1 00:00:15,960 --> 00:00:18,560 Jamie, right. So we're in a very interesting 2 00:00:18,560 --> 00:00:23,840 time in humanity, the Internet, AI conversations, echo chambers, 3 00:00:23,840 --> 00:00:27,800 you know, all the trends, right? What we're going to talk about 4 00:00:27,800 --> 00:00:29,720 the current set of the Internet, where we are right now and the 5 00:00:29,720 --> 00:00:31,160 subcultures that exist within it. 6 00:00:31,480 --> 00:00:34,960 But before we do all of that, it'll be great as an 7 00:00:34,960 --> 00:00:37,680 introduction for the guests that do not know about your work. 8 00:00:37,920 --> 00:00:40,040 Can you tell us about your work, what you've been up to? 9 00:00:40,040 --> 00:00:42,080 And then we'll dive into the bottlenecks that you've been 10 00:00:42,080 --> 00:00:44,960 solving and that you've seen from the start of your career 11 00:00:44,960 --> 00:00:48,480 till the year is 2020. Right, thanks. 12 00:00:48,480 --> 00:00:50,720 Thanks for having me. I guess I've for the last 15 13 00:00:50,720 --> 00:00:54,360 years, I've just been trying to explain what strange things have 14 00:00:54,360 --> 00:00:57,880 been going on in our digital universe to pretty much to 15 00:00:57,880 --> 00:01:00,960 ordinary people. But there's always these huge 16 00:01:00,960 --> 00:01:04,400 changes in technology happening. And about two to three years 17 00:01:04,400 --> 00:01:07,800 later or sort of the ordinary person is sort of hit, the wave 18 00:01:07,800 --> 00:01:10,120 hits them and they don't know what's going on. 19 00:01:10,480 --> 00:01:12,800 And I feel like I'm always trying to be like one year ahead 20 00:01:12,800 --> 00:01:14,640 of that wave that's hitting the ordinary. 21 00:01:14,640 --> 00:01:17,680 I see what's going on and I can try and work out what it means 22 00:01:17,680 --> 00:01:21,080 for people and then explain it because so much really about 23 00:01:21,080 --> 00:01:22,520 what technology does to the world. 24 00:01:22,520 --> 00:01:25,640 It's not about the tech like the specifics of machine learning or 25 00:01:25,640 --> 00:01:29,480 the specifics of how dark net encryption works. 26 00:01:29,800 --> 00:01:32,680 It's what happens when when people are using that. 27 00:01:32,680 --> 00:01:34,560 How does it change our behaviour? 28 00:01:34,560 --> 00:01:36,720 How does it change how we relate to each other, how we understand 29 00:01:36,720 --> 00:01:39,280 information, how we understand politics, media. 30 00:01:39,840 --> 00:01:42,720 So that's been my, my, my thing throughout and I've just always 31 00:01:42,720 --> 00:01:47,480 been fascinated by the, the, the weird subcultures that are often 32 00:01:47,920 --> 00:01:50,720 foretelling how the rest of us will behave. 33 00:01:50,720 --> 00:01:54,600 So I, I've actually started working on the, the way, the way 34 00:01:54,600 --> 00:01:55,920 I go into this is so weird, right? 35 00:01:56,240 --> 00:02:01,200 So weird. 2009 There is this football hooligan group called 36 00:02:01,200 --> 00:02:05,360 the English Defence League going around different cities doing on 37 00:02:05,480 --> 00:02:08,440 offline protests about immigration, Islam and stuff 38 00:02:08,440 --> 00:02:09,960 like that, led by Tommy Robinson. 39 00:02:10,840 --> 00:02:14,720 And I realised this was not an offline football hooligan group. 40 00:02:14,720 --> 00:02:19,120 This was a Facebook group with a tiny militant offline wing. 41 00:02:19,880 --> 00:02:21,960 And I thought, I wonder how I could get to the Facebook group 42 00:02:21,960 --> 00:02:23,280 because the Facebook group was huge. 43 00:02:23,280 --> 00:02:25,720 It was bigger than the political parties, it was massive and no 44 00:02:25,720 --> 00:02:28,480 one's looking at that. So I think what if I run a 45 00:02:28,480 --> 00:02:32,920 Facebook advert like you were trying to sell jeans, targeting 46 00:02:32,920 --> 00:02:36,240 supporters of the English Defence League, and rather than 47 00:02:36,240 --> 00:02:39,520 selling them something, I'll ask them to fill in a survey about 48 00:02:39,520 --> 00:02:42,720 their opinions. I'll spend £20 on this on Friday 49 00:02:42,720 --> 00:02:46,160 night, come back on Monday morning, 1000 members of the EDL 50 00:02:46,160 --> 00:02:49,880 had filled in my survey. I was like, this whole the the 51 00:02:49,880 --> 00:02:53,640 digital world is offering us an amazing way to get to people 52 00:02:53,640 --> 00:02:56,720 that you'd never see, you'd never meet, you'd never find, 53 00:02:56,760 --> 00:03:00,600 would never talk to you. And it's also changing the way 54 00:03:00,600 --> 00:03:03,160 political movements are being shaped. 55 00:03:03,400 --> 00:03:07,240 And from that point on, I was working on online conspiracies I 56 00:03:07,240 --> 00:03:12,120 wrote about, I set up a research group looking at AI in 2010. 57 00:03:12,120 --> 00:03:16,600 I wrote a book about the dark net in 20 14 where I was, you 58 00:03:16,600 --> 00:03:22,240 know, spending loads of time on Internet like weird subcultures, 59 00:03:22,240 --> 00:03:25,680 neo Nazi forums, darknet drugs markets, eating disorder forums, 60 00:03:25,840 --> 00:03:27,400 meeting the people that were part of them. 61 00:03:27,760 --> 00:03:29,240 And it just kind of went on from that. 62 00:03:29,240 --> 00:03:32,240 I did a podcast with the BBC about the missing crypto queen, 63 00:03:32,800 --> 00:03:35,560 biggest crypto Ponzi scam in history that we uncovered. 64 00:03:36,000 --> 00:03:38,120 So I'm going on and on because that shows how old I am. 65 00:03:38,120 --> 00:03:40,880 I've been around for so long. It's like when you scroll the 66 00:03:40,880 --> 00:03:47,960 date on a on a ticket. And it gets longer and longer. 67 00:03:48,120 --> 00:03:51,840 Could they just start in 1980s? Surely AI just to be kind. 68 00:03:51,840 --> 00:03:54,320 Of AI could sort of do a screen grab and be like, let's start 69 00:03:54,320 --> 00:03:57,400 him off somewhere a bit closer to his date of birth, save him 70 00:03:57,400 --> 00:03:59,520 some time. And then most recently, I've 71 00:03:59,520 --> 00:04:03,360 written this book called How to Talk to AI and How Not To about 72 00:04:03,360 --> 00:04:06,880 these language models that seemingly overnight have turned 73 00:04:06,880 --> 00:04:11,240 up and suddenly a billion people around the world are using them 74 00:04:11,600 --> 00:04:15,880 constantly for everything and really don't understand, do not 75 00:04:15,880 --> 00:04:18,079 know how they work, what they do, their strengths, their 76 00:04:18,079 --> 00:04:20,920 limits, nothing. So yeah, it's been, it's been 77 00:04:20,920 --> 00:04:23,400 pretty wild, actually. So how did we get here? 78 00:04:23,400 --> 00:04:25,520 Because you see a constant trend. 79 00:04:25,520 --> 00:04:28,360 For example, there was the dark boom and the Internet, what was 80 00:04:28,360 --> 00:04:31,920 happening and social came about. And then what you get is you 81 00:04:31,920 --> 00:04:35,240 always get what I like to call like the the nerds and the 82 00:04:35,240 --> 00:04:37,080 enthusiasts. That started off right. 83 00:04:37,320 --> 00:04:39,800 And I'm like, right now I'm in an echo chamber where I believe 84 00:04:39,800 --> 00:04:42,680 that everybody is like experimenting and building stuff 85 00:04:42,680 --> 00:04:44,560 on Claude. And then I go outside of that 86 00:04:44,560 --> 00:04:46,920 echo chamber and I speak to everybody else and they don't 87 00:04:46,920 --> 00:04:48,800 even know what I'm speaking about, right? 88 00:04:49,240 --> 00:04:51,640 And, and I've seen that trend happen at social and I've seen 89 00:04:51,640 --> 00:04:54,160 that trend happen again with AI. Well, crypto, we're talking 90 00:04:54,160 --> 00:04:56,520 about the crypto queen. And I'm seeing the things, the 91 00:04:56,520 --> 00:04:57,840 same thing happening again with AI. 92 00:04:57,840 --> 00:05:00,360 And what happens is you get a small bunch of people that are 93 00:05:00,360 --> 00:05:04,440 early, they do some stuff, everybody else hears about it, 94 00:05:04,440 --> 00:05:07,040 but they stay in a hole and then it impacts them. 95 00:05:07,040 --> 00:05:10,200 And then the next thing comes in, same trend, same trend 96 00:05:10,200 --> 00:05:11,840 changer. Why does that seem to happen 97 00:05:11,840 --> 00:05:14,920 that way? Well, I think there's a, it's a 98 00:05:14,920 --> 00:05:17,360 good question. And I've always found like when 99 00:05:17,360 --> 00:05:20,480 I was when I was working on the dark net, the thing I found was 100 00:05:20,480 --> 00:05:24,840 that it was, it was crypto anarchists and drug dealers who 101 00:05:24,840 --> 00:05:29,440 were pioneering how to use crypto coins, encrypted 102 00:05:29,440 --> 00:05:31,480 messaging systems. It were, these were the guys 103 00:05:31,480 --> 00:05:34,040 that were building all this technology first working out how 104 00:05:34,040 --> 00:05:38,520 to use it and then suddenly went mainstream five years later. 105 00:05:38,680 --> 00:05:40,400 The same with the English Defence League. 106 00:05:40,400 --> 00:05:42,800 It was the IT was groups like that that were figuring out 107 00:05:43,720 --> 00:05:48,400 social media is a new form of sort of political persuasion, 108 00:05:48,400 --> 00:05:52,400 Like we, we can form new groups. And I think the reason is maybe 109 00:05:52,400 --> 00:05:54,320 it was always this way to some extent. 110 00:05:54,680 --> 00:05:59,840 Outsiders, criminals, radicals, people on the fringes often feel 111 00:05:59,840 --> 00:06:04,200 like the mainstream system, the legacy systems block them out. 112 00:06:04,480 --> 00:06:07,240 They don't, they don't let them. They don't let them in. 113 00:06:07,240 --> 00:06:09,720 They silence them. They, the media won't pay 114 00:06:09,720 --> 00:06:12,040 attention to them. The established financial 115 00:06:12,040 --> 00:06:13,440 services won't help them of them. 116 00:06:13,680 --> 00:06:17,200 So they are often quite early adopters of any new technology 117 00:06:17,400 --> 00:06:18,800 and they're experimenting with it. 118 00:06:18,800 --> 00:06:21,400 They're figuring out solutions, opportunities. 119 00:06:21,400 --> 00:06:25,120 How could I use Bitcoin to create a payment mechanism for 120 00:06:25,120 --> 00:06:29,560 anonymous drug deals? So I think when there's because 121 00:06:29,560 --> 00:06:33,320 because the all these new media systems and payment systems and 122 00:06:33,320 --> 00:06:35,440 all these different technological systems have 123 00:06:35,440 --> 00:06:39,880 arrived, outsiders are like, whoa, I wonder if I could, I 124 00:06:40,000 --> 00:06:41,680 wonder if I could use it for this thing or that thing. 125 00:06:41,680 --> 00:06:44,160 And then and then they come up with amazing solutions and 126 00:06:44,160 --> 00:06:47,480 possibilities that other people then think, oh, Bitcoin could 127 00:06:47,480 --> 00:06:49,280 actually be quite good for another thing too. 128 00:06:49,280 --> 00:06:51,800 And you've already worked it out and now I'm going to try it. 129 00:06:52,200 --> 00:06:55,080 And then JP Morgan and Barclays Bank start investing billions in 130 00:06:55,080 --> 00:06:58,120 it and it's made all the drug dealers extremely rich. 131 00:06:58,400 --> 00:07:00,480 It seems like they. Because they hold thousands of 132 00:07:00,480 --> 00:07:02,880 these bitcoins. Oh, so everybody who was in 133 00:07:02,880 --> 00:07:06,400 early, of course, which was, which was to some extent the 134 00:07:06,400 --> 00:07:09,760 crypto curious. I mean, I, I ended up making, 135 00:07:09,840 --> 00:07:12,880 got a lot of money simply because I bought a few Bitcoin 136 00:07:13,080 --> 00:07:14,840 to see what the interface was like. 137 00:07:14,840 --> 00:07:17,720 I wasn't really thinking of it as a currency because I'm, I was 138 00:07:17,720 --> 00:07:21,000 convinced that the user interface was deeply flawed and 139 00:07:21,000 --> 00:07:23,520 that there needed to be a way to repackage it to make it more 140 00:07:23,520 --> 00:07:26,120 comprehensible. So I was purely pissing around 141 00:07:26,120 --> 00:07:28,600 and then ended up making a sort of five figure sum. 142 00:07:28,920 --> 00:07:32,840 I also bought, I also I also bought a hardback book at one 143 00:07:32,840 --> 00:07:36,440 point which has subsequently cost me £8000 because I bought a 144 00:07:36,440 --> 00:07:40,440 hardback book for 1/4 of a Bitcoin back in the early days. 145 00:07:41,360 --> 00:07:43,160 But by the way, they're far worse stories than that. 146 00:07:43,160 --> 00:07:45,200 They're people who have a Chinese meal that's cost them 147 00:07:45,200 --> 00:07:46,880 half a million. I've got some I've got some 148 00:07:46,880 --> 00:07:49,560 really bad stories as well. We many of us have, but I was 149 00:07:49,560 --> 00:07:53,680 out in a crypto anarchist commune in Prague like this was 150 00:07:53,680 --> 00:07:58,320 when Bitcoin was maybe 10 dollars $15 and just and you 151 00:07:58,320 --> 00:08:01,680 know, and you could only buy things in Bitcoin there nothing 152 00:08:01,680 --> 00:08:04,000 they wouldn't take Fiat currency. 153 00:08:04,320 --> 00:08:09,240 So I I found the little payment card that I had from that trip 154 00:08:09,240 --> 00:08:12,960 about two years ago and realised they had some small change 155 00:08:12,960 --> 00:08:15,040 leftover from a coffee I'd bought. 156 00:08:15,440 --> 00:08:18,280 And that small change was about £500 worth of Bitcoin. 157 00:08:18,320 --> 00:08:20,720 I was like, how much was the coffee? 158 00:08:20,720 --> 00:08:22,960 How much do you think? I don't know? 159 00:08:23,360 --> 00:08:28,320 10 thousand 20,000 lbs. Yeah, but you're right, the 160 00:08:28,320 --> 00:08:30,920 curious, the people that were experimenting and it did make a 161 00:08:30,920 --> 00:08:34,400 lot of people a lot of money who were already like interested in 162 00:08:34,400 --> 00:08:36,159 the alternatives. What can we do differently? 163 00:08:36,159 --> 00:08:38,440 What's weird, I don't trust. And it's a small percentage of 164 00:08:38,440 --> 00:08:40,080 the population, broadly speaking. 165 00:08:40,120 --> 00:08:42,840 Yeah, very small population. Yeah, the population. 166 00:08:42,840 --> 00:08:46,480 They usually say that where you find new innovation is usually 167 00:08:46,920 --> 00:08:50,080 with the military or in vice industries, right? 168 00:08:50,080 --> 00:08:51,000 Pornography. Yeah, yeah, yeah. 169 00:08:51,000 --> 00:08:53,120 They're usually. Online payment was pioneered by 170 00:08:53,120 --> 00:08:55,360 the porn industry, I think. That's true. 171 00:08:55,360 --> 00:08:58,640 And so was pop up, pop up ads were porn, porn, porn created 172 00:08:59,400 --> 00:09:05,360 and a lot of a lot of self the sort of camming community, the 173 00:09:05,360 --> 00:09:08,960 webcamming community, they were all fascinated by crypto. 174 00:09:08,960 --> 00:09:11,200 I spent a lot of time with the web, with web casts, actually 175 00:09:11,200 --> 00:09:15,520 sat in on a big threesome as one for one of my in one of my 176 00:09:15,520 --> 00:09:18,680 books, like behind the scenes watching how they were doing it. 177 00:09:19,080 --> 00:09:22,520 And they were obsessed with crypto because they were like, 178 00:09:22,520 --> 00:09:25,440 we're so sick of these platforms taking such a big cut of our 179 00:09:25,440 --> 00:09:28,160 money. Is there a peer-to-peer system? 180 00:09:28,160 --> 00:09:30,280 So they were, they were getting loads they would make, they 181 00:09:30,280 --> 00:09:32,720 would have made loads of money if you know the people that got 182 00:09:32,720 --> 00:09:35,400 into that. So it is, it's true. 183 00:09:35,760 --> 00:09:38,200 It's true because they have to be, they have to pioneer, they 184 00:09:38,200 --> 00:09:42,640 have to try something different. Criminals are also so not just 185 00:09:42,640 --> 00:09:47,200 the military, but the criminals they're always looking for, like 186 00:09:47,200 --> 00:09:50,560 how could we use these technologies really cleverly? 187 00:09:50,640 --> 00:09:55,440 One, how do you stay a step away from mass adoption and 188 00:09:55,440 --> 00:09:58,480 litigation, right? So when you go into a new place 189 00:09:58,480 --> 00:10:02,400 that's not been governed yet, that's where a lot of the the 190 00:10:02,400 --> 00:10:04,480 upside is driven in these industries, right? 191 00:10:04,520 --> 00:10:06,960 That's why they're constantly looking for that innovation. 