:44:30.720 BuzzFeed back in the day, instead of having a dozen 0:44:30.760 --> 0:44:34.320 writers writing clickbait articles, why not have just one human 0:44:34.680 --> 0:44:38.560 that is a prompt engineer that's managing a virtual AI 0:44:39.600 --> 0:44:43.799 clickbait room that's just pumping out these articles that you 0:44:43.840 --> 0:44:46.160 know they were paying someone down to forty grand a 0:44:46.200 --> 0:44:47.160 year to write previously. 0:44:47.760 --> 0:44:50.160 Yeah, I mean it's a great question, like that was 0:44:50.200 --> 0:44:54.160 a horrific, horrible job to have. Not too many years ago. 0:44:54.320 --> 0:44:56.319 So it's great to have a computer do it, but 0:44:56.719 --> 0:44:58.839 that means that we need these other people to go 0:44:58.920 --> 0:45:02.040 on to be to have writing jobs that are more 0:45:02.080 --> 0:45:05.560 satisfying to them than that. But that's not necessarily the case, 0:45:05.640 --> 0:45:10.360 because as these things get smarter and better, they're just 0:45:10.400 --> 0:45:12.120 going to be relied upon more. We're not going to 0:45:12.160 --> 0:45:15.160 go back. There's no going back now. It just happened 0:45:15.200 --> 0:45:18.719 like it just happened basically as of March twenty twenty three. 0:45:19.480 --> 0:45:23.600 And one of the big problems that people have already 0:45:23.640 --> 0:45:31.280 projected running into is if computers replace humans, say writers, 0:45:33.200 --> 0:45:36.480 basically entirely. Eventually all the stuff that humans have written 0:45:36.480 --> 0:45:38.799 on the Internet is going to become dated. It's going 0:45:38.840 --> 0:45:41.560 to stop, yeah, and it will have been replaced and 0:45:41.640 --> 0:45:48.640 picked up on by generative, pre trained transformers. Right, And 0:45:48.680 --> 0:45:52.040 eventually all the writing on the Internet after a certain 0:45:52.120 --> 0:45:54.480 date will have been written by computers, but will be 0:45:54.480 --> 0:45:57.600 being scraped by computers. When humans go ask the computer 0:45:57.680 --> 0:46:01.160 a question, the computer then goes in reference is something 0:46:01.200 --> 0:46:04.040 written by a computer, So humans will be completely taken 0:46:04.080 --> 0:46:06.239 out of the equation in that in that respect, we'll 0:46:06.239 --> 0:46:09.360 be getting all of our information, at least non historical 0:46:09.400 --> 0:46:13.279 information from non humans, and that could be a really 0:46:13.320 --> 0:46:15.839 big problem, not just in the fact that we're losing 0:46:16.000 --> 0:46:18.959 jobs or in the fact that computers are now telling 0:46:19.040 --> 0:46:21.600 us all of our information, but also that there's some 0:46:23.400 --> 0:46:27.520 there's some part of what humans put into things that 0:46:27.600 --> 0:46:31.120 will be lost that I think we're going to demand. 0:46:31.239 --> 0:46:34.319 I saw somebody put it like, I think, I can't 0:46:34.320 --> 0:46:37.000 remember who it was, but they said, we're going like 0:46:37.280 --> 0:46:39.960 people will go seek out human written stuff. There will 0:46:40.000 --> 0:46:43.040 always be audiences for human written stuff. Yeah. Maybe, like 0:46:43.080 --> 0:46:47.120 you said, will rely on computers to write the Verizon pamphlets, 0:46:47.480 --> 0:46:49.439 but we're not going to rely on computers to write 0:46:49.440 --> 0:46:52.200 great works of literature or to create great works of art, 0:46:52.320 --> 0:46:55.120 Like we're just not going to They'll still do that. 0:46:55.160 --> 0:46:56.840 They're going to be writing books and movies and all that, 0:46:56.880 --> 0:46:59.640 but there will always be a taste in a market 0:46:59.680 --> 0:47:03.399 for human created stuff. This guy said, I think he's right. 