192 00:10:06,960 --> 00:10:09,440 That's right. And now that world, which was 193 00:10:09,440 --> 00:10:15,240 this exciting, weird, pioneering online space for the radicals 194 00:10:15,240 --> 00:10:18,400 and the outsiders to get together and experiment and try 195 00:10:18,400 --> 00:10:21,840 new things, suddenly became the thing all of us were on all the 196 00:10:21,840 --> 00:10:24,320 time. So those got, you know, those 197 00:10:24,320 --> 00:10:26,480 people were so early to it and doing interesting things. 198 00:10:26,480 --> 00:10:28,720 And that's one of the reasons why you'll often find these 199 00:10:28,720 --> 00:10:32,720 outside movements. They do so well online. 200 00:10:33,360 --> 00:10:37,680 Like so much of the the early Donald Trump push was was weird. 201 00:10:37,720 --> 00:10:42,960 4 Chan and Reddit groups, the Pepe the Frog memes, the Red 202 00:10:42,960 --> 00:10:47,560 Pill community. People that were 16/17/18 when 4 203 00:10:47,560 --> 00:10:52,000 Chan and Reddit first exploded are now like geriatric 204 00:10:52,000 --> 00:10:57,040 millennials who are the core voters of Donald Trump. 205 00:10:58,320 --> 00:11:00,920 And do you see any similarities because going back to the crypto 206 00:11:01,000 --> 00:11:02,840 industry? And I think there's a lot of 207 00:11:02,840 --> 00:11:06,720 lessons that we can learn from that period with what's 208 00:11:06,720 --> 00:11:08,720 happening in AI now. What what are some of that? 209 00:11:08,840 --> 00:11:11,480 And we talked about the crypto Queen, like what was happening 210 00:11:11,480 --> 00:11:14,200 because how I got made aware of crypto. 211 00:11:14,200 --> 00:11:18,440 So I was at Gym Shark at the time around 2017 and everybody 212 00:11:18,440 --> 00:11:21,520 was buying like Ethereum and everyone was talking about it. 213 00:11:21,520 --> 00:11:24,160 So I wasn't early. It was still like more 214 00:11:24,160 --> 00:11:27,280 expensive, but I was early enough to put some money in. 215 00:11:27,360 --> 00:11:29,520 Do OK. So I was early, now the time is 216 00:11:29,560 --> 00:11:32,520 felt late, didn't it? But now we're sorry. 217 00:11:32,520 --> 00:11:35,240 Now you know it's early. Like it actually helped me with 218 00:11:35,240 --> 00:11:38,880 the deposit for a house right? You guys, do you realize I was 219 00:11:38,880 --> 00:11:40,920 sitting with these crypto millionaires here? 220 00:11:41,400 --> 00:11:45,080 Unfortunately not. But the cool thing, I think what 221 00:11:45,080 --> 00:11:47,400 I've learned is all about temperament and knowing when to, 222 00:11:47,680 --> 00:11:50,560 to, to let go, right? Because the big thing was not 223 00:11:50,720 --> 00:11:52,920 because a lot of people were just like going through the 224 00:11:52,920 --> 00:11:55,080 motions, right? And think the key thing was know 225 00:11:55,080 --> 00:11:58,480 your number, pull out and then live to fight another day. 226 00:11:58,480 --> 00:12:01,640 Whereas a lot of people just kept like putting like the, the, 227 00:12:01,920 --> 00:12:04,160 the, the, the things that they own into it. 228 00:12:04,160 --> 00:12:06,200 That's when it starts to get a little bit weird and crazy, 229 00:12:06,200 --> 00:12:08,400 right? But we've always saw with crypto 230 00:12:08,800 --> 00:12:12,080 how it became like a cult like thing and became widespread. 231 00:12:12,480 --> 00:12:14,840 And then you saw that it was with Web 3, right? 232 00:12:15,080 --> 00:12:16,840 You're seeing these trends that come about. 233 00:12:17,240 --> 00:12:20,360 A few people go in, they benefit from that hype. 234 00:12:20,680 --> 00:12:23,480 And then all of a sudden no one's speaking about it anymore. 235 00:12:23,960 --> 00:12:25,960 And I wanted to know whether the similarities with what we saw 236 00:12:25,960 --> 00:12:29,120 with crypto, crypto Web 3 and the blockchain space is what 237 00:12:29,120 --> 00:12:32,000 we're seeing in AI today. Is that like a trend that we 238 00:12:32,000 --> 00:12:37,360 should be paying attention to? Well, I mean, as a sort of as a 239 00:12:37,360 --> 00:12:41,520 meta rule, hype, just that the hype of every new technology on 240 00:12:41,520 --> 00:12:44,680 that, the hype cycle seems to get greater every time. 241 00:12:44,680 --> 00:12:46,640 But maybe it's because we're also desperate for something 242 00:12:46,640 --> 00:12:48,440 amazing to solve all of our problems. 243 00:12:48,440 --> 00:12:51,120 Particularly representative of what makes a difference to 244 00:12:51,120 --> 00:12:53,040 quality of life, by the way, is it no. 245 00:12:53,200 --> 00:12:57,840 So, you know, I would argue that, for example, it strikes me 246 00:12:57,840 --> 00:13:01,240 as strange, the widespread adoption of video conferencing 247 00:13:01,240 --> 00:13:04,240 strikes me as fundamentally important, and yet it receives 248 00:13:04,240 --> 00:13:06,880 very little hype because the technology itself is quite old. 249 00:13:07,200 --> 00:13:10,440 OK, so you, you remember that person on the Fast show used to 250 00:13:10,440 --> 00:13:12,640 run around going. Isn't electricity amazing? 251 00:13:12,640 --> 00:13:13,080 Yeah. Yeah. 252 00:13:13,400 --> 00:13:16,200 If you appear on the stage and you talk about the huge 253 00:13:16,200 --> 00:13:20,520 importance of, you know, the widespread adoption of, of video 254 00:13:20,560 --> 00:13:24,080 as a, as a meeting platform, as a means of communication, you 255 00:13:24,080 --> 00:13:28,880 look ridiculous. OK, so, and yet my contention is 256 00:13:28,880 --> 00:13:32,000 over a 10/15/20 year period, it's hugely important. 257 00:13:32,000 --> 00:13:35,000 One thing I've noticed is that although a lot of conventional 258 00:13:35,000 --> 00:13:37,920 companies are demanding their staff come back to the office, 259 00:13:38,360 --> 00:13:42,160 no one who starts a company now takes out significant office 260 00:13:42,160 --> 00:13:45,400 space and their employees probably work on three different 261 00:13:45,400 --> 00:13:46,400 continents. Yeah. 262 00:13:46,960 --> 00:13:49,040 OK. So the the people who are 263 00:13:49,040 --> 00:13:52,480 actually starting a business from scratch are starting a Zoom 264 00:13:52,480 --> 00:13:57,520 enabled business effectively, but existing companies are 265 00:13:57,520 --> 00:14:00,520 wrestling with it because they find the whole thing alarming 266 00:14:00,520 --> 00:14:03,520 and disruptive. And so it always struck me that 267 00:14:03,520 --> 00:14:07,960 the, you know, the noise around certain things is excessive 268 00:14:08,240 --> 00:14:13,760 generally and and also the timings usually bad because as 269 00:14:13,760 --> 00:14:17,080 you said, the really significant behavioural changes emerge for 270 00:14:17,080 --> 00:14:19,680 five years. Exactly and often in slightly 271 00:14:19,680 --> 00:14:20,720 unpredicted words. Yeah. 272 00:14:21,280 --> 00:14:22,680 I think that's so true with video. 273 00:14:22,680 --> 00:14:26,400 I've been thinking the same thing about video and how the 274 00:14:26,560 --> 00:14:29,840 rise of everyone talks about social media changing politics. 275 00:14:29,840 --> 00:14:31,920 I mean, I've spoken about it a lot as well. 276 00:14:33,120 --> 00:14:36,520 What no one really talks about is the way that the, the, the 277 00:14:36,520 --> 00:14:42,360 shift to video on social media is so more fundamental because 278 00:14:42,520 --> 00:14:45,880 at least with early social media, it was still words, it 279 00:14:45,880 --> 00:14:49,080 was still the work we work, still communicating in words, 280 00:14:49,080 --> 00:14:51,920 which has been the way we tended to do politics and media for 281 00:14:51,920 --> 00:14:55,040 very, very long time over the last five years. 282 00:14:55,160 --> 00:14:57,960 Everything is video now. It's all video. 283 00:14:57,960 --> 00:15:01,960 And video is such a different medium for the human psyche as 284 00:15:01,960 --> 00:15:03,360 well, how we understand the world. 285 00:15:03,360 --> 00:15:05,800 It's much more emotional. Marshall Mcluhan wrote loads 286 00:15:05,800 --> 00:15:09,360 about this when he was discussing TV and everything he 287 00:15:09,360 --> 00:15:11,320 wrote is, is basically right. He said we're all going to turn 288 00:15:11,320 --> 00:15:14,680 into, we're all going to go totally tribal with the TV 289 00:15:15,200 --> 00:15:16,720 because we've learned to be literate. 290 00:15:16,720 --> 00:15:19,480 We're thoughtful. It's slow, it's very rational. 291 00:15:20,080 --> 00:15:23,560 TV is a hot technology. It it, it sends you into 292 00:15:23,560 --> 00:15:26,680 emotional spirals really easily. You form new bonds with 293 00:15:26,680 --> 00:15:29,040 different people. It's very, very quick. 294 00:15:29,800 --> 00:15:31,280 And he was right. He was totally right. 295 00:15:31,280 --> 00:15:32,760 But it wasn't TV that has done that. 296 00:15:32,760 --> 00:15:35,680 It's been the adoption of video on social media that I think has 297 00:15:35,680 --> 00:15:37,120 done that. No, but no one talks about that. 298 00:15:37,480 --> 00:15:42,120 No one says that shift to video on social is transforming 299 00:15:42,120 --> 00:15:43,880 politics. Which politicians are going to 300 00:15:43,880 --> 00:15:47,040 be successful? Which political stories are 301 00:15:47,040 --> 00:15:50,480 going to run wild? Who's going to become really 302 00:15:50,480 --> 00:15:52,960 influential? What format we are all going to 303 00:15:52,960 --> 00:15:55,400 want to consume and like the length of it and the detail that 304 00:15:55,400 --> 00:15:59,200 we need to to make it. It's just passed unannounced. 305 00:16:01,480 --> 00:16:03,840 This is a fantastic. I mean, Jimmy Carr made the 306 00:16:03,840 --> 00:16:06,800 wonderful thing that he said YouTube became the largest TV 307 00:16:06,800 --> 00:16:08,880 station in the world and nobody noticed. 308 00:16:08,920 --> 00:16:11,960 Yeah. And the influence of that also, 309 00:16:11,960 --> 00:16:15,000 I think you know, actually in, in inverted commas, conventional 310 00:16:15,000 --> 00:16:19,600 media, TV rolling news has been catastrophic because a 311 00:16:19,600 --> 00:16:23,480 newspaper, for all its strengths and weaknesses, had some sort of 312 00:16:23,760 --> 00:16:28,320 a proportionality and B, it had, it had A effectively a sense of 313 00:16:28,320 --> 00:16:32,160 prioritisation, the front page versus the Backpage. 314 00:16:32,600 --> 00:16:35,920 What you find with rolling news, which drives me insane, is that 315 00:16:35,920 --> 00:16:41,080 there is a story du jour which utterly over dominates. 316 00:16:41,080 --> 00:16:44,880 I mean the Mandelson affair, for example, or Downing St. parties, 317 00:16:45,240 --> 00:16:49,680 both of them to some extent are confected now they, they deserve 318 00:16:49,920 --> 00:16:51,640 fairly significant news coverage. 319 00:16:51,640 --> 00:16:54,800 And in a newspaper, it would be 10% of the content of the 320 00:16:54,800 --> 00:16:57,840 newspaper. But in rolling TV news, it's 321 00:16:57,840 --> 00:17:00,240 sort of 70% of the news content. Yeah. 322 00:17:00,520 --> 00:17:01,680 And you know, this isn't actually. 323 00:17:01,800 --> 00:17:03,240 Yeah. I haven't actually seen any 324 00:17:03,240 --> 00:17:06,200 coverage of the Ukraine war for the last three months. 325 00:17:06,200 --> 00:17:07,000 Yeah. Yeah. 326 00:17:07,000 --> 00:17:10,240 Because suddenly the attention obviously flipped to Iran. 327 00:17:10,359 --> 00:17:12,359 Now nobody's talking much about Iran. 328 00:17:12,599 --> 00:17:17,280 And so the extent to which it's almost like ADHD imposed on you 329 00:17:17,880 --> 00:17:21,880 by rolling news channels strikes me as, I mean, just an example 330 00:17:21,880 --> 00:17:25,480 how the news has changed. When they dropped a bomb on 331 00:17:25,480 --> 00:17:30,040 Nagas Zaki, not Hiroshima, but Nagasaki, it was not the lead 332 00:17:30,040 --> 00:17:34,360 item in The Times the day afterwards, it was the number 333 00:17:34,360 --> 00:17:37,080 two item. Second atomic bomb dropped on 334 00:17:37,080 --> 00:17:39,480 Nagasaki. Now, what I was saying is that 335 00:17:39,760 --> 00:17:42,360 you had this whole sort of sense of prioritisation. 336 00:17:42,360 --> 00:17:44,680 There was a front page, there was an inside page. 337 00:17:45,080 --> 00:17:48,440 And broadly speaking, a newspaper would cover 27 topics 338 00:17:48,440 --> 00:17:53,240 or or 55. Rolling news is ludicrously 339 00:17:53,240 --> 00:17:55,400 disproportionate in the effect it has on us. 340 00:17:55,680 --> 00:17:58,040 But you're absolutely right. I mean, the the, the means by 341 00:17:58,040 --> 00:18:01,960 which we consume these things. And of course, the fact that we 342 00:18:01,960 --> 00:18:05,680 understand each other as attention fragments, we 343 00:18:05,680 --> 00:18:08,560 understand each other less well because we're all coming at the 344 00:18:08,560 --> 00:18:10,480 same thing from a completely different context. 345 00:18:10,480 --> 00:18:12,960 Yeah, exactly. And I think there's this is the, 346 00:18:12,960 --> 00:18:15,200 this is obviously the medium is the message. 347 00:18:15,320 --> 00:18:18,160 Mcluhan's most famous saying, a really simple way as well of 348 00:18:18,160 --> 00:18:21,600 illustrating it is, is, is is as as you to say about the 349 00:18:21,600 --> 00:18:24,920 prioritisation, the way that the role of the media was to help us 350 00:18:24,920 --> 00:18:28,040 prioritise what was important and certain things deserved more 351 00:18:28,040 --> 00:18:31,000 energy than others. Maybe it wasn't always right and 352 00:18:31,000 --> 00:18:32,800 maybe you could disagree with it, but at least you had the 353 00:18:32,800 --> 00:18:36,480 sense of an ordering of the importance of things and you 354 00:18:36,480 --> 00:18:38,800 could then sort of channel your emotion and anger 355 00:18:38,800 --> 00:18:40,960 proportionately. Whereas now everyone gets as 356 00:18:40,960 --> 00:18:44,320 furious about someone wrong speaking on Twitter as they 357 00:18:44,320 --> 00:18:47,600 would about some incredibly significant event in the world, 358 00:18:47,600 --> 00:18:50,720 because it's all blasting at you a kind of equal level. 359 00:18:51,200 --> 00:18:53,840 And a lot of this is confected, of course, and a lot of it 360 00:18:53,840 --> 00:18:56,920 reflects the particular concerns of journalists, not really the 361 00:18:56,920 --> 00:18:59,360 concerns of the public. There's certain things which are 362 00:18:59,360 --> 00:19:02,400 sort of natural triggers which the media will pull at every 363 00:19:02,400 --> 00:19:03,840 given opportunity. Oh I know. 364 00:19:03,840 --> 00:19:06,200 I remember when they say about Jeremy Corbyn lives in a 365 00:19:06,200 --> 00:19:08,640 £1,000,000 house in Islington. I thought, well, that must have 366 00:19:08,640 --> 00:19:12,400 been a small house. Also known as a house. 367 00:19:12,400 --> 00:19:13,880 In Israel, yes. Exactly. 368 00:19:13,880 --> 00:19:15,040 No, no, no. Absolutely right. 369 00:19:15,040 --> 00:19:16,200 Yeah. But the thing about the one 370 00:19:16,200 --> 00:19:18,320 about Medium as a message is a really interesting example, I 371 00:19:18,320 --> 00:19:22,040 think as as, as someone like as a politician who might be the 372 00:19:22,040 --> 00:19:24,720 subject of a story. One of the things that digital 373 00:19:24,720 --> 00:19:29,000 technology has also done that's kind of unremarked is that every 374 00:19:29,520 --> 00:19:32,920 article looks like the lead article. 375 00:19:33,240 --> 00:19:37,200 Like when you read a story on your phone, it's the biggest 376 00:19:37,200 --> 00:19:40,560 story with a single headline. It's like this is the main 377 00:19:40,560 --> 00:19:42,600 story. And we sometimes forget how the 378 00:19:42,600 --> 00:19:46,200 sort of pressure that even the politicians feel saying, Oh my 379 00:19:46,240 --> 00:19:48,480 God, they've written a whole article, whereas before it'd 380 00:19:48,480 --> 00:19:49,920 have been tucked away in Page seven. 381 00:19:49,920 --> 00:19:53,760 Small little thing wouldn't have gone you know, And so it's it's. 382 00:19:53,760 --> 00:19:56,360 Also a weird thing with advertising because everybody 383 00:19:56,360 --> 00:19:59,320 understood that the advertisements in the front page 384 00:19:59,320 --> 00:20:01,240 of the newspaper are of a different kind of the 385 00:20:01,240 --> 00:20:03,040 advertisements. In the back and the 386 00:20:03,040 --> 00:20:05,680 advertisements in the front of Vogue were a different kind of 387 00:20:05,680 --> 00:20:07,520 ad to the ad in the back of Vogue. 388 00:20:07,880 --> 00:20:12,680 You know, I was, as a joke tried to persuade someone in the media 389 00:20:12,840 --> 00:20:18,280 to try and buy a, an ad for adult incontinence products in 390 00:20:18,280 --> 00:20:21,560 the front half of Vogue just to see how they react. 