0:47:04.400 --> 0:47:08.440 Yeah. And Justine Bateman, I don't know if you saw that. 0:47:09.560 --> 0:47:10.800 I don't know if it was a blog poster. 0:47:11.880 --> 0:47:14.080 Are you having a hallucination right now? Did you mean 0:47:14.200 --> 0:47:14.880 Justine Bateman? 0:47:15.400 --> 0:47:20.560 Yeah, yeah, Justin Bateman, the Jason Bateman's sister the actor, 0:47:20.680 --> 0:47:22.960 and she's done all kinds of things since then. I 0:47:23.000 --> 0:47:26.440 know she has a computer science degree, so she's very 0:47:26.440 --> 0:47:28.880 smart and knows a lot about this stuff. But she 0:47:29.000 --> 0:47:32.080 basically said, and this is beyond just the chatbot stuff. 0:47:32.520 --> 0:47:36.400 But she was like, right now, there are major Hollywood 0:47:36.400 --> 0:47:41.719 stars being scanned, and there may be a brand new 0:47:41.760 --> 0:47:45.600 Tom Cruise movie in sixty years. Yeah, long after he's 0:47:45.680 --> 0:47:49.239 dead starring Tom Cruise. He may be making movies for 0:47:49.239 --> 0:47:52.480 the next two hundred years. And like, is this what 0:47:52.520 --> 0:47:55.279 you want? Actors? Do you want to be scanned and 0:47:55.400 --> 0:47:58.880 have them use your image like this in perpetuity for 0:47:59.160 --> 0:48:02.120 you know, there will be money involved. It's not like 0:48:02.160 --> 0:48:03.640 they can just say, Okay, we can just do whatever 0:48:03.640 --> 0:48:06.360 we want. But what if they're like, here's a billion dollars, 0:48:06.480 --> 0:48:11.520 Tom Cruise, just for the use of your image in perpetuity, 0:48:11.760 --> 0:48:15.239 because we will be able to duplicate that so realistically 0:48:15.800 --> 0:48:20.320 that people won't know human voices, same thing that's already happening. 0:48:20.880 --> 0:48:26.400 What yeah, it is. So that stuff is kind of scary. 0:48:26.560 --> 0:48:29.400 And you know when you read I didn't really know 0:48:29.440 --> 0:48:32.960 this was kind of already happening in companies. But Olivia 0:48:33.040 --> 0:48:37.080 found this stuff IBM CEO Arvin Krishna said just last 0:48:37.080 --> 0:48:39.879 month in May that he believed thirty percent of back 0:48:39.880 --> 0:48:44.719 office jobs could be replaced over five years, and it 0:48:44.800 --> 0:48:48.760 was pausing hiring for close to eight thousand positions because 0:48:48.800 --> 0:48:52.160 they might be able to use AI instead. And then 0:48:52.280 --> 0:48:56.759 Dropbox talked about the AI era when they announced a 0:48:56.840 --> 0:49:00.239 round of layoffs. So it is happening right now in 0:49:00.320 --> 0:49:00.719 real time. 0:49:00.800 --> 0:49:04.680 Pretty amazing. Yeah, that's I mean, there's proof positive right there, 0:49:05.040 --> 0:49:08.479 Like that guy couldn't even wait a couple months a year, 0:49:09.120 --> 0:49:12.280 Like this really started up in March, and he's saying 0:49:12.320 --> 0:49:14.719 this in May. Already in May, they're like, wait, wait, 0:49:14.760 --> 0:49:17.759 stop hiring. We're gonna eventually replace these guys with AI 0:49:18.160 --> 0:49:21.480 so soon that we're going to stop hiring those positions 0:49:21.480 --> 0:49:24.040 for now until the AI is competent enough to take over. 0:49:24.600 --> 0:49:27.319 I mean, how many people does ib employ? What, what's 0:49:27.360 --> 0:49:28.160 thirty percent of that? 0:49:28.400 --> 0:49:30.960 I don't know. I would say at least at least 0:49:31.000 --> 0:49:34.879 one hundred people, right, So, yeah, like you said, it's 0:49:34.880 --> 0:49:37.040 happening already. And then one other thing to look out 0:49:37.040 --> 0:49:40.600 for too, that's I believe is already at least theoretically 0:49:40.640 --> 0:49:45.879 possible since AI can write code. Now they'll be able 0:49:45.920 --> 0:49:50.640 to create new large language models themselves, So the computers 0:49:50.640 --> 0:49:53.400 will be able to create new AI. 0:49:54.080 --> 0:49:56.160 Well, that's the singularity, right. 0:49:56.320 --> 0:50:00.799 No, the singularity is when one of them a understands 0:50:00.880 --> 0:50:05.480 what it is and become Yes, that's the singularity. 0:50:05.840 --> 0:50:07.560 But this leads to that though, doesn't it It does. 0:50:07.840 --> 0:50:11.600 It's hypothetically yes, but we just understand what's going on 0:50:11.680 --> 0:50:13.880 so little that you just can't say either way. Really, 0:50:13.920 --> 0:50:16.239 you definitely can't say that it no, it won't happen. 0:50:16.280 --> 0:50:18.359 It's just fantasy. And you also can't say yes, it's 0:50:18.360 --> 0:50:19.280 definitely gonna happen. 0:50:19.960 --> 0:50:25.520 Yeah. And here's the thing, man, I'm not a paranoid technophobe. 0:50:26.080 --> 0:50:27.960 You don't any measure foiled cap on. 0:50:28.719 --> 0:50:33.120 No, by any measure. I'm a pretty positive thinker, and 0:50:33.200 --> 0:50:35.239 this this is pretty scary to me. 0:50:37.560 --> 0:50:44.160 I'm just gonna leave that there, agreed, Chuck. Okay, if 0:50:44.200 --> 0:50:47.520 you want to know more about large language models, everybody 0:50:47.640 --> 0:50:51.279 just to start looking around Earth and when you see 0:50:51.320 --> 0:50:54.000 people running from explosions, go toward it and ask what's 0:50:54.040 --> 0:50:54.440 going on? 0:50:55.040 --> 0:50:56.800 You almost said, type it into a search engine. 0:50:57.080 --> 0:50:59.520 Yeah, steer clear of those. Yeah, there's so much more, 0:50:59.520 --> 0:51:02.120 we could have talked about. But this is if you 0:51:02.160 --> 0:51:04.520 ask me, this is round one. I think we definitely 0:51:04.560 --> 0:51:06.000 need to do at least one or so. 0:51:06.080 --> 0:51:08.680 More on this, Okay, Yeah, and then one day, like 0:51:08.719 --> 0:51:11.600 I said, Ai, Josh and Chuckle, just wrap it all 0:51:11.680 --> 0:51:13.480 up and spank it on the bottom and say no 0:51:13.640 --> 0:51:14.279 problems here. 0:51:14.400 --> 0:51:17.120 Hopefully they'll give us a billion dollars rather than like 0:51:17.160 --> 0:51:20.040 a month free of blue Apron instead. 0:51:20.680 --> 0:51:22.359 Yeah, I mean we could talk here. 0:51:23.840 --> 0:51:27.000 Well, since Chuck said we can talk real confidential like 0:51:27.400 --> 0:51:28.760 that means it's time for listener. 0:51:28.800 --> 0:51:34.880 May I'm gonna call this conception, not inception, but conception. 