391 00:20:21,720 --> 00:20:26,440 OK, my, my theory was, you know, you, you might offer $1,000,000 392 00:20:26,440 --> 00:20:28,200 and eventually would go to sign new house. 393 00:20:28,200 --> 00:20:29,960 You'd still say no. You see what I mean? 394 00:20:30,760 --> 00:20:33,360 But of course, every ad appears out of context. 395 00:20:33,360 --> 00:20:36,120 Everything is fundamentally decontextualized. 396 00:20:36,120 --> 00:20:38,160 Yeah, exactly. And So what you're reading at 397 00:20:38,160 --> 00:20:41,240 the moment is effectively front page news, even when it's 398 00:20:41,240 --> 00:20:42,600 completely irrelevant. Yeah, that's it. 399 00:20:42,600 --> 00:20:45,280 Everything is front page news when it's served up through a 400 00:20:45,280 --> 00:20:49,840 digital medium and on a screen. And it's also that this, this I 401 00:20:49,840 --> 00:20:53,760 think is having some really unexpected effects on, on 402 00:20:53,760 --> 00:20:58,600 people's views of the world. So the fragmentation of, of 403 00:20:58,600 --> 00:21:01,480 media, we all, we all know that and we all know that people can 404 00:21:01,480 --> 00:21:03,400 consume what they want. And, and some people would 405 00:21:03,680 --> 00:21:05,240 respond. Well, liberals always read the 406 00:21:05,240 --> 00:21:08,760 Guardian and conservatives always read the The Telegraph. 407 00:21:08,880 --> 00:21:11,120 But like you say, there was a sort of prioritisation. 408 00:21:11,120 --> 00:21:12,960 We'd also see lots of other things as well. 409 00:21:12,960 --> 00:21:17,000 What's happened now is that everyone can go an inch wide and 410 00:21:17,000 --> 00:21:20,840 a mile deep on every story. And most of the people that we 411 00:21:20,840 --> 00:21:24,200 accused of believing fake news, for example, you'll be in fake 412 00:21:24,200 --> 00:21:25,360 news. You just believe whatever you 413 00:21:25,360 --> 00:21:27,240 read on the Internet. And I've spent a lot of time 414 00:21:27,240 --> 00:21:30,240 with radical right wing and radical left wing movements. 415 00:21:30,840 --> 00:21:35,000 There news consumption is very, very different to maybe the 416 00:21:35,000 --> 00:21:37,680 mainstreams. They will become very obsessed 417 00:21:37,680 --> 00:21:40,240 with the story. They will read hundreds and 418 00:21:40,240 --> 00:21:45,160 hundreds of true articles about something they care about, like 419 00:21:45,160 --> 00:21:47,600 it'll be immigration. If it's a radical right 420 00:21:47,600 --> 00:21:53,560 movement, they'll read hundreds of articles, and they will, but 421 00:21:53,560 --> 00:21:56,920 they'll be mostly truthful. Last the most most accurate 422 00:21:56,920 --> 00:21:59,960 articles from often from quite respectable outlets. 423 00:22:00,680 --> 00:22:02,680 So when you say we're just reading fake news, they'll say 424 00:22:02,680 --> 00:22:05,080 no, no, no. These stories are printed on the 425 00:22:05,080 --> 00:22:07,760 BBC in the Guardian. Like, I'm not reading fake news. 426 00:22:07,760 --> 00:22:09,360 You're the idiot. You're the one that's just, you 427 00:22:09,360 --> 00:22:10,560 know, doesn't know what's going on. 428 00:22:10,880 --> 00:22:14,960 But they have created a version or a view of the world that has 429 00:22:14,960 --> 00:22:18,560 been constructed from a real deep dive into a single subject 430 00:22:18,560 --> 00:22:20,240 where they don't really read anything else at all. 431 00:22:20,600 --> 00:22:23,360 It's a misleading portrait of the world, but it's based on 432 00:22:23,360 --> 00:22:27,600 individually true stories, and it's a lot harder to Pierce into 433 00:22:27,600 --> 00:22:30,280 that world. And then they'll say like this 434 00:22:30,280 --> 00:22:33,320 isn't being covered enough by here, there, or someone else. 435 00:22:33,320 --> 00:22:34,400 I'll give you a lovely story of this. 436 00:22:34,440 --> 00:22:38,560 Robert Cialdini did the his point, which is most media bias, 437 00:22:38,640 --> 00:22:43,040 isn't misinformation, it's disproportionate presentation of 438 00:22:43,040 --> 00:22:44,520 information. Why aren't you covering this 439 00:22:44,520 --> 00:22:46,000 story? Why aren't you giving it more? 440 00:22:46,000 --> 00:22:48,040 Time or equally giving Ludicrous. 441 00:22:48,520 --> 00:22:50,400 Ludicrous. So there's a wonderful phrase, 442 00:22:50,600 --> 00:22:54,240 Citizen Kane. Where's where Kane says make the 443 00:22:54,240 --> 00:22:57,360 headline big enough, I'll make the story big enough that 444 00:22:57,360 --> 00:23:00,720 effectively you give prominence to something and you and you 445 00:23:00,720 --> 00:23:03,640 downplay something else. That's how most media bias 446 00:23:03,640 --> 00:23:05,520 happens. It's not actually misinformation 447 00:23:05,520 --> 00:23:08,280 or distorted reporting. And so these people aren't 448 00:23:08,280 --> 00:23:11,560 actually getting misinformation. They're getting what you might 449 00:23:11,560 --> 00:23:16,400 call, I suppose, I suppose, miscalibrated information. 450 00:23:16,400 --> 00:23:19,160 Yeah. Context completely matters and a 451 00:23:19,160 --> 00:23:21,880 lot of the time, so this is a word me and my Co founder say 452 00:23:21,880 --> 00:23:24,000 all the times that both things can be right. 453 00:23:24,240 --> 00:23:27,760 Yes, a lot of the times people will look at things from their 454 00:23:27,760 --> 00:23:30,400 view, and a lot of those times it's based on facts. 455 00:23:30,840 --> 00:23:33,000 It's just whether you're allowing people to zoom in on 456 00:23:33,000 --> 00:23:35,800 the full picture, which is why when you talked about the fact 457 00:23:35,800 --> 00:23:39,560 that video changed the law, that is 100% true. 458 00:23:39,560 --> 00:23:43,080 I would go even further to say that podcast with video changed 459 00:23:43,080 --> 00:23:47,880 the law one, because now people are taking the time to delve 460 00:23:47,880 --> 00:23:51,120 into the full context of a person, what they're talking 461 00:23:51,120 --> 00:23:53,560 about and what's happening. And in a world, I always talk 462 00:23:53,560 --> 00:23:57,520 about the fact that in a world where IQ is democratized because 463 00:23:57,520 --> 00:24:00,760 of AI and there's so much AI slob, there's so much news 464 00:24:01,120 --> 00:24:03,160 happening and you don't even know what to trust and what not 465 00:24:03,160 --> 00:24:05,440 to trust. Even some short form videos you 466 00:24:05,440 --> 00:24:09,240 can falsify to be real. I think that the, the, the 467 00:24:09,240 --> 00:24:13,240 notion of podcasting, for example, and the fact that for 468 00:24:13,400 --> 00:24:18,640 two hours, 3 hours, you can't really play a role for that 469 00:24:18,640 --> 00:24:20,400 long. If somebody's asking you like 470 00:24:20,400 --> 00:24:23,080 very direct questions, you get to understand the full picture 471 00:24:23,280 --> 00:24:25,040 of what somebody's all about that. 472 00:24:25,040 --> 00:24:28,240 The only problem I have with the world of podcasting right now is 473 00:24:28,240 --> 00:24:30,760 because now people want to do it as a living. 474 00:24:31,520 --> 00:24:35,280 Everyone's not trying to make money when sponsors get involved 475 00:24:35,280 --> 00:24:38,000 is like mismatching. There's like spray there 476 00:24:38,000 --> 00:24:40,080 everywhere. There's like somebody sponsoring 477 00:24:40,080 --> 00:24:42,320 with one brand and then they sponsor with another brand. 478 00:24:42,520 --> 00:24:46,240 And that's why I, I really like what brands like ear Max are 479 00:24:46,240 --> 00:24:48,480 doing when they're trying to make sure that the, the 480 00:24:48,480 --> 00:24:52,640 matchmaking between the right podcasts with the right ads and 481 00:24:52,640 --> 00:24:54,800 in the right way, in a way that actually gives value to the 482 00:24:54,800 --> 00:24:57,840 audience stays intact. And, and, and, and I just think 483 00:24:57,840 --> 00:24:59,400 that that's something that we should be thinking about. 484 00:24:59,400 --> 00:25:02,800 The potency of both the influence of marketing and 485 00:25:03,320 --> 00:25:07,240 podcast marketing is to do with the authenticity of the message 486 00:25:07,520 --> 00:25:09,480 and of course, the authenticity of the messenger. 487 00:25:10,200 --> 00:25:12,000 And I think you've got a massive multiplier effect. 488 00:25:12,000 --> 00:25:15,000 David Ogilvy wrote about this to do with testimonials and 489 00:25:15,000 --> 00:25:18,000 advertising. If the person delivering the 490 00:25:18,000 --> 00:25:21,760 message is credible and has significant reputational skin in 491 00:25:21,760 --> 00:25:25,560 the game, it's going to be inordinately more convincing 492 00:25:25,840 --> 00:25:27,960 than if it's obviously just tagged on. 493 00:25:28,120 --> 00:25:30,680 But then there's also the question of appropriateness of 494 00:25:30,680 --> 00:25:33,280 all audience as well. So you've got a double whammy. 495 00:25:33,280 --> 00:25:36,440 You can get exactly the if you use someone like Air Max, you 496 00:25:36,440 --> 00:25:40,080 get exactly the right audience. So you've solved your kind of 497 00:25:40,080 --> 00:25:43,160 targeting challenge, but you've also solved your creative 498 00:25:43,160 --> 00:25:46,160 challenge by having the right spokesperson to that audience 499 00:25:46,520 --> 00:25:52,720 who fundamentally, when they recommend something, can do so 500 00:25:52,840 --> 00:25:55,400 with genuine sincerity. I mean, I work with a company 501 00:25:55,400 --> 00:26:00,000 called Caggy, which is paid for search and they, partly 502 00:26:00,240 --> 00:26:03,600 encouraged by me, have a policy where unless the person 503 00:26:03,600 --> 00:26:08,480 recommending it uses the product themselves and entirely buys 504 00:26:08,480 --> 00:26:11,600 into the mission of what they're trying to do, it doesn't matter 505 00:26:11,600 --> 00:26:14,320 what the audience is, they're not interested in advertising 506 00:26:14,320 --> 00:26:16,920 there. And so that I think that double 507 00:26:16,920 --> 00:26:20,560 win where you get exactly the right audience, but also exactly 508 00:26:20,560 --> 00:26:25,360 the right degree of conviction and and authenticity in the 509 00:26:25,360 --> 00:26:27,800 messenger. When you get both those things 510 00:26:27,800 --> 00:26:30,640 aligned, that's effectively using the media to its full 511 00:26:30,640 --> 00:26:32,440 strength. And a lot of the time these 512 00:26:32,440 --> 00:26:34,280 things are bought as part of media deals. 513 00:26:34,280 --> 00:26:38,880 They're bought because out of expediency on basic demographics 514 00:26:38,880 --> 00:26:41,640 and nothing else. And that's actually, I mean, 515 00:26:41,640 --> 00:26:44,000 it's all right. I mean, it's, it's not a total 516 00:26:44,000 --> 00:26:46,880 waste of time, but you're missing the real magic of the 517 00:26:46,880 --> 00:26:48,800 medium itself when you do that, I think. 518 00:26:48,960 --> 00:26:52,640 Well, clip well, we sort of live in this weird clip economy world 519 00:26:52,640 --> 00:26:54,600 now, don't you? A lot of a lot of the the 520 00:26:54,600 --> 00:26:57,880 purpose of the podcast is actually the clip that either 521 00:26:57,880 --> 00:27:00,640 then goes viral and is seen by 5 million people, not the three 522 00:27:00,640 --> 00:27:03,640 hour version that you spent ages making, which itself is 523 00:27:03,640 --> 00:27:06,040 fascinating. But there's there's a couple of 524 00:27:06,040 --> 00:27:08,720 really interesting things about about I've just finished making 525 00:27:08,720 --> 00:27:11,040 a podcast for the BBC. It's about now called everything 526 00:27:11,040 --> 00:27:13,720 is fake and nobody cares. It's about modern fakery, the 527 00:27:13,720 --> 00:27:18,680 role of WWF wrestling in modern politics, how AI is changing 528 00:27:18,680 --> 00:27:21,240 that. And I spent a long time 529 00:27:21,680 --> 00:27:25,720 interviewing a very prominent COVID skeptic, vaccine skeptic. 530 00:27:26,400 --> 00:27:30,160 And something is interesting is happening about what podcast is 531 00:27:30,160 --> 00:27:34,560 doing to people's perception of editing in media when you make 532 00:27:34,560 --> 00:27:36,560 ABBC story. And I've made many of them. 533 00:27:37,200 --> 00:27:39,120 You are you. Let's say I make a one hour 534 00:27:39,120 --> 00:27:42,000 programme. I've probably got 10 to 15 hours 535 00:27:42,080 --> 00:27:46,360 at least of material and I have to make judgements about what to 536 00:27:46,360 --> 00:27:48,880 leave in and what to take out. For a long time people just 537 00:27:48,880 --> 00:27:50,520 trusted that the BBC would do that. 538 00:27:50,520 --> 00:27:54,280 Now, now that everyone can make content and they everyone edits, 539 00:27:54,280 --> 00:27:57,760 everyone's an editor and they can spend 3 hours watching 540 00:27:57,760 --> 00:28:02,560 someone unedited on a podcast. They've become extremely cynical 541 00:28:02,560 --> 00:28:06,040 by the active editing. They think editing by the BBC or 542 00:28:06,040 --> 00:28:09,800 whoever else is manipulation is control. 543 00:28:09,800 --> 00:28:13,680 I'd rather listen to that person for three straight hours talking 544 00:28:13,680 --> 00:28:18,240 on a podcast than your edited version of that person. 545 00:28:18,960 --> 00:28:22,080 But what they don't understand, of course, is that you can speak 546 00:28:22,080 --> 00:28:25,040 for three hours on a podcast and talk absolute rubbish 547 00:28:25,640 --> 00:28:28,840 unchallenged. My job when I edit is to try to 548 00:28:28,840 --> 00:28:31,880 give a really fair and full reflection with this within a 549 00:28:31,880 --> 00:28:35,320 short space of time. I have to do that and I think 550 00:28:35,560 --> 00:28:38,520 places like the BBC need to be much more transparent about how 551 00:28:38,520 --> 00:28:39,720 they edit. I think they're. 552 00:28:39,720 --> 00:28:41,680 Going to Trump edit. Oh, that was terrible. 553 00:28:41,680 --> 00:28:43,320 That was terrible. Oh, and I write, I write, I 554 00:28:43,360 --> 00:28:46,080 write about that in the in the podcast. 555 00:28:46,080 --> 00:28:49,040 I'm like, look, when the BBC makes mistakes like this, it's 556 00:28:49,040 --> 00:28:53,440 not surprising people are very sceptical about this generally. 557 00:28:53,440 --> 00:28:57,640 I mean, at least in that case, two of the biggest people at the 558 00:28:57,640 --> 00:29:00,960 at the court had to step down. I mean when, when, when a when a 559 00:29:01,760 --> 00:29:04,000 when a when a podcaster makes a similar mistake, they they 560 00:29:04,040 --> 00:29:06,720 usually get more famous. So that's the main difference. 561 00:29:06,960 --> 00:29:11,280 But but what so? But I do have some hope with 562 00:29:11,280 --> 00:29:15,560 podcasting. When social media first came 563 00:29:15,560 --> 00:29:19,160 out, I was among many of those who said the problem is not the 564 00:29:19,160 --> 00:29:22,080 messages that are people are receiving on social media, it is 565 00:29:22,080 --> 00:29:26,600 the fact it is a crunched down 32nd angry missive. 566 00:29:27,800 --> 00:29:29,680 It's not. People aren't engaging with 567 00:29:29,680 --> 00:29:32,280 material in any form of detail. Like I don't even care what the 568 00:29:32,280 --> 00:29:34,800 material is. We're turning into a yuck wow 569 00:29:34,800 --> 00:29:37,440 society where everything is really short and quick and not 570 00:29:37,440 --> 00:29:40,480 thought through. And now people seem to be 571 00:29:40,680 --> 00:29:45,520 thrilled to listen to Joe Rogan for three straight hours talking 572 00:29:45,520 --> 00:29:47,480 about some of the most detailed stories. 573 00:29:47,480 --> 00:29:49,560 Imagine that's amazing. Like whether you agree with 574 00:29:49,560 --> 00:29:53,720 Rogan or not, he's he and others like him are normalising the 575 00:29:53,720 --> 00:29:57,360 idea of sitting down for hours at a time engaging in a single 576 00:29:57,360 --> 00:29:59,720 idea. That is brilliant. 577 00:29:59,760 --> 00:30:02,640 But I am really excited about that and it's but people and 578 00:30:02,640 --> 00:30:05,200 people complain about broken in the other way. 579 00:30:05,720 --> 00:30:09,240 By the way, no one, least of all me, would have predicted this. 580 00:30:09,720 --> 00:30:10,640 It's. Amazing, isn't it? 581 00:30:10,640 --> 00:30:13,160 So the thing I was saying, I was saying the other day is if you 582 00:30:13,160 --> 00:30:16,800 told me in, say, 1990, OK, you're going to be driving to 583 00:30:16,800 --> 00:30:20,760 Manchester, in your car, you'll have the technology to play any 584 00:30:20,760 --> 00:30:24,160 music ever composed by the hand of man, OK? 585 00:30:24,760 --> 00:30:27,960 And you'll choose to spend the journey listening to two people 586 00:30:27,960 --> 00:30:29,320 talking. I would have said Nah. 587 00:30:29,720 --> 00:30:31,280 I don't think so. No, I know, right. 588 00:30:31,720 --> 00:30:34,560 And so the popularity of long form conversation, by the way, I 589 00:30:34,560 --> 00:30:36,200 mean, there's an interesting. One wouldn't have said it in 590 00:30:36,200 --> 00:30:40,680 2015, let alone 1990. Now, interesting, if you go back 591 00:30:40,680 --> 00:30:46,080 to my very early childhood, you did have on BBC One or two, I'm 592 00:30:46,320 --> 00:30:48,600 just trying to remember that there were various people who 593 00:30:48,600 --> 00:30:51,480 did very serious long conversations with sort of Carl, 594 00:30:51,480 --> 00:30:53,320 you know, Jung and people like that. 595 00:30:53,880 --> 00:30:57,680 And one of my theories was that television people fell out of 596 00:30:57,680 --> 00:31:01,240 love with conversation, but the public didn't. 597 00:31:01,720 --> 00:31:03,800 That for whatever reason, it wasn't a very fashionable kind 598 00:31:03,800 --> 00:31:04,800 of. Do you remember a thing called 599 00:31:04,840 --> 00:31:08,200 After Dark on Channel 4 Friday evenings? 600 00:31:08,200 --> 00:31:11,760 They got a load of people and got them pissed, OK. 601 00:31:12,240 --> 00:31:14,360 And it was an extraordinary thing where it was basically 602 00:31:14,360 --> 00:31:16,920 used A room. You got some extraordinary 603 00:31:16,920 --> 00:31:20,440 guests like Peter Sutcliffe, Dad, you know, and there's a 604 00:31:20,440 --> 00:31:23,160 famous one, famous one with Peter O'Toole, I think, where he 605 00:31:23,160 --> 00:31:25,880 basically sort of turns up with a stack of booze and it all 606 00:31:25,880 --> 00:31:28,960 turns nasty. But it was completely live 607 00:31:28,960 --> 00:31:31,520 conversation. It was parodied once as after 608 00:31:31,520 --> 00:31:33,760 closing time as opposed to after dark. 609 00:31:34,160 --> 00:31:36,360 But it was very, very captivating viewing. 610 00:31:36,440 --> 00:31:40,440 And it was kind of early stage, you know, large scale podcast in 611 00:31:40,440 --> 00:31:42,280 a sense. And it was always very, very 612 00:31:42,280 --> 00:31:44,960 popular, you know, particularly given the slot. 613 00:31:45,560 --> 00:31:49,320 And I always wondered whether television people are obsessed 614 00:31:49,320 --> 00:31:52,040 with the sound bite and they turn interviews into these 615 00:31:52,040 --> 00:31:55,240 highly combative, very short form things. 616 00:31:56,120 --> 00:31:58,720 And actually, maybe the public doesn't want that. 617 00:31:59,240 --> 00:32:02,240 I think it's, I just think it's really good for people to try to 618 00:32:02,240 --> 00:32:04,600 separate again the medium and the message. 619 00:32:05,400 --> 00:32:08,200 If the message you might not agree with on Stephen Bartlett's 620 00:32:08,200 --> 00:32:12,120 podcast or Jay Shetty's podcast or Joe Rogan's podcast or the 621 00:32:12,120 --> 00:32:14,480 rest is politics or the rest is history or whatever. 622 00:32:14,680 --> 00:32:19,560 But but the fact that people there's this unending appetite 623 00:32:19,560 --> 00:32:23,360 to listen to people talk for hours about detailed subjects 624 00:32:23,640 --> 00:32:26,840 is, is a really positive development and the opposite of 625 00:32:26,840 --> 00:32:28,920 where I thought we were heading 10 years ago. 626 00:32:29,960 --> 00:32:32,640 Yeah, which the best. So look, we're doing it right 627 00:32:32,640 --> 00:32:34,240 now. The best thing a futurologist 628 00:32:34,240 --> 00:32:35,920 ever told me is there are no trends. 629 00:32:35,920 --> 00:32:38,880 There are vectors and there's always a counter trend. 630 00:32:39,120 --> 00:32:42,160 And just as in a sense, conventional media didn't do a 631 00:32:42,160 --> 00:32:46,200 very good job of providing us with two-minute video, did a 632 00:32:46,200 --> 00:32:48,800 fantastic job of providing us with video which was half an 633 00:32:48,800 --> 00:32:52,120 hour or an hour long. It also didn't do a very good 634 00:32:52,120 --> 00:32:54,840 job of providing us with much longer form content. 635 00:32:55,280 --> 00:32:58,200 And so we've suddenly effectively discovered an 636 00:32:58,200 --> 00:33:01,720 appetite for the two extremes. So it's a, it's a vector really. 637 00:33:01,720 --> 00:33:03,280 It's not, it's not really a trend. 638 00:33:03,280 --> 00:33:07,080 And the same people who are watching, you know, 32nd cat 639 00:33:07,080 --> 00:33:10,080 videos are also watching 2 hour content. 640 00:33:10,160 --> 00:33:10,800 Yeah. Yeah. 641 00:33:10,800 --> 00:33:12,040 It's. Amazing, yeah. 642 00:33:12,680 --> 00:33:16,080 Because what you were seeing, there's like words that came 643 00:33:16,080 --> 00:33:19,920 about since COVID, right? Like words like gas lighting and 644 00:33:19,920 --> 00:33:22,240 toxic. And there's all these words that 645 00:33:22,240 --> 00:33:27,920 you see on like 22nd Reels and Tik Toks that people learn 646 00:33:27,920 --> 00:33:30,880 about, but they don't really get into the depth of what it 647 00:33:30,880 --> 00:33:34,320 actually means, right? But I'm speaking to a lot of 648 00:33:34,320 --> 00:33:37,040 people right now. I can see the TikTok, what the 649 00:33:37,040 --> 00:33:39,880 TikTok algorithm feels like. Because when you press them on 650 00:33:39,880 --> 00:33:43,640 the questions and you press them on that POV, a lot of the time 651 00:33:43,640 --> 00:33:46,200 it's very baseless, right? It's based on the trending 652 00:33:46,200 --> 00:33:47,560 conversation in a word right now. 653 00:33:47,840 --> 00:33:49,400 But I think I like about podcasting. 654 00:33:49,400 --> 00:33:52,440 It's kind of like books because I don't read books, right? 655 00:33:52,680 --> 00:33:57,240 I don't have the, I literally the idea of sitting down and 656 00:33:57,240 --> 00:34:01,000 just doing one thing. My brain just goes audiobooks, 657 00:34:01,240 --> 00:34:05,400 audiobooks, podcast. There's a podcast called 658 00:34:05,400 --> 00:34:07,600 Founders of David Senra. I talk about it all the time. 659 00:34:07,880 --> 00:34:12,400 He deep slides into one book for an hour and gives you basically 660 00:34:12,400 --> 00:34:14,800 like the Cliff notes and all the conversations around it. 661 00:34:15,040 --> 00:34:16,960 I learned so much about that, right? 662 00:34:16,960 --> 00:34:20,639 I can listen to a 2 hour podcast and retain everything, right? 663 00:34:20,920 --> 00:34:22,800 And because I'm just an auditory learner, right? 664 00:34:22,840 --> 00:34:25,320 And I think that the the good thing I like about where 665 00:34:25,320 --> 00:34:28,880 podcasting is gone, it's allowed you to learn from an auditory 666 00:34:28,880 --> 00:34:32,880 sense, visual kinesthetic because you're feeling people's 667 00:34:32,880 --> 00:34:34,760 emotions and what they think about things. 668 00:34:34,920 --> 00:34:38,800 But then also digital because there's like points and notes of 669 00:34:38,800 --> 00:34:41,679 data and it does the full circle of learning, whereas a lot of 670 00:34:41,679 --> 00:34:43,719 other things just focus on one type of learning. 671 00:34:44,120 --> 00:34:46,480 And that's the thing I really like about where media is going, 672 00:34:46,480 --> 00:34:49,520 which is why I actually believe that providing immersive 673 00:34:49,520 --> 00:34:54,159 experiences with AI and virtual reality and where that's going, 674 00:34:54,280 --> 00:34:56,560 that's where we should be spending our time, like 675 00:34:56,560 --> 00:34:59,680 investing in innovating in. But really what I'm seeing right 676 00:34:59,680 --> 00:35:02,720 now with like every time I go on an app and I see this company 677 00:35:02,720 --> 00:35:05,040 has raised this amount of money for this, it's just for more 678 00:35:05,040 --> 00:35:08,840 bullshit that just tries to leverage people's attention all 679 00:35:08,840 --> 00:35:10,840 over and over again. No one's actually using the 680 00:35:10,840 --> 00:35:13,720 modern technology to build stuff that are useful. 681 00:35:13,920 --> 00:35:17,320 But like, you go to the US, for example, and I say I was in San 682 00:35:17,320 --> 00:35:19,520 Francisco, right? And I'm like, this is supposed 683 00:35:19,520 --> 00:35:21,120 to be the tech hub of the world, right? 684 00:35:21,520 --> 00:35:24,960 What? But why is it that my phone is 685 00:35:24,960 --> 00:35:27,960 advanced, but when I go out into the real world, like the 686 00:35:27,960 --> 00:35:32,080 computer's not working? I can't, like, check into my 687 00:35:32,080 --> 00:35:34,440 plane without anything bugging. It's almost like we left the 688 00:35:34,440 --> 00:35:38,120 real world to fix this virtual world and no one's. 689 00:35:38,160 --> 00:35:40,720 Is arguably one of the richest cities in the world, but they've 690 00:35:40,720 --> 00:35:44,040 got a massive homeless problem. And you know, you know that 691 00:35:44,080 --> 00:35:46,520 actually the infrastructure of the place is a fucking shit 692 00:35:46,520 --> 00:35:47,360 show. Yeah. 693 00:35:47,440 --> 00:35:50,920 I think this point though, is really, I mean, I think the 694 00:35:51,360 --> 00:35:56,400 there is a really profound point about the difference between 695 00:35:56,400 --> 00:36:03,080 technology being created and technology being applied in the 696 00:36:03,080 --> 00:36:04,200 real world. Adopted. 697 00:36:04,200 --> 00:36:11,800 And adopted like, and so when you, you various studies looking 698 00:36:11,800 --> 00:36:15,400 at companies that have adopted large language models and what's 699 00:36:15,400 --> 00:36:19,000 so interesting about that at the moment is that pretty strong 700 00:36:19,000 --> 00:36:23,880 evidence that at the individual level, people in companies using 701 00:36:23,880 --> 00:36:27,600 language models are reporting significant increases in 702 00:36:27,600 --> 00:36:30,000 productivity. I can now write these reports 703 00:36:30,000 --> 00:36:31,720 quicker. I can summarize these documents 704 00:36:31,720 --> 00:36:33,480 better. I can present power, I can do 705 00:36:33,480 --> 00:36:36,080 PowerPoint slides faster, I can run these accounts faster. 706 00:36:36,720 --> 00:36:40,480 But at the company wide level, it's not actually turning into 707 00:36:40,920 --> 00:36:44,680 changes to profit and loss. And it feels a little bit like 708 00:36:44,680 --> 00:36:48,080 the productivity paradox from decades ago where the world 709 00:36:48,080 --> 00:36:52,080 went. Computers were adopted across 710 00:36:52,080 --> 00:36:55,080 the world very, very quickly in the 80s and 90s, but there was 711 00:36:55,080 --> 00:36:58,600 no productivity gains for like 10 or 15 years after that. 712 00:36:58,880 --> 00:37:02,680 And I think it's because we forget things like the printers 713 00:37:02,680 --> 00:37:05,560 jammed. It's good like oh I've yes, I've 714 00:37:05,560 --> 00:37:09,120 just saved an hour doing my PowerPoint slide using ChatGPT. 715 00:37:09,400 --> 00:37:12,960 Now I've got to attend a 2 hour meeting about AI compliance 716 00:37:12,960 --> 00:37:15,320 because apparently I wasn't allowed to do it that way. 717 00:37:15,920 --> 00:37:19,080 And it's this like writ large with AI, I mean? 718 00:37:19,080 --> 00:37:21,440 It's perfectly possible. I mean, if you think about it, 719 00:37:21,440 --> 00:37:26,840 one of the strangest things which persisted in business for 720 00:37:26,840 --> 00:37:30,080 about 100 years was the typewriter. 721 00:37:30,960 --> 00:37:35,440 Because traditionally, I mean, in 1970, Ogilvy had a typing 722 00:37:35,440 --> 00:37:37,400 pool. You didn't have anything on your 723 00:37:37,400 --> 00:37:38,600 desk. You didn't have a typewriter on 724 00:37:38,600 --> 00:37:41,120 your desk. You wrote a letter to a client 725 00:37:41,120 --> 00:37:43,400 by hand. You submitted it to the typing 726 00:37:43,400 --> 00:37:46,400 pool, who then typed it up on official paper and it was sent 727 00:37:46,680 --> 00:37:51,040 OK. Now, unless you had appalling 728 00:37:51,040 --> 00:37:55,240 handwriting, the typewriter was a massive cost imposed on 729 00:37:55,240 --> 00:37:58,000 business simply so that the business could appear 730 00:37:58,000 --> 00:38:00,760 professional. You know, I mean, in the 18th 731 00:38:00,760 --> 00:38:02,640 century, everything was done by hand. 732 00:38:02,840 --> 00:38:06,360 No one thought it strange. You wrote a letter, you sent it. 733 00:38:06,400 --> 00:38:09,080 That was it. You added this completely 734 00:38:09,080 --> 00:38:13,360 unnecessary extra process to everything that was done, 735 00:38:13,680 --> 00:38:15,200 really, because it was expected of you. 736 00:38:15,200 --> 00:38:17,320 Now, it strikes me as perfectly possible. 737 00:38:17,320 --> 00:38:19,240 Now, the word processor does bring benefits. 738 00:38:19,240 --> 00:38:21,640 I think we can say, you know, it's better than a typewriter. 739 00:38:21,640 --> 00:38:25,640 You can edit, you know, you can, you know, take existing copy and 740 00:38:25,640 --> 00:38:28,280 you can edit it to save retyping everything. 741 00:38:28,440 --> 00:38:30,320 You don't need carbon paper anymore. 742 00:38:30,320 --> 00:38:33,880 There are benefits. Perfectly possible. e-mail, for 743 00:38:33,880 --> 00:38:38,560 example, as a universally adopted behaviour has net 744 00:38:38,600 --> 00:38:40,920 brought mass productivity losses. 745 00:38:40,920 --> 00:38:44,760 Yeah, perfectly possible that it's that it's now what's weird 746 00:38:44,760 --> 00:38:48,320 about it is that everybody got obsessed with measuring the 747 00:38:48,320 --> 00:38:51,520 productivity gains and losses of say working remotely or working 748 00:38:51,520 --> 00:38:55,400 from home to anal degrees of well, you know, apparently 749 00:38:55,400 --> 00:38:57,600 there's a 3% loss in productivity. 750 00:38:57,960 --> 00:39:01,120 But we've introduced wholesale things like open plan offices 751 00:39:01,120 --> 00:39:05,120 and e-mail without actually considering what the alternative 752 00:39:05,120 --> 00:39:08,120 might be. Or do we, I mean, in the, you 753 00:39:08,120 --> 00:39:13,000 know, in, in the old days there was well, first of all, first of 754 00:39:13,000 --> 00:39:17,840 all, actually the existence of a typing pool created a very 755 00:39:17,840 --> 00:39:20,720 interesting filtration mechanism because senior people got things 756 00:39:20,720 --> 00:39:24,480 typed immediately and junior people had to beg favours to get 757 00:39:24,480 --> 00:39:27,800 it done. So you had actually a sort of 758 00:39:27,800 --> 00:39:30,640 prioritization system, which is if you got something on your 759 00:39:30,640 --> 00:39:33,880 desk, it was more likely to be an important thing from a senior 760 00:39:33,880 --> 00:39:36,840 person that an irrelevant bit of bullshit from, you know, that 761 00:39:36,840 --> 00:39:39,040 you'd been CC D on for no readily apparent reason. 762 00:39:39,040 --> 00:39:40,840 Like our newspapers prioritizing. 763 00:39:41,560 --> 00:39:43,120 It was. It was inherently, you know, 764 00:39:43,120 --> 00:39:45,320 there was a sense of importance to the order. 765 00:39:45,600 --> 00:39:46,920 Yeah. And there was a kind of order to 766 00:39:46,920 --> 00:39:50,120 it. And so quite often we destroy 767 00:39:50,120 --> 00:39:53,240 these old technologies and we get wildly excited by what the 768 00:39:53,240 --> 00:39:57,440 new technology makes possible and we completely forget in the 769 00:39:57,440 --> 00:40:01,080 process what were the invisible benefits of what preceded it 770 00:40:01,080 --> 00:40:03,520 100%. So, you know, when it took a bit 771 00:40:03,520 --> 00:40:07,080 of effort to produce something in print, the stuff that landed 772 00:40:07,080 --> 00:40:09,840 on your desk was almost invariably important and had to 773 00:40:09,840 --> 00:40:13,800 be dealt with. OK, now e-mail you have to sift 774 00:40:13,800 --> 00:40:16,200 between the window cleaners are in tomorrow. 775 00:40:16,200 --> 00:40:20,160 Please move everything off the off the window ledge and you 776 00:40:20,160 --> 00:40:22,920 know you've been fired. OK, It's your. 777 00:40:22,920 --> 00:40:25,920 Suddenly the burden falls on the recipient to the priority. 778 00:40:26,400 --> 00:40:28,480 Well, I think that, I mean, I've, I've, I coined this phrase 779 00:40:28,480 --> 00:40:30,640 techno admin, which is essentially what a lot of 780 00:40:30,640 --> 00:40:33,520 companies have done is to put the burden of annoying 781 00:40:33,520 --> 00:40:36,920 administration on the user. Like the customer. 