0:51:35.040 --> 0:51:36.680 Oh I saw this one. I don't know how I 0:51:36.719 --> 0:51:37.319 feel about this. 0:51:39.000 --> 0:51:41.200 Hey, guys. Last year, my wife and I were attempting 0:51:41.239 --> 0:51:43.040 to get pregnant. A couple of months in we made 0:51:43.040 --> 0:51:45.040 plans to stay with some friends in another town for 0:51:45.080 --> 0:51:47.600 a weekend, and when we can arrive, it happened to 0:51:47.640 --> 0:51:52.279 coincide with my wife's ovulation cycle. As shy people, we 0:51:52.320 --> 0:51:55.800 both felt a little bit awkward about, you know, hugging 0:51:55.880 --> 0:51:58.960 and kissing in a friend's guest room, but we really 0:51:59.040 --> 0:52:01.360 didn't want to miss that chance and that time of 0:52:01.360 --> 0:52:04.319 the month, so we went about getting in the mood 0:52:04.320 --> 0:52:07.080 as quietly as possible, and my wife suggested we play 0:52:07.080 --> 0:52:09.600 a podcast from my phone so that, you know, if 0:52:09.640 --> 0:52:12.440 any noise is made outside the room, it would sound 0:52:12.480 --> 0:52:14.560 like we were just doing a little pre bedtime listening. 0:52:15.920 --> 0:52:18.360 I knew I needed something with nice, simple production values, 0:52:18.440 --> 0:52:21.120 so we wouldn't get distracted, of course, by the whiz 0:52:21.200 --> 0:52:23.640 bang sounds and whatnot. And since you were my intro 0:52:23.760 --> 0:52:26.040 to the world of podcast, I've always had a steady 0:52:26.040 --> 0:52:29.840 supply of yours downloaded. I picked the least interesting sounding 0:52:29.880 --> 0:52:33.640 one in the feed at the time, How Coal Works. 0:52:33.960 --> 0:52:38.600 Okay, I thought that one turned out to be surprisingly interesting. Yeah, 0:52:38.640 --> 0:52:40.239 I could see how we would have thought that though. 0:52:40.800 --> 0:52:43.359 Yeah, for sure. We put that on and we did 0:52:43.360 --> 0:52:47.000 our business. Six weeks later we got a positive pregnancy test, 0:52:47.880 --> 0:52:50.439 and now over a year later, we've welcomed our son 0:52:50.520 --> 0:52:53.640 into the world. Name is Cole, and of course we 0:52:53.719 --> 0:52:57.279 named him Cole. That is what this person said. 0:52:57.400 --> 0:52:59.200 Wait a minute, Wait a minute, they really did name 0:52:59.239 --> 0:52:59.680 them Cole. 0:53:00.600 --> 0:53:04.319 No, he said it as a joke. Oh but great 0:53:04.360 --> 0:53:07.400 minds right, good joke for both of them. It's almost 0:53:07.440 --> 0:53:13.120 like you're both chatbox. And this person said You're fine 0:53:13.160 --> 0:53:15.600 to read this, but give me a fake name. And 0:53:15.640 --> 0:53:18.360 so I just want to say thanks to Gene for 0:53:18.440 --> 0:53:19.520 writing in about this. 0:53:20.040 --> 0:53:21.840 Gene is in Gene Transfer. 0:53:23.000 --> 0:53:25.400 Sure, thanks a lot, Gene. 0:53:25.400 --> 0:53:29.759 We appreciate that. I think again, I'm still figuring that 0:53:29.800 --> 0:53:32.719 one out. And if you want to be like Gene, 0:53:32.840 --> 0:53:36.360 I'm making air quotes here, you can send us an 0:53:36.400 --> 0:53:39.160 email to wrap it up. Spank it on the bottom. 0:53:39.320 --> 0:53:40.400 Only humans can do that. 0:53:41.120 --> 0:53:43.160 I wonder if when you said spanking on the bottom, 0:53:43.239 --> 0:53:44.840 if that created any issues. 0:53:45.160 --> 0:53:49.760 Yeah, I hadn't thought about that. Maybe playfully how about that? Sure, 0:53:51.080 --> 0:53:58.680 and send it off to stuff podcast at iHeartRadio dot com. 0:53:58.920 --> 0:54:01.799 Stuff you Should Know is a production of iHeartRadio. For 0:54:01.880 --> 0:54:06.080 more podcasts my heart Radio, visit the iHeartRadio app, Apple Podcasts, 0:54:06.200 --> 0:54:08.040 or wherever you listen to your favorite shows.