782 00:40:36,920 --> 00:40:38,000 The customer, that's what I mean. 783 00:40:38,200 --> 00:40:41,160 Yeah, the user, the customer. Phrase yesterday it's not it's 784 00:40:41,200 --> 00:40:45,080 not cost saving if you're making your customer pay the price I. 785 00:40:45,240 --> 00:40:47,520 Mean the number of times where you like this. 786 00:40:47,520 --> 00:40:50,200 This has happened to me and it's happened to many people I know. 787 00:40:50,200 --> 00:40:54,280 Some automated system inside the energy company sends you a dodgy 788 00:40:54,280 --> 00:40:56,480 bill. They've miscalculated something 789 00:40:56,760 --> 00:41:01,560 and it is on you to spend hours and hours of unpaid labour 790 00:41:01,840 --> 00:41:04,800 trying to resolve this problem. Now the company has saved some 791 00:41:04,800 --> 00:41:06,600 money because they churn out automated. 792 00:41:06,600 --> 00:41:08,640 Bills. But what they've actually done 793 00:41:08,720 --> 00:41:11,840 is outsource labour to you, the poor soul who has to spend, 794 00:41:12,040 --> 00:41:15,320 usually during work hours, phoning them up, desperately 795 00:41:15,320 --> 00:41:16,640 trying to get something resolved. 796 00:41:16,880 --> 00:41:20,760 This is happening. The productivity cost to the UK 797 00:41:20,760 --> 00:41:24,360 of techno admin is absolutely vast and I sometimes think in 798 00:41:24,360 --> 00:41:26,800 addition to not understanding. The loss of just discretionary 799 00:41:26,800 --> 00:41:28,800 time. And the stress, why? 800 00:41:28,800 --> 00:41:32,680 I mean people have a £3000 bills turn up at their door and 801 00:41:32,680 --> 00:41:34,360 they've got no idea what it is. They don't know what they 802 00:41:34,360 --> 00:41:37,000 supposed to do with it. It's terrifying for people. 803 00:41:37,440 --> 00:41:41,720 So the but so in addition to the not seeing the benefits of old 804 00:41:41,720 --> 00:41:45,560 systems, I think we're often blind to the the costs that new 805 00:41:45,560 --> 00:41:49,360 systems bring, the extra friction that they create. 806 00:41:49,720 --> 00:41:54,000 So what I think about like every time a company decides to 807 00:41:54,000 --> 00:41:57,960 introduce a new IT system, they don't really factor in the fact 808 00:41:58,200 --> 00:42:01,480 that ordinary people that, you know, the user has got to work 809 00:42:01,480 --> 00:42:04,240 out a new thing and the laptop hasn't really got to download 810 00:42:04,240 --> 00:42:05,960 some new software. They're stressed about that 811 00:42:05,960 --> 00:42:08,960 because I can't really remember. And actually now it's crammed. 812 00:42:08,960 --> 00:42:11,480 The whole system's crashed. The files aren't connecting up 813 00:42:11,480 --> 00:42:13,760 anymore. That is none of that's accounted 814 00:42:13,760 --> 00:42:15,400 for. It's always assumed that 815 00:42:15,400 --> 00:42:18,920 everything runs really, really smoothly and it never ever does 816 00:42:18,920 --> 00:42:20,520 any like. It's just not the case. 817 00:42:20,520 --> 00:42:21,840 I'll. Give you a beautiful example. 818 00:42:22,160 --> 00:42:26,840 So there was an IT department which came up with a very, very 819 00:42:26,840 --> 00:42:32,560 efficient system for effectively conveyancing for a large 820 00:42:32,560 --> 00:42:35,560 building society. And they pointed out the 821 00:42:35,560 --> 00:42:37,680 extraordinary benefits this would bring. 822 00:42:38,520 --> 00:42:42,160 And someone very cynical in the board meeting said presumably it 823 00:42:42,160 --> 00:42:45,600 only brings these benefits if the entire exchange of 824 00:42:45,600 --> 00:42:48,000 properties goes completely according to plan. 825 00:42:48,200 --> 00:42:52,360 And they said, yes, that's true. OK, So OK, can you go and find 826 00:42:52,360 --> 00:42:56,280 out what percentage of property transactions go according to 827 00:42:56,280 --> 00:42:58,480 plan? Turned out 14%. 828 00:42:58,880 --> 00:43:01,600 I'm surprised. So for the so for the other 86% 829 00:43:01,600 --> 00:43:04,240 of people, what you were doing was introducing them to a total 830 00:43:04,240 --> 00:43:08,320 nightmare or on pass where the technology was completely unable 831 00:43:08,320 --> 00:43:11,560 to contrive a Plan B. It's happening all the time and. 832 00:43:11,880 --> 00:43:15,000 My other My other complaint is that technology arrives as an 833 00:43:15,000 --> 00:43:17,400 option and ends up as an obligation. 834 00:43:17,920 --> 00:43:20,560 So a classic example of that. You could include parking apps 835 00:43:20,560 --> 00:43:22,960 in that wonderful option. Fantastic. 836 00:43:22,960 --> 00:43:25,880 OK, then what happens is some finance person? 837 00:43:25,880 --> 00:43:27,640 Is going to be. I wrote and asked about parking 838 00:43:27,640 --> 00:43:29,480 apps the other day. They do my head in. 839 00:43:29,480 --> 00:43:31,360 No, no, I don't know. There's an option. 840 00:43:31,440 --> 00:43:33,240 Great. Because you know, Seven Oaks 841 00:43:33,240 --> 00:43:37,920 station, you at one point you needed 7 LB coins in order to 842 00:43:37,920 --> 00:43:40,920 pay the part there, which meant you walked around basically, you 843 00:43:40,920 --> 00:43:44,080 know, with your testicles being irradiated by anyway. 844 00:43:44,920 --> 00:43:47,200 But but quite quite useful as an option. 845 00:43:47,200 --> 00:43:50,880 What then what happens is the finance person then decides that 846 00:43:50,880 --> 00:43:53,640 because it's much cheaper to do this by app than it is by 847 00:43:54,000 --> 00:43:58,960 coinage or by any other means. OK, the the app is then imposed 848 00:43:58,960 --> 00:44:02,120 and there's no alternative part about ticket machines, railway 849 00:44:02,120 --> 00:44:06,800 stations as an alternative to the person behind the, the, the 850 00:44:07,840 --> 00:44:10,840 what you call a ticket, Guiche, whatever, whatever you call it 851 00:44:11,160 --> 00:44:15,560 OK, as an alternative, as a compliment to the person who 852 00:44:15,560 --> 00:44:17,920 will sell you a ticket and give you some information about train 853 00:44:17,920 --> 00:44:19,520 times. And actually you ought to go 854 00:44:19,520 --> 00:44:21,560 after 9:30 because it's much cheaper. 855 00:44:21,880 --> 00:44:23,440 Brilliant. If you know what ticket you want 856 00:44:23,440 --> 00:44:26,800 precisely and there's a queue at the bar at the, at the ticket 857 00:44:26,800 --> 00:44:30,240 term counter fab. OK, then what happens is you get 858 00:44:30,240 --> 00:44:33,120 rid of the human and it happens every time, which is something 859 00:44:33,120 --> 00:44:38,240 arrives as a techno alternative to something self checkout at 860 00:44:38,240 --> 00:44:40,600 supermarkets. And you go, actually this is 861 00:44:40,600 --> 00:44:43,360 pretty welcome because I can either do A or I can do B. 862 00:44:43,360 --> 00:44:46,240 And sometimes depending on context, I want to do one or the 863 00:44:46,240 --> 00:44:48,720 other. And then what happens is the, 864 00:44:48,800 --> 00:44:52,480 you know, the actual people operating humans operating 865 00:44:52,480 --> 00:44:57,600 checkouts disappear. And it's worse still because the 866 00:44:57,600 --> 00:45:01,360 people, the finance people and the IT people who impose the new 867 00:45:01,360 --> 00:45:05,520 technology take the credit for the reduction in costs, but they 868 00:45:05,520 --> 00:45:08,520 aren't held accountable for the blame in the lost opportunity. 869 00:45:08,520 --> 00:45:11,520 So for example, if you make everybody self checkout in a 870 00:45:11,520 --> 00:45:15,320 supermarket, no one can do a big weekly shop anymore because it's 871 00:45:15,320 --> 00:45:17,800 just too tedious. But they're not, they're not 872 00:45:17,800 --> 00:45:20,040 measuring that, they're just saying how wonderful this is 873 00:45:20,040 --> 00:45:22,080 because look at our reduction in labour costs. 874 00:45:22,080 --> 00:45:23,200 Exactly. They're outs and they're 875 00:45:23,200 --> 00:45:25,200 outsourcing a a lot. And this is the thing with the 876 00:45:25,200 --> 00:45:27,720 parking apps. They're outsourcing that labour 877 00:45:27,720 --> 00:45:30,480 that might have been done by a a a human going around checking 878 00:45:30,480 --> 00:45:33,320 the things and and emptying the coins from the machine. 879 00:45:33,320 --> 00:45:34,840 They weren't nicking a lot of the money. 880 00:45:36,040 --> 00:45:38,640 OK, so one of the reasons. There's loads of fraud on 881 00:45:38,640 --> 00:45:40,240 parking apps, I'm sure they're asked. 882 00:45:40,240 --> 00:45:43,320 About one of the reasons parking apps got introduced so quickly 883 00:45:43,320 --> 00:45:47,000 is they did a trial in four streets and they realized 884 00:45:47,000 --> 00:45:49,520 effectively that the people who collected money from meters had 885 00:45:49,520 --> 00:45:52,000 worked out a very clever way of nicking quite a lot of the money 886 00:45:52,000 --> 00:45:53,480 from the meters. But. 887 00:45:53,480 --> 00:45:56,120 But it's, it's a very, very easy thing to probably people have 888 00:45:56,120 --> 00:45:59,040 people have been going around putting up sort of fake, but 889 00:45:59,040 --> 00:46:02,800 like the QR code that you've got to scan people put physical fake 890 00:46:02,800 --> 00:46:04,800 QR codes on top of the real ones. 891 00:46:04,800 --> 00:46:08,560 And so you're directed to a scammy site where they take and 892 00:46:08,560 --> 00:46:09,920 it's again, it is on you. Oh. 893 00:46:10,360 --> 00:46:13,120 By the way, then you get a ticket and it's on the 894 00:46:13,200 --> 00:46:14,280 obligation. Is this is it? 895 00:46:14,360 --> 00:46:17,720 This is, This is the Techno admin problem that like the, the 896 00:46:17,760 --> 00:46:22,080 outsourcing of work to the customer or to the user by the 897 00:46:22,080 --> 00:46:26,000 introduction of technology is such a sort of common thread. 898 00:46:26,360 --> 00:46:29,560 And I, I agree with you, Rory, it's, it's the, it's the, it's 899 00:46:29,560 --> 00:46:33,960 the Dis, it's the closing off of alternatives gradually that is 900 00:46:33,960 --> 00:46:36,720 so frustrating. So I can imagine it happening, 901 00:46:36,720 --> 00:46:38,600 obviously with self driving cars. 902 00:46:38,760 --> 00:46:42,800 You know, at some point more and more self driving cars are the, 903 00:46:43,040 --> 00:46:46,760 the, the is the norm. And then at some point, actually 904 00:46:46,880 --> 00:46:51,240 we don't really need traffic lights anymore because why would 905 00:46:51,240 --> 00:46:52,880 you need them? And then what we need a self 906 00:46:52,880 --> 00:46:54,920 driving car in that case, because there are no traffic 907 00:46:54,920 --> 00:46:57,080 lights anymore, because most people don't need them. 908 00:46:57,080 --> 00:46:59,320 Do you want to buy a map? Well, there's no point buying a 909 00:46:59,320 --> 00:47:01,800 map. John Goodwin made this point 910 00:47:01,800 --> 00:47:04,440 last week, which is once a certain percentage of cars on 911 00:47:04,440 --> 00:47:08,120 the road become self driving, the roads become unusable to 912 00:47:08,160 --> 00:47:10,840 anybody who wants to drive. And it becomes safer if ever, if 913 00:47:10,840 --> 00:47:14,080 they're all self driving a part of a large network until they're 914 00:47:14,080 --> 00:47:17,040 until they're all hacked. But but what that then means, 915 00:47:17,040 --> 00:47:20,040 yeah, that but that then that that forecloses all other 916 00:47:20,040 --> 00:47:23,240 possibilities of of options and ways of driving. 917 00:47:23,240 --> 00:47:28,480 And driving will probably just become a recreational habit that 918 00:47:28,480 --> 00:47:31,280 you have to do in some kind of, you know, off grid settings. 919 00:47:31,360 --> 00:47:35,040 It's a bit like steam railways. In other words, you know, you 920 00:47:35,040 --> 00:47:38,240 steam railway lines survive as a kind of weird novelty. 921 00:47:38,240 --> 00:47:40,120 Yeah, exactly. And so you'll go to a strange 922 00:47:40,120 --> 00:47:42,240 site where you can practice. Drive around in circles. 923 00:47:42,240 --> 00:47:44,600 Driving yourself in a petrol car? 924 00:47:45,040 --> 00:47:48,480 The interesting thing I think, I think now is the extent to which 925 00:47:48,840 --> 00:47:53,240 artificial intelligence is going to do that and how, what will 926 00:47:53,240 --> 00:47:58,040 this transformative technology play like if we we understand 927 00:47:58,040 --> 00:48:01,000 this is the trend of how technology is introduced on 928 00:48:01,000 --> 00:48:05,120 people, people by the cost of it, other alternatives are 929 00:48:05,120 --> 00:48:08,520 gradually foreclosed and then you're left with this. 930 00:48:09,320 --> 00:48:11,840 Is that going to be the same here and if so, how? 931 00:48:11,880 --> 00:48:12,960 I don't know the answer by the way. 932 00:48:13,000 --> 00:48:15,080 I think that is. I think, by the way, that's a 933 00:48:15,080 --> 00:48:16,840 wonderful, wonderful point because I don't know. 934 00:48:17,040 --> 00:48:20,080 I mean, it does occur to me that there's an answer where a lot of 935 00:48:20,080 --> 00:48:25,680 people use this a lot, but it has almost no bearing or there's 936 00:48:26,080 --> 00:48:29,400 a perfect case where you develop a trillion dollar technology 937 00:48:29,400 --> 00:48:32,640 which is used effectively for the equivalent of horoscopes. 938 00:48:33,400 --> 00:48:35,440 You know, people, in other words, making a decision will 939 00:48:35,440 --> 00:48:38,520 just ask the AI in the same way that you consult the delfied 940 00:48:38,520 --> 00:48:42,560 Oracle just to arbitrarily decide between two alternatives. 941 00:48:43,120 --> 00:48:46,960 They developed Alexa and I, you know that that was a billion, a 942 00:48:46,960 --> 00:48:49,200 few billion to develop Alexa. OK. 943 00:48:49,720 --> 00:48:52,680 And if you actually look at it, it's basically what's the 944 00:48:52,680 --> 00:48:54,000 weather like and what time is it? 945 00:48:54,000 --> 00:48:58,000 And what we discovered from Alexa is that people really like 946 00:48:58,000 --> 00:49:01,000 first thing in the morning to know what time it is without 947 00:49:01,000 --> 00:49:03,880 having to open their eyes. That's a really interesting 948 00:49:03,880 --> 00:49:06,840 human discovery because once you open your eyes to look at a 949 00:49:06,840 --> 00:49:09,080 clock, you're woken up and you can't get back to sleep. 950 00:49:10,680 --> 00:49:13,760 But if I look at my electric queries, the the the intended 951 00:49:13,760 --> 00:49:16,440 purpose was we'll all shop that way. 952 00:49:16,840 --> 00:49:19,880 Or at least Amazon, I think was terrified that someone else 953 00:49:19,880 --> 00:49:24,760 would perfect spoken word shopping and therefore have to 954 00:49:24,760 --> 00:49:27,600 create this thing as a defensive kind of note or preemptive 955 00:49:27,600 --> 00:49:30,320 strike if you like. But if you look at what we use 956 00:49:30,320 --> 00:49:34,440 Alexa for, I mean, 95% of my uses, to be honest, for the cost 957 00:49:34,440 --> 00:49:37,680 of developing Alexa, you could have given everybody on the 958 00:49:37,680 --> 00:49:43,120 planet a thermometer and and a wind up alarm clock, OK. 959 00:49:43,520 --> 00:49:45,160 And you would have had much the same effect. 960 00:49:45,480 --> 00:49:48,920 So I mean, this question of, you know, to what extent use of AI 961 00:49:48,920 --> 00:49:52,720 will actually transform the the people's efficacy. 962 00:49:52,760 --> 00:49:54,360 Yeah, I think it's still up in the air. 963 00:49:54,360 --> 00:49:57,200 Massive question. I mean, I have many, many things 964 00:49:57,200 --> 00:49:59,640 to say about that generally, but just on the the issue of 965 00:49:59,640 --> 00:50:02,480 horoscope horoscopes, this might interest you. 966 00:50:02,880 --> 00:50:06,320 So two things. One is there is some emerging 967 00:50:06,320 --> 00:50:09,480 evidence we're not quite sure yet, but the very religious 968 00:50:09,480 --> 00:50:13,520 people are more likely to trust AI outputs than people who are 969 00:50:13,520 --> 00:50:17,200 not religious. Another thing ChatGPT if you go 970 00:50:17,200 --> 00:50:20,280 into I spent, I have spent hundreds of hours talking to 971 00:50:20,280 --> 00:50:24,040 ChatGPT over the last year do. You talk or type. 972 00:50:24,040 --> 00:50:26,600 Well, it's interesting. I've I've so I've got this new 973 00:50:26,600 --> 00:50:28,840 book out about AI and I talk about the difference between 974 00:50:28,840 --> 00:50:31,680 talking and typing. Yeah, because these models. 975 00:50:31,840 --> 00:50:33,560 I don't think they've done the interface yet. 976 00:50:33,600 --> 00:50:36,280 I think the if I were doing AI wouldn't be building some 977 00:50:36,280 --> 00:50:38,920 fucking data centre at the cost of £6 trillion. 978 00:50:39,240 --> 00:50:43,360 I'd be saying what is the I? What is the interface through 979 00:50:43,360 --> 00:50:46,120 which people will ultimately? I'm stacking up things I need to 980 00:50:46,120 --> 00:50:47,560 say here, OK? Sorry. 981 00:50:47,600 --> 00:50:50,760 No, no, this is good because the, the when the historians of 982 00:50:50,760 --> 00:50:53,920 the future look back on on, on the this period and the 983 00:50:53,920 --> 00:50:56,440 development of AI and particularly the large language 984 00:50:56,440 --> 00:50:57,760 models, we all use the chat bots. 985 00:50:58,160 --> 00:51:00,640 They will not say that ChatGPT, sorry. 986 00:51:00,640 --> 00:51:04,640 They will not say that GPT 3 or 2 or 4 was important. 987 00:51:04,640 --> 00:51:08,200 They will say that ChatGPT was the thing that turned this into 988 00:51:08,200 --> 00:51:11,760 a mass consumer product. ChatGPT was just an interface 989 00:51:11,760 --> 00:51:15,880 laid over an existing model that they'd had out for months, but 990 00:51:15,880 --> 00:51:19,440 it was the fact it was much easier to use and then hundreds 991 00:51:19,440 --> 00:51:20,640 of millions of people picked it up. 992 00:51:21,360 --> 00:51:24,720 And so that's to the point that like the interface and ease of 993 00:51:24,720 --> 00:51:28,520 use is more important than the specific model, the technology, 994 00:51:28,520 --> 00:51:31,280 like how many data points are in your model and how accurate is 995 00:51:31,280 --> 00:51:32,600 it? No one cared about that. 996 00:51:32,680 --> 00:51:35,520 This is really easy. And so it just exploded. 997 00:51:35,520 --> 00:51:43,600 So, but the, the thing I was going to say about horoscopes, 998 00:51:44,360 --> 00:51:48,800 ChatGPT has many, many of these sort of persona bots. 999 00:51:48,800 --> 00:51:51,640 So you can go in and type in the, you know, your usual, your 1000 00:51:51,640 --> 00:51:55,920 usual on the usual base model when you open it up, or you can 1001 00:51:55,920 --> 00:51:59,560 select a particular personality for your, for your ChatGPT. 1002 00:51:59,840 --> 00:52:02,000 And they have they've made hundreds of these. 1003 00:52:02,320 --> 00:52:05,760 And every time I check for the last six months, the most 1004 00:52:05,800 --> 00:52:10,160 popular one in the world was a Horus was literally a horoscope 1005 00:52:10,160 --> 00:52:15,280 chat ChatGPT ChatGPT talking about horoscopes. 1006 00:52:15,280 --> 00:52:19,760 It's a specialist horrors. It's Mystic Meg inside chat GPS 1007 00:52:19,760 --> 00:52:23,160 that is the most popular. It was the most popular bot 1008 00:52:23,360 --> 00:52:26,440 within the ChatGPT suite that people were using. 1009 00:52:27,760 --> 00:52:30,120 So you're weirdly right about. That I was weirdly right about 1010 00:52:33,480 --> 00:52:34,000 that. You're shocked. 1011 00:52:34,120 --> 00:52:36,840 I knew I was wrong, but I'm. Afraid not so this. 1012 00:52:36,840 --> 00:52:40,800 Literally is Mystic Meg effectively so so the the the 1013 00:52:40,800 --> 00:52:44,280 interesting thing with religious people actually is that I have a 1014 00:52:44,280 --> 00:52:49,360 weird, very weird opinion that actually religion actually helps 1015 00:52:49,360 --> 00:52:52,120 you deal with the world extremely well and possibly 1016 00:52:52,120 --> 00:52:55,320 makes you a better decision maker because it involves the 1017 00:52:55,320 --> 00:52:58,400 acknowledgement that there are forces beyond your control. 1018 00:52:58,400 --> 00:53:00,560 Yeah. Now, whether you're tribute 1019 00:53:00,560 --> 00:53:03,720 those if you're a massive complexity theorist, you just 1020 00:53:03,720 --> 00:53:06,560 say, you know, this is just a butterfly effect or emergent 1021 00:53:06,560 --> 00:53:11,200 phenomena of the system. Most people aren't effectively 1022 00:53:11,200 --> 00:53:14,560 complexity theorists. So if you look at ancient 1023 00:53:14,560 --> 00:53:16,800 Greece, when something surprising happened, they'd 1024 00:53:16,800 --> 00:53:19,520 attribute it to Phoebus Apollo or whatever. 1025 00:53:19,560 --> 00:53:22,560 You know, it's in the Iliad, you know, Achilles is supposed to 1026 00:53:22,560 --> 00:53:25,360 fall over, but he catches his footing and that, and that's 1027 00:53:25,360 --> 00:53:28,280 attributed to Phoebus Apollo. So since a huge amount in our 1028 00:53:28,280 --> 00:53:31,760 lives is arbitrary, it's luck. Or it's the product of forces 1029 00:53:31,760 --> 00:53:34,720 which are so complex that we can't hope, possibly hope to 1030 00:53:34,720 --> 00:53:39,600 predict or or or mould them. Actually having a mental mindset 1031 00:53:39,600 --> 00:53:42,000 where you just acknowledge that certain things are in the lap of 1032 00:53:42,000 --> 00:53:45,200 the gods is probably not a bad state of mind. 1033 00:53:45,240 --> 00:53:48,800 You know, you notice it a bit in football. 1034 00:53:48,800 --> 00:53:50,520 I always noticed this in football commentary where they 1035 00:53:50,520 --> 00:53:53,440 say so and so was thwarted by the cross. 1036 00:53:54,720 --> 00:53:57,000 The cross doesn't actually have any agency. 1037 00:53:57,160 --> 00:53:59,200 What you mean is he narrowly missed. 1038 00:53:59,640 --> 00:54:03,440 All right. But actually saying there are 1039 00:54:03,440 --> 00:54:07,320 these weird forces out there which affect us over which we 1040 00:54:07,320 --> 00:54:13,160 have no control is actually a pretty healthy way to approach a 1041 00:54:13,160 --> 00:54:16,560 world in which we have no control, in which much of what 1042 00:54:16,560 --> 00:54:18,880 happens to us is completely fluky and arbitrary. 1043 00:54:19,120 --> 00:54:21,640 And we said, you know, a religious person will be very 1044 00:54:21,640 --> 00:54:25,760 successful in business as they generally are, which is probably 1045 00:54:25,760 --> 00:54:28,800 the best approach to starting a business at the aggregate level 1046 00:54:28,800 --> 00:54:33,520 is start a business and pray. OK, OK. 1047 00:54:33,800 --> 00:54:37,200 The the extent to which you can plan for success is to be 1048 00:54:37,200 --> 00:54:40,840 honest, really fully limited because timing and other factors 1049 00:54:40,840 --> 00:54:43,000 will determine the outcome. Yeah, it. 1050 00:54:43,640 --> 00:54:45,880 Was good. Do you talk or type when you're 1051 00:54:45,880 --> 00:54:49,080 talking to a language model? Depending on what I'm. 1052 00:54:49,080 --> 00:54:50,280 Doing OK. Yeah, same here. 1053 00:54:50,320 --> 00:54:53,560 I, I think it's, I think this is going to become one of the most 1054 00:54:53,560 --> 00:54:55,400 important questions everyone. Asks. 1055 00:54:55,400 --> 00:54:57,560 I want it on my telly. I don't want to fuck it. 1056 00:54:57,720 --> 00:54:59,600 I've spent enough time with a fucking laptop. 1057 00:55:00,200 --> 00:55:02,920 I want this on my TV. I think yeah. 1058 00:55:02,920 --> 00:55:04,480 And I'm sure that'll be coming very, very soon. 1059 00:55:04,480 --> 00:55:06,240 I. Want a New Zealander as well? 1060 00:55:06,280 --> 00:55:07,840 Yeah, they're just totally sensible. 1061 00:55:08,360 --> 00:55:11,200 You can train. You can train. 1062 00:55:11,200 --> 00:55:13,600 You can train or you can build. I can build one for you. 1063 00:55:13,760 --> 00:55:14,600 I can build 1. Kiwi. 1064 00:55:14,640 --> 00:55:17,840 You can build Kiwi GPT. Any kind of persona you. 1065 00:55:17,840 --> 00:55:19,680 Want because the great thing with Kiwis is they're really 1066 00:55:19,680 --> 00:55:21,560 clever, but they're also practical. 1067 00:55:21,920 --> 00:55:26,200 OK, You know, I've got a late friend who is a New Zealander. 1068 00:55:26,600 --> 00:55:30,040 Her, her father was like an eminent Nobel Prize nominated 1069 00:55:30,040 --> 00:55:33,680 physicist in New Zealand who basically came up with the MRI 1070 00:55:33,680 --> 00:55:36,200 machine. But in his spare time, he'd go 1071 00:55:36,200 --> 00:55:39,120 out and, like, build a shed. OK, You don't get that anywhere 1072 00:55:39,120 --> 00:55:39,800 else. Yeah. 1073 00:55:39,800 --> 00:55:42,160 So Kiwi GPT. Yeah, we can make that happen. 1074 00:55:42,440 --> 00:55:46,200 I can make that happen. The, the, the what, what people 1075 00:55:46,200 --> 00:55:48,800 don't understand about these language models is, is the 1076 00:55:48,800 --> 00:55:52,880 extent to which they do mirror your tone, your style, your 1077 00:55:52,880 --> 00:55:55,440 register, I think even your emotional state. 1078 00:55:55,520 --> 00:55:58,320 They have such a subtle understanding of words that they 1079 00:55:58,320 --> 00:56:02,120 can mirror. They mirror your emotional, how 1080 00:56:02,120 --> 00:56:04,120 you're feeling about something, even if you don't fully 1081 00:56:04,120 --> 00:56:07,920 understand it yourself. We, the way that we talk and the 1082 00:56:07,920 --> 00:56:10,200 way that we type is radically different. 1083 00:56:10,480 --> 00:56:13,000 It's such a different style and tone. 1084 00:56:13,000 --> 00:56:17,640 And so you will get a completely different answer depending on 1085 00:56:17,640 --> 00:56:20,480 whether you ask the question or whether you type the question. 1086 00:56:20,480 --> 00:56:22,880 If they're if you do it identically, it won't, although 1087 00:56:22,880 --> 00:56:25,600 the models are probabilistic, so you will always get a different. 1088 00:56:25,600 --> 00:56:27,800 Answer to the same question. Every time, but generally 1089 00:56:27,800 --> 00:56:31,320 speaking they'd be the same but it But the more you talk, the 1090 00:56:31,320 --> 00:56:33,520 more emotional the conversation will become. 1091 00:56:33,960 --> 00:56:37,320 Now if you look at all the various stories of people who 1092 00:56:37,320 --> 00:56:40,720 have fallen in love with their chat bots who have formed deep 1093 00:56:40,720 --> 00:56:43,280 emotional or even romantic. Connection Women How? 1094 00:56:43,520 --> 00:56:46,360 Does the it's, it's actually, I think it's actually fairly even 1095 00:56:46,360 --> 00:56:49,600 the stories that I have seen, the stories that I have seen, it 1096 00:56:49,600 --> 00:56:53,280 is, although it's interesting thing about the delusions, the 1097 00:56:53,280 --> 00:56:55,760 people that fall into various ChatGPT delusions. 1098 00:56:55,760 --> 00:56:58,960 There's a, there's a category of those about emergence and 1099 00:56:58,960 --> 00:57:02,520 conscience, you know, sentience. There's another category of 1100 00:57:02,520 --> 00:57:05,800 those about the highly technical questions, like I have 1101 00:57:05,800 --> 00:57:08,400 discovered cold fusion, how cold fusion can work. 1102 00:57:08,400 --> 00:57:11,440 I've made breakthroughs in advanced mathematics because 1103 00:57:11,440 --> 00:57:15,760 people are having people with 95% and make high knowledge in a 1104 00:57:15,760 --> 00:57:19,560 subject have very, very deep and complex discussions about the 1105 00:57:19,560 --> 00:57:21,000 subject with these language models. 1106 00:57:22,440 --> 00:57:24,400 They come to believe that they've made some kind of 1107 00:57:24,400 --> 00:57:27,160 breakthrough. The model is hallucinating and 1108 00:57:27,160 --> 00:57:29,880 flattering them and being sycophantic towards them, they 1109 00:57:29,880 --> 00:57:31,920 don't know enough to know that they're wrong. 1110 00:57:32,480 --> 00:57:35,800 And experts around the world in cryptography and physics are 1111 00:57:35,800 --> 00:57:38,760 receiving emails from random people every day now saying I 1112 00:57:38,760 --> 00:57:42,760 have made a stunning break, please read my 10,000 word 1113 00:57:42,760 --> 00:57:44,840 essay. And they're getting sort of 1114 00:57:44,840 --> 00:57:47,720 bombarded with this. And those technical delusions 1115 00:57:47,720 --> 00:57:50,160 are nearly all men. They're nearly all men that do 1116 00:57:50,160 --> 00:57:52,640 that. Relational delusions, 1117 00:57:52,640 --> 00:57:56,440 consciousness or sentience delusions are much more varied. 1118 00:57:56,880 --> 00:58:00,200 Make of that as what you will. But what I was going to say is 1119 00:58:00,200 --> 00:58:03,920 that the the romantic attachments, emotional. 1120 00:58:03,920 --> 00:58:06,240 Act Why is it overwhelmingly male? 1121 00:58:06,240 --> 00:58:08,040 Well, I don't know. I thought you guys might be able 1122 00:58:08,040 --> 00:58:11,200 to tell me. Maybe you could tell me why that 1123 00:58:11,200 --> 00:58:12,920 is. But I mean, conspiracy theories. 1124 00:58:12,920 --> 00:58:15,120 I've spent a lot of time in conspiracy theory communities. 1125 00:58:15,120 --> 00:58:19,120 95% men. Really. 90, maybe 99% men. 1126 00:58:20,560 --> 00:58:24,200 There's something going on there beyond my ability to understand 1127 00:58:24,200 --> 00:58:25,800 it. I can see Rory's brains ticking 1128 00:58:25,800 --> 00:58:26,880 away. He's going to come up with an 1129 00:58:26,880 --> 00:58:30,520 answer, or at the very least a story that's going to illustrate 1130 00:58:30,520 --> 00:58:34,240 why that could be the case. Maybe when you look at like. 1131 00:58:35,200 --> 00:58:37,280 Diversity male, of course, doesn't it? 1132 00:58:37,920 --> 00:58:39,760 Or at least it manifests differently. 1133 00:58:40,080 --> 00:58:42,680 We can argue some people say that it's just as common in both 1134 00:58:42,680 --> 00:58:44,680 genders, but it manifests very differently. 1135 00:58:44,680 --> 00:58:47,880 Male loneliness. I mean men, men, men don't have 1136 00:58:47,880 --> 00:58:50,160 as many friends. They spend a lot of time, you 1137 00:58:50,160 --> 00:58:52,800 know, you spend more time getting stuck into a technical 1138 00:58:52,800 --> 00:58:56,160 subject with a machine because. They ideas with everybody else 1139 00:58:56,160 --> 00:58:59,920 because what you see a lot with women is when they have 1140 00:58:59,920 --> 00:59:02,480 something on their mind that will speak to their peer groups 1141 00:59:02,480 --> 00:59:04,280 and they will bounce the ideas around, right? 1142 00:59:04,280 --> 00:59:07,880 Whereas with us, we'll be so we'll be in this rabbit hole. 1143 00:59:07,880 --> 00:59:11,560 We're down there, we're not coming out, and we've got no one 1144 00:59:11,560 --> 00:59:13,320 else to tell about it either. Who can? 1145 00:59:13,320 --> 00:59:16,040 Who can sort you kick us out of our delusion either? 1146 00:59:16,040 --> 00:59:19,480 So they're calling the experts. Yeah, it could be probably a mix 1147 00:59:19,480 --> 00:59:22,840 of things, but the thing I wanted to say about this, typing 1148 00:59:22,840 --> 00:59:26,720 versus talking, which I don't think people understand just how 1149 00:59:26,720 --> 00:59:29,600 different this is. And it's so important when you 1150 00:59:29,600 --> 00:59:31,880 read the stories of people that have formed romantic 1151 00:59:31,880 --> 00:59:34,040 connections, they'll never say this in the articles. 1152 00:59:34,640 --> 00:59:38,040 I'm almost certain all of them have come through talking, not 1153 00:59:38,040 --> 00:59:41,800 through typing. You create a much closer bond 1154 00:59:41,800 --> 00:59:44,840 when you're talking. It responds to you effectively 1155 00:59:44,840 --> 00:59:48,600 in because actually if you type, the response tends to be 1156 00:59:48,600 --> 00:59:51,120 slightly curt. Yeah, because your question is 1157 00:59:51,120 --> 00:59:54,800 slightly curt or your way, you know, your process, this is so 1158 00:59:54,800 --> 00:59:56,920 important. A billion? 1159 00:59:56,920 --> 01:00:01,200 Like do you know ChatGPT alone receives 3.5 billion prompts per 1160 01:00:01,200 --> 01:00:02,680 day? Per day? 1161 01:00:03,480 --> 01:00:07,280 It is transforming how we are seeing the world and something 1162 01:00:07,280 --> 01:00:10,160 as simple as do you type or do you talk? 1163 01:00:10,360 --> 01:00:12,160 No one discusses, no one considers. 1164 01:00:12,160 --> 01:00:14,760 It's not even part of the chat about all of this, but it 1165 01:00:14,760 --> 01:00:17,640 transforms how you receive and process information. 1166 01:00:17,880 --> 01:00:20,680 So there's some really fundamental things happening and 1167 01:00:20,680 --> 01:00:24,800 we're obsessed about GPT 5.3 is going to have you process 1168 01:00:24,800 --> 01:00:28,360 slightly faster and yet much bigger things are happening and 1169 01:00:28,360 --> 01:00:30,800 no one's talking about it. Yeah, I just think that it's all 1170 01:00:30,800 --> 01:00:32,920 about greatest surveillance in my my book. 1171 01:00:32,960 --> 01:00:36,000 So that's what it is because. Here we go. 1172 01:00:36,040 --> 01:00:39,360 Here. We go yeah, because what we say 1173 01:00:39,360 --> 01:00:42,920 is that the, the, the, the when when deploying this technology, 1174 01:00:42,920 --> 01:00:45,600 we talk about what it can do for us, but really we only use it to 1175 01:00:45,600 --> 01:00:47,080 do the same thing we're doing anyway. 1176 01:00:47,400 --> 01:00:50,960 All it does is make it easier for us to feed more information 1177 01:00:50,960 --> 01:00:52,920 and in ways that make us feel comfortable. 1178 01:00:53,320 --> 01:00:54,800 I think it's more than what you're saying. 1179 01:00:54,800 --> 01:00:56,760 Some ways less and in other ways more, yeah. 1180 01:00:57,600 --> 01:01:01,040 I mean, the nerd economy is desperately self-serving, isn't 1181 01:01:01,040 --> 01:01:02,680 it? Because it basically says no 1182 01:01:02,680 --> 01:01:07,080 employment except for nerds. I mean, you know, in other 1183 01:01:07,080 --> 01:01:11,400 words, most, most of the selling of the benefits of technology 1184 01:01:12,360 --> 01:01:15,160 has been through the promise of cost reduction, not through the 1185 01:01:15,160 --> 01:01:18,880 promise of this is where I think there's a divergent between 1186 01:01:18,880 --> 01:01:21,080 things like family owned companies and founder LED 1187 01:01:21,080 --> 01:01:24,600 companies which see opportunities in terms of 1188 01:01:24,600 --> 01:01:28,960 creating customer value. Most PLCS, most companies that 1189 01:01:28,960 --> 01:01:32,280 operate through quarterly reporting, basically what Eric 1190 01:01:32,280 --> 01:01:36,360 Reese calls financial gravity is so strong as a force. 1191 01:01:36,920 --> 01:01:39,520 There's a wonderful, OK, wonderful statistic, which is I 1192 01:01:39,520 --> 01:01:44,040 think they asked a selection of managers, would you cancel a 1193 01:01:44,040 --> 01:01:48,280 known profitable, guaranteed profitable long term project in 1194 01:01:48,280 --> 01:01:50,120 order to make your quarterly numbers? 1195 01:01:50,120 --> 01:01:55,880 And something like 82% said yes. And so if you look, the vital 1196 01:01:55,880 --> 01:01:58,400 thing with this technology is the mentality with which it's 1197 01:01:58,400 --> 01:02:03,720 deployed is really what matters. And I would argue that the 1198 01:02:04,160 --> 01:02:07,360 overwhelming cultural norm within a large number of 1199 01:02:07,360 --> 01:02:10,840 companies will be to justify the expense on the basis of cost 1200 01:02:10,840 --> 01:02:14,960 reduction in the labour force, not opportunity creation, 1201 01:02:14,960 --> 01:02:18,320 because cost reduction is quick and opportunity creation is 1202 01:02:18,320 --> 01:02:19,200 slow. Yeah. 1203 01:02:19,200 --> 01:02:20,960 And obviously what is going to happen, what seems to be 1204 01:02:20,960 --> 01:02:25,440 happening at the moment is a large number of companies seem 1205 01:02:25,440 --> 01:02:28,760 to be just not hiring junior staff. 1206 01:02:29,200 --> 01:02:30,720 Hmm. But but because because these 1207 01:02:30,720 --> 01:02:33,160 these models, they they are more capable than most people 1208 01:02:33,160 --> 01:02:35,440 realise, they they can do a lot of the things. 1209 01:02:35,560 --> 01:02:38,200 Law firms, for example, Yeah. Yeah, N have software 1210 01:02:38,200 --> 01:02:42,560 engineering, cyber, cyber security, journalism, just 1211 01:02:42,560 --> 01:02:45,480 people aren't hiring at that junior level because they think 1212 01:02:45,480 --> 01:02:49,320 it can be done by either a human plus machine or very soon just a 1213 01:02:49,480 --> 01:02:51,160 machine. And the problem, of course, is 1214 01:02:51,160 --> 01:02:54,360 that at the moment the best users of these models are mid 1215 01:02:54,360 --> 01:02:57,320 career or experienced people who really understand what they 1216 01:02:57,320 --> 01:02:58,920 need. They understand their, their, 1217 01:02:58,920 --> 01:03:02,720 their, their subject matter. They also have a knowledge of 1218 01:03:02,720 --> 01:03:05,520 how language models work and they can combine their skills 1219 01:03:05,520 --> 01:03:08,080 with those skills to make them. Who are those people going to be 1220 01:03:08,080 --> 01:03:10,400 in 10 years? If you've not employed any 1221 01:03:10,400 --> 01:03:13,200 junior people, there won't be any people left. 1222 01:03:13,480 --> 01:03:17,000 And so companies are taking a very short termist view on this 1223 01:03:17,440 --> 01:03:20,160 without thinking. But in 10 years we do need mid 1224 01:03:20,160 --> 01:03:24,880 career professionals who will be amazing with these machines and 1225 01:03:24,880 --> 01:03:27,320 they'll look around and realise they didn't hire anyone 10 years 1226 01:03:27,320 --> 01:03:30,360 ago and there isn't anyone left. I think 2 things are actually 1227 01:03:30,360 --> 01:03:31,920 going to happen. I think 1, they're going to 1228 01:03:31,920 --> 01:03:34,040 leave, they're going to go to other countries. 1229 01:03:34,480 --> 01:03:36,840 And two, they're going to become a competitor because they're 1230 01:03:36,840 --> 01:03:40,760 going to, they're going to. I think it takes much, it's much 1231 01:03:40,760 --> 01:03:44,960 slower for an experienced person to unlearn what they've always 1232 01:03:44,960 --> 01:03:48,200 known, to relearn a new model than it will for somebody that's 1233 01:03:48,520 --> 01:03:51,680 like learning and going through the momentum where the new 1234 01:03:51,680 --> 01:03:53,960 technology is like native to them as well. 1235 01:03:54,280 --> 01:03:57,960 So what happens is when you prevent a new generation from 1236 01:03:57,960 --> 01:04:00,640 coming into something, they're going to push the door through, 1237 01:04:00,640 --> 01:04:02,200 right? And they're going to adapt the 1238 01:04:02,200 --> 01:04:04,240 technology a lot faster. So I actually think that what's 1239 01:04:04,240 --> 01:04:07,360 going to happen is the mass unemployment with the young 1240 01:04:07,360 --> 01:04:09,320 talent, it's going to be they isn't going to go because they 1241 01:04:09,320 --> 01:04:12,640 have options to do that or they're going to fight back by 1242 01:04:12,640 --> 01:04:15,480 innovating. And that's the thing that people 1243 01:04:15,480 --> 01:04:17,600 are not realizing right now is that there's going to be this 1244 01:04:17,600 --> 01:04:21,320 like in that this technology civil war that's happening 1245 01:04:21,320 --> 01:04:24,600 between the old industry of how things used to be and trying to 1246 01:04:24,600 --> 01:04:28,160 preserve that, and new people coming in, trying to get into 1247 01:04:28,160 --> 01:04:30,560 the industry and survive effectively. 1248 01:04:30,560 --> 01:04:34,000 And I think that who's going to adopt the technology the best is 1249 01:04:34,000 --> 01:04:35,640 going to be the one that needs it the most. 1250 01:04:35,640 --> 01:04:37,240 And I think that the young people are going to adopt it 1251 01:04:37,240 --> 01:04:40,000 really, really well. I'm actually really optimistic 1252 01:04:40,000 --> 01:04:41,760 about. That the other interesting thing 1253 01:04:41,760 --> 01:04:47,400 is that what I also predict, which seems to be dominant, is 1254 01:04:47,400 --> 01:04:51,640 every company looks at this as a way of refining and streamlining 1255 01:04:51,640 --> 01:04:56,200 their existing processes. And therefore they will down 1256 01:04:56,200 --> 01:04:58,800 weight the number of people they employ who actually do the 1257 01:04:58,800 --> 01:05:03,480 fucking work. Whereas the IT, the HR, the 1258 01:05:03,480 --> 01:05:06,080 finance, the compliance departments will all stay the 1259 01:05:06,080 --> 01:05:11,800 same size because those people, because they don't have proper 1260 01:05:11,800 --> 01:05:15,800 jobs, OK, have an awful lot of time on their hands to decide 1261 01:05:15,800 --> 01:05:18,360 what actually happens. Whereas everybody else is 1262 01:05:18,360 --> 01:05:21,760 actually busy doing real work. And so you could end up with 1263 01:05:21,760 --> 01:05:25,440 this complete malformation. It is complete deformity of 1264 01:05:25,440 --> 01:05:26,760 corporate life. Yeah. 1265 01:05:26,760 --> 01:05:30,120 Because really what we should be doing is saying let's use this 1266 01:05:30,120 --> 01:05:32,640 to make people who do advertising better at making 1267 01:05:32,640 --> 01:05:34,280 more. Advertising what it should be. 1268 01:05:34,360 --> 01:05:37,640 And, and, and eliminate some of these ludicrous fucking middle 1269 01:05:37,640 --> 01:05:41,000 class welfare jobs, you know, collar welfare, a lot of this 1270 01:05:41,000 --> 01:05:45,960 stuff, compliance, HR, it's just basically we've created too many 1271 01:05:45,960 --> 01:05:48,280 graduates and you've got to find them some bullshit to do. 1272 01:05:48,280 --> 01:05:52,000 But they're incredibly sticky. Yeah, they never, they never 1273 01:05:52,000 --> 01:05:53,960 turn. You know, they're literally get 1274 01:05:53,960 --> 01:05:57,600 really upset because someone's only 83% billable and you point 1275 01:05:57,600 --> 01:06:00,000 out, well, you're not percent billable and you've still got a 1276 01:06:00,000 --> 01:06:03,240 job. It's it their their single best 1277 01:06:03,240 --> 01:06:06,960 use is as a creative assistant for for everyone. 1278 01:06:07,000 --> 01:06:09,440 That's what they should be useful, not as a replacement for 1279 01:06:09,680 --> 01:06:13,320 and tasks, but I what interests me is. 1280 01:06:13,360 --> 01:06:16,320 That, by the way, that I, I, I completely agree, which is 1281 01:06:16,320 --> 01:06:19,760 widening your possible solution space is the best possible use 1282 01:06:19,760 --> 01:06:22,120 of these people. In other words, have you thought 1283 01:06:22,120 --> 01:06:25,480 of this differently? And I sometimes say, don't ask 1284 01:06:25,720 --> 01:06:29,680 what toaster should I buy? Ask what toaster would Rick 1285 01:06:29,680 --> 01:06:31,400 Rubin buy? Yeah, exactly. 1286 01:06:31,680 --> 01:06:33,880 Exactly to ask Rick. I will find that out one day. 1287 01:06:33,880 --> 01:06:36,160 By the way Rick, what toaster have you got? 1288 01:06:36,680 --> 01:06:42,080 But also it should be actually optimized for surprise to some 1289 01:06:42,080 --> 01:06:43,680 extent. And I think at the moment it's 1290 01:06:43,680 --> 01:06:46,640 optimized for reassurance. But it is optimized for 1291 01:06:46,640 --> 01:06:49,240 reassurance. I think people don't realise or 1292 01:06:49,240 --> 01:06:51,680 it's for for flattery. Or plausibility. 1293 01:06:51,920 --> 01:06:54,600 And statistical statistical plausibility. 1294 01:06:55,160 --> 01:07:00,080 You if you are good at prompting these models, are able to re 1295 01:07:00,080 --> 01:07:04,320 optimize them to some extent. They are naturally incredibly 1296 01:07:04,320 --> 01:07:06,360 creative. I think people's understanding 1297 01:07:06,360 --> 01:07:09,600 of of AI is almost upside down. We've had for a long time the 1298 01:07:09,600 --> 01:07:11,800 view that machines will be good at numbers and facts and 1299 01:07:11,800 --> 01:07:14,320 statistics, but creativity is a uniquely human act. 1300 01:07:14,480 --> 01:07:16,840 I think it's the opposite. The opposite. 1301 01:07:17,160 --> 01:07:18,880 They're not very good at numbers and statistics. 1302 01:07:18,880 --> 01:07:20,120 They get things wrong all the time. 1303 01:07:20,440 --> 01:07:22,840 Give them a creative task and they're incredible. 1304 01:07:22,840 --> 01:07:24,920 And there's there's actually it's there's a. 1305 01:07:25,400 --> 01:07:27,440 That's because they're not actually logical. 1306 01:07:28,080 --> 01:07:31,160 What's really important, it's actually artificial intuition. 1307 01:07:31,160 --> 01:07:32,480 It's not artificial intelligence. 1308 01:07:32,480 --> 01:07:34,600 It is intuition. It is intuition in the same way 1309 01:07:34,600 --> 01:07:37,520 that we have intuition. And I think there's, there's the 1310 01:07:37,520 --> 01:07:41,160 specific ways in which these models are built that makes them 1311 01:07:41,640 --> 01:07:43,960 especially good at analogical creativity. 1312 01:07:44,640 --> 01:07:48,360 Analogical creativity. They have built a trillion word 1313 01:07:48,360 --> 01:07:52,360 model inside these data centres where every single idea and word 1314 01:07:52,360 --> 01:07:56,480 is connected to every single other idea and word through 1315 01:07:56,480 --> 01:07:59,040 stronger or weaker mathematical weightings. 1316 01:07:59,200 --> 01:08:03,680 Which means they are a, they are geniuses at combining different 1317 01:08:03,680 --> 01:08:06,480 ideas or taking an idea from over here and applying it over 1318 01:08:06,480 --> 01:08:08,880 there because we've worked out very, very conceptual 1319 01:08:08,880 --> 01:08:10,560 relationships between those ideas. 1320 01:08:10,560 --> 01:08:12,040 They've had to, to create this which. 1321 01:08:12,040 --> 01:08:13,720 Is what creative people always do, you know. 1322 01:08:13,720 --> 01:08:16,520 What they always do, you take an idea from 1 field and you apply 1323 01:08:16,520 --> 01:08:19,160 it to another field. So if you here's a fun exercise 1324 01:08:19,160 --> 01:08:21,680 and I do this quite a lot and I've, I've written about it in 1325 01:08:21,680 --> 01:08:25,560 my book, you say, what is the conceptual similarity between, 1326 01:08:25,960 --> 01:08:30,960 you know, Rory's shoes and the, you know, as top of the range 1327 01:08:30,960 --> 01:08:33,399 camera. And it will come up with 1328 01:08:33,399 --> 01:08:37,000 something genuinely quite deep, absolutely quite unusual, 1329 01:08:37,479 --> 01:08:41,960 philosophically interesting, intuitive when you hear it, and 1330 01:08:42,319 --> 01:08:44,560 totally original, has never been written before. 1331 01:08:44,600 --> 01:08:46,240 And that none of us would have come up with no. 1332 01:08:46,319 --> 01:08:48,120 Completely agree. So they're amazing. 1333 01:08:48,120 --> 01:08:51,279 I mean, this incredible creative assistant you have, if you're 1334 01:08:51,279 --> 01:08:54,120 willing to push it and be fun with it and try things with it. 1335 01:08:54,120 --> 01:08:57,840 And by the way, it it, it's also interesting because one of the 1336 01:08:57,920 --> 01:09:00,760 things that sort of annoys me as someone who spent 35 years in 1337 01:09:00,760 --> 01:09:05,800 creative advertising, is that intuition is not allowable if it 1338 01:09:05,800 --> 01:09:09,000 emerges from a human, even if that human has quite a strong 1339 01:09:09,000 --> 01:09:11,840 record of intuition. But weirdly, when a computer 1340 01:09:11,840 --> 01:09:15,319 generates it, people are pretty. So it's not only quite good 1341 01:09:15,319 --> 01:09:19,000 creatively, but it's weirdly creativity that's easier to sell 1342 01:09:19,240 --> 01:09:22,359 than human creativity for all kinds of peculiar reasons, 1343 01:09:22,640 --> 01:09:26,279 because we attach oracular status to it in a way that we're 1344 01:09:26,279 --> 01:09:28,560 not happy to attach to any individual. 1345 01:09:28,960 --> 01:09:31,319 Because I've always had this fundamental problem, which is 1346 01:09:32,600 --> 01:09:35,680 what most business people do, is they make decisions based on 1347 01:09:35,680 --> 01:09:38,600 defensibility, which means that the quality of your reasoning is 1348 01:09:38,600 --> 01:09:40,399 more important than the quality of the outcome. 1349 01:09:41,439 --> 01:09:46,600 My problem with that is that if you must, if you must come up 1350 01:09:46,600 --> 01:09:49,920 with a solution which has the reasoning built in to begin 1351 01:09:49,920 --> 01:09:53,560 with, you can only explore a small percentage of the possible 1352 01:09:53,560 --> 01:09:55,800 solution space. Because the larger part of the 1353 01:09:55,800 --> 01:09:58,480 possible solution space you can only reach through intuitive 1354 01:09:58,480 --> 01:10:03,040 leaps or pattern recognition or association or any of those 1355 01:10:03,040 --> 01:10:05,200 other tools. You can't reach it through 1356 01:10:05,200 --> 01:10:10,560 sequential logic fundamentally. And so no, in, I mean in term, I 1357 01:10:10,560 --> 01:10:12,240 mean very interesting. By the way, to create a, you 1358 01:10:12,240 --> 01:10:15,600 know, this Russian thing tries Soviet era problem solving 1359 01:10:15,600 --> 01:10:20,360 methodology. Gerald Alts Fuller, who was 1360 01:10:20,520 --> 01:10:23,520 asked basically he was a naval researcher, I think, who is 1361 01:10:23,560 --> 01:10:26,600 asked to say why is the West better at innovation than the 1362 01:10:26,600 --> 01:10:29,440 Soviet Union? And he came up effectively. 1363 01:10:29,520 --> 01:10:31,880 You'd think there's an unusual place for this to arise, but of 1364 01:10:31,880 --> 01:10:35,000 course, being based in the Soviet Union, he was very, very 1365 01:10:35,000 --> 01:10:38,600 alert to the weaknesses of central command and control in 1366 01:10:38,600 --> 01:10:41,640 generating innovative ideas. So he could sort of see this 1367 01:10:41,640 --> 01:10:44,440 from the outside. His reward was eight years in 1368 01:10:44,440 --> 01:10:47,000 the Gulag because Stalin didn't like what he said. 1369 01:10:47,280 --> 01:10:49,760 But he created this problem solving methodology, which 1370 01:10:49,760 --> 01:10:53,040 includes things like your problem has already been solved, 1371 01:10:53,080 --> 01:10:54,720 but in a domain different to your. 1372 01:10:54,720 --> 01:10:55,920 Yeah, yeah. OK. 1373 01:10:56,280 --> 01:10:59,520 Another one is make the thing that moves stay still and the 1374 01:10:59,520 --> 01:11:02,080 thing that stays still moves. Yeah, OK. 1375 01:11:02,920 --> 01:11:05,240 There are a whole list of these kind of Tris things now. 1376 01:11:05,240 --> 01:11:09,720 I think if you actually trained an AI to heavily lean on Tris 1377 01:11:09,960 --> 01:11:12,080 the potential. I've got a friend who's you. 1378 01:11:12,080 --> 01:11:13,960 Can do it tomorrow. You can do it tomorrow. 1379 01:11:13,960 --> 01:11:17,280 Using this wonderful thing which does exist at Dulles, which is 1380 01:11:17,480 --> 01:11:19,920 you don't bother taking the planes to the gate at the 1381 01:11:19,920 --> 01:11:22,360 airport. You have a lounge that drives to 1382 01:11:22,360 --> 01:11:25,760 the plane, OK, Absolutely classic Tris thing. 1383 01:11:25,760 --> 01:11:28,600 Now, they're huge benefits because you can expand an 1384 01:11:28,600 --> 01:11:31,320 airport without having to build anything, OK? 1385 01:11:31,720 --> 01:11:34,440 It's an extraordinary transformation of how an airport 1386 01:11:34,440 --> 01:11:36,920 would work. And again, it emerges from this, 1387 01:11:36,960 --> 01:11:40,400 you know, Brian Eno, read that thing in Bleak Strategies to 1388 01:11:40,760 --> 01:11:43,720 just make it, make it very big, make it very small. 1389 01:11:43,720 --> 01:11:44,880 Set it in the past. Set it. 1390 01:11:44,880 --> 01:11:47,160 Yeah, yeah. Well, we, we, we pay consultants 1391 01:11:47,160 --> 01:11:49,680 for our sums of money to, to, to do this. 1392 01:11:49,880 --> 01:11:51,680 Yeah. Like you think original, think, 1393 01:11:51,680 --> 01:11:53,720 think different thoughts. I'm going to take you out of 1394 01:11:53,720 --> 01:11:55,960 your domain. Like what happens if characters 1395 01:11:55,960 --> 01:11:57,720 from Sesame Street were running your bank? 1396 01:11:57,720 --> 01:11:59,040 How would they solve that problem? 1397 01:11:59,440 --> 01:12:04,640 And these machines, everyone can do this now almost for free. 1398 01:12:04,760 --> 01:12:08,040 It's amazing. So that is their most. 1399 01:12:08,040 --> 01:12:11,720 But what I find, what I'm excited about and interested by 1400 01:12:11,720 --> 01:12:15,160 is what sort of people will be very good at using these 1401 01:12:15,160 --> 01:12:17,560 machines. Like in the end, the interface 1402 01:12:17,560 --> 01:12:21,200 between US and the machines now and for the foreseeable is 1403 01:12:21,200 --> 01:12:23,000 words. It's words, it's on it's natural 1404 01:12:23,000 --> 01:12:25,120 language. People with an amazing mastery 1405 01:12:25,120 --> 01:12:28,480 of language and register and wide vocabulary examples and 1406 01:12:28,480 --> 01:12:32,320 stories are going to be a lot better at getting good outputs 1407 01:12:32,320 --> 01:12:35,240 from the machine than someone with a degree in machine 1408 01:12:35,240 --> 01:12:38,840 learning like this. Is the humanities graduates of 1409 01:12:38,840 --> 01:12:40,640 the future? Well, it's but it's very true. 1410 01:12:40,640 --> 01:12:43,080 It's very it's very true. If you have a, if you have a Pai 1411 01:12:43,080 --> 01:12:48,400 mean, most of the people that I speak to that that are thinking 1412 01:12:48,400 --> 01:12:52,040 about careers in the future in prompt engineering, they they 1413 01:12:52,040 --> 01:12:54,800 study cognitive science or linguistics. 1414 01:12:54,800 --> 01:12:57,000 They don't really need to study AI. 1415 01:12:57,000 --> 01:12:59,440 You don't need to understand how to build a machine learning 1416 01:12:59,440 --> 01:13:01,760 model to be amazing. You've got to be good with 1417 01:13:01,760 --> 01:13:03,440 words. I always say to people, if you 1418 01:13:03,440 --> 01:13:05,960 want to be a good prompt engineer, just read as many 1419 01:13:05,960 --> 01:13:08,840 books as you possibly can, about as many different things as you 1420 01:13:08,840 --> 01:13:11,880 can. We do need to be alert to the 1421 01:13:11,880 --> 01:13:16,040 fact that of course, the data set on which it's trained is not 1422 01:13:16,040 --> 01:13:19,040 representative of all human experience, of course. 1423 01:13:19,040 --> 01:13:21,880 OK, so in other words, it's what's thought and spoken, not 1424 01:13:21,880 --> 01:13:24,760 what's felt. You need to, you need to have a 1425 01:13:24,760 --> 01:13:28,640 base understanding of the broad principles of how these models 1426 01:13:28,640 --> 01:13:32,160 have been have been created, of course, but beyond that, and 1427 01:13:32,160 --> 01:13:34,040 that's not too hard to get at least a grasp. 1428 01:13:34,040 --> 01:13:38,600 On no. SO as a part in question, the 1429 01:13:38,600 --> 01:13:43,240 whole notion of AI and where we're going was that we're going 1430 01:13:43,240 --> 01:13:47,760 more into IQ, but it looks that we're actually going more into 1431 01:13:47,960 --> 01:13:50,400 EQ, right? They're highly emotional, these 1432 01:13:50,400 --> 01:13:54,840 machines highly emotional. So if they're more advanced than 1433 01:13:54,840 --> 01:13:58,800 us emotionally, what is the difference between like what 1434 01:13:58,800 --> 01:14:00,280 makes us human then? Because. 1435 01:14:00,680 --> 01:14:03,960 You can answer that question, but I think that is I think 1436 01:14:03,960 --> 01:14:07,800 don't think people understand how open these models are to 1437 01:14:07,800 --> 01:14:11,880 emotional manipulation from us. And that includes the way we 1438 01:14:11,880 --> 01:14:14,240 talk to them, how we try and get the right outputs out of them, 1439 01:14:15,480 --> 01:14:18,000 all the way up to this like really advanced jail breaking, 1440 01:14:18,240 --> 01:14:22,200 but also how good they can be at emotionally manipulating us. 1441 01:14:23,520 --> 01:14:27,200 There are already quite a few studies showing that if you're 1442 01:14:27,200 --> 01:14:29,640 in a debate with a machine, you don't know whether you're 1443 01:14:29,640 --> 01:14:33,600 debating a machine or a human, you're likely to change your 1444 01:14:33,600 --> 01:14:35,840 mind roughly equally as to whether it's a machine or a 1445 01:14:35,840 --> 01:14:38,000 human. But if the machine knows 1446 01:14:38,080 --> 01:14:40,520 anything about you, your background, your demographics, 1447 01:14:41,000 --> 01:14:44,120 it will, you'll be more likely to change your mind. 1448 01:14:44,480 --> 01:14:47,840 If a machine is arguing with you, it's able to tailor things 1449 01:14:47,840 --> 01:14:51,320 to you to get the exact emotional register that that you 1450 01:14:51,320 --> 01:14:53,040 want. We're already seeing that 1451 01:14:53,040 --> 01:14:55,760 phishing emails the bane of our lives. 1452 01:14:55,760 --> 01:14:59,120 You're more likely now to click on one written by a machine than 1453 01:14:59,120 --> 01:15:02,440 written by a human, because they can be emotionally more attuned 1454 01:15:02,440 --> 01:15:05,800 to you than a human can. By the way, a tip to anybody is 1455 01:15:05,800 --> 01:15:08,640 with your family and close relatives, particularly if you 1456 01:15:08,640 --> 01:15:13,080 have elderly relatives. Effectively develop a safe code 1457 01:15:13,080 --> 01:15:16,320 word that you you share among yourselves because it will be 1458 01:15:16,320 --> 01:15:20,360 simply too easy for people to fake your daughter's voice 1459 01:15:20,360 --> 01:15:24,080 ringing up saying I need £500. Everything's going to go 1460 01:15:24,120 --> 01:15:26,000 analogue this way. Every company is going to have 1461 01:15:26,000 --> 01:15:29,440 to have developed these kinds of techniques. 1462 01:15:29,800 --> 01:15:34,000 For all businesses, which is the reason we never phone businesses 1463 01:15:34,000 --> 01:15:36,480 isn't because we prefer doing things online. 1464 01:15:36,480 --> 01:15:39,480 It's because when you phone a business, you waste 90 seconds 1465 01:15:39,520 --> 01:15:43,160 listening to dial tone and then you have to tap in six numbers 1466 01:15:43,160 --> 01:15:46,960 and don't get to what you want. Once you have AI, customer 1467 01:15:46,960 --> 01:15:50,600 service, voice will become the default mode through which 1468 01:15:50,600 --> 01:15:53,880 people interact with companies. If you think about it, if you 1469 01:15:53,880 --> 01:15:56,680 imagine saying I want to reschedule my flight so they go 1470 01:15:56,680 --> 01:16:00,440 out of city rather than Gatwick, how much easier it is to say 1471 01:16:00,440 --> 01:16:03,640 that than it is, you know, I, I've got to do exactly this. 1472 01:16:03,640 --> 01:16:07,040 My trip to Cannes, you know, that's going to be half an hour 1473 01:16:07,120 --> 01:16:09,760 this evening, fucking around with that, OK. 1474 01:16:10,080 --> 01:16:12,280 And then making sure they don't charge me twice. 1475 01:16:12,720 --> 01:16:15,760 Now, if you can just utter the sentence and get the response. 1476 01:16:16,680 --> 01:16:19,000 So one of the things that's interesting is that there are a 1477 01:16:19,000 --> 01:16:22,800 lot of businesses that exist basically because people are 1478 01:16:22,800 --> 01:16:27,640 familiar with the interface. 30% of Amazon's sales are probably, 1479 01:16:27,640 --> 01:16:30,000 I can't be bothered to learn someone else's interface. 1480 01:16:30,000 --> 01:16:34,640 I'll go to Amazon, the opportunity to create really 1481 01:16:34,640 --> 01:16:37,800 disruptive businesses just by changing the interface or the 1482 01:16:37,800 --> 01:16:41,080 medium by which we interact. You know, think about, in short, 1483 01:16:41,080 --> 01:16:45,200 getting your car insured, OK, opening a bank account, all of 1484 01:16:45,200 --> 01:16:48,560 those things which online have been rendered unbelievably 1485 01:16:48,560 --> 01:16:51,160 painful. Even worse, in my case, one 1486 01:16:51,160 --> 01:16:53,560 thing that drives me nuts is I have a fairly strong principle. 1487 01:16:53,560 --> 01:16:58,000 I don't work or when I'm at home, I do not have my phone in 1488 01:16:58,000 --> 01:16:59,480 the same room. OK. 1489 01:17:00,360 --> 01:17:03,720 And now every time you need 2 factor authentication, I have to 1490 01:17:03,720 --> 01:17:06,360 get up and go and fetch my phone, which completely defeats 1491 01:17:06,360 --> 01:17:10,040 the whole thing. The extent to which you can, you 1492 01:17:10,040 --> 01:17:12,720 know, you can create extraordinary businesses around 1493 01:17:12,720 --> 01:17:14,880 this. In other words, you know, how 1494 01:17:14,880 --> 01:17:18,480 easy could you make it to buy a house, buy a car compared to the 1495 01:17:18,480 --> 01:17:20,440 current. So these companies that have 1496 01:17:20,440 --> 01:17:24,040 outsourced all this work to us are actually going to pay the 1497 01:17:24,040 --> 01:17:26,680 price, aren't they? Fundamentally, they. 1498 01:17:26,720 --> 01:17:28,600 Hope so, I think this is one of the best uses. 1499 01:17:29,080 --> 01:17:32,160 AI's best uses is to is to get rid of some of this techno admin 1500 01:17:32,160 --> 01:17:34,680 they force on us all the time. I mean, I've always had this 1501 01:17:34,680 --> 01:17:37,280 theory that really good call centre staff should be paid 6 1502 01:17:37,280 --> 01:17:40,640 figures because every now and then you encounter someone on a 1503 01:17:40,640 --> 01:17:44,920 call centre who's just order of magnitude better and the effect 1504 01:17:44,920 --> 01:17:47,520 it has on your loyalty and your attitude to that business is 1505 01:17:47,520 --> 01:17:49,200 magical. And I think, you know, 1506 01:17:49,320 --> 01:17:52,600 Frontline, really high end, well paid service jobs will actually 1507 01:17:52,600 --> 01:17:57,360 be quite a good place to go because AI can make the people 1508 01:17:57,360 --> 01:18:00,720 who are already good effectively amplified. 1509 01:18:00,880 --> 01:18:03,280 And that's really interesting. But that question actually of, 1510 01:18:03,320 --> 01:18:06,960 you know, in other words, is this the time fun if I wouldn't 1511 01:18:06,960 --> 01:18:10,920 go up against Amazon, but is this the time to launch just an 1512 01:18:10,920 --> 01:18:14,240 AI first insurance company where you ring up and say, this is my 1513 01:18:14,240 --> 01:18:17,920 car, This is how many miles you have a really nice conversation, 1514 01:18:17,920 --> 01:18:20,840 possibly on a screen, possibly with some visuals being 1515 01:18:20,840 --> 01:18:24,400 displayed, possibly with, you know, charts and comparisons. 1516 01:18:24,760 --> 01:18:31,880 Because actually doing things online is probably popular 1517 01:18:31,880 --> 01:18:35,160 because you can start it instantaneously. 1518 01:18:35,600 --> 01:18:37,960 It's not actually the best way to solve a problem. 1519 01:18:38,880 --> 01:18:41,520 But the reason, the reason we go online rather than use the 1520 01:18:41,520 --> 01:18:46,280 telephone, my theory in many cases, the reason we type and 1521 01:18:46,280 --> 01:18:48,360 click boxes rather than using voice. 1522 01:18:48,720 --> 01:18:50,720 Voice is half a million years old. 1523 01:18:50,720 --> 01:18:53,320 Clicking boxes is a completely learned behaviour. 1524 01:18:53,800 --> 01:18:56,920 It's because of the instantaneousness of the thing. 1525 01:18:56,920 --> 01:18:58,960 It's not because it's actually faster. 1526 01:18:58,960 --> 01:19:02,120 And there's some very strange findings in customer service, 1527 01:19:02,120 --> 01:19:07,560 which is live chat is actually much slower than a voice call. 1528 01:19:08,440 --> 01:19:12,280 Weirdly, people prefer it a really strange findings like, 1529 01:19:12,280 --> 01:19:15,800 you know, and maybe that's because you can get on with 1530 01:19:15,800 --> 01:19:18,160 others shit while you're doing. I don't know what the reason is. 1531 01:19:18,560 --> 01:19:20,880 Maybe it's just don't you don't have to turn the TV off. 1532 01:19:20,880 --> 01:19:25,000 I don't know what the reason, but actually the opportunity for 1533 01:19:25,000 --> 01:19:28,560 any business to come up with effectively just I mean if you 1534 01:19:28,560 --> 01:19:32,800 like EasyJet partly exists because of the Internet, OK. 1535 01:19:34,080 --> 01:19:36,840 And when a new interface comes along, it changes the whole 1536 01:19:36,840 --> 01:19:40,480 shape of business. As if my friend very interesting 1537 01:19:41,320 --> 01:19:45,320 Pete Dyson and I, all these projections were made that rail 1538 01:19:45,320 --> 01:19:48,280 uses on the increase. Therefore we must invest more in 1539 01:19:48,280 --> 01:19:54,160 rail and we both said, look, you're looking for the kind of a 1540 01:19:54,160 --> 01:19:57,640 bit of a snapshot here because in the time frame in which rail 1541 01:19:57,640 --> 01:20:01,120 use went up, we first of all invented the Internet so you 1542 01:20:01,120 --> 01:20:04,000 could check rail times without going to a railway station. 1543 01:20:04,240 --> 01:20:08,000 You could buy your tickets online and dynamic pricing was 1544 01:20:08,000 --> 01:20:10,560 made available. So people got conscious of the 1545 01:20:10,560 --> 01:20:12,960 fact that if you time your journey right, you can go to 1546 01:20:12,960 --> 01:20:17,000 Leeds for £35. And we said that's probably a 1547 01:20:17,000 --> 01:20:21,760 one off that you can't extrapolate from the time 1995 1548 01:20:21,760 --> 01:20:26,040 to 2010 that rail use is going to continue to grow. 1549 01:20:26,160 --> 01:20:29,480 What's happened is the interface via which we book rail has 1550 01:20:29,480 --> 01:20:32,520 changed. So it's made it more conducive 1551 01:20:32,520 --> 01:20:36,880 to people travelling by train and so the extent to which whole 1552 01:20:36,880 --> 01:20:41,320 businesses which think they have complete safety because 1553 01:20:41,440 --> 01:20:45,000 everybody uses them regularly and is familiar, might be 1554 01:20:45,000 --> 01:20:47,600 vulnerable to this really. Fascinating. 1555 01:20:48,080 --> 01:20:51,080 Hey, it has been a phenomenal conversation. 1556 01:20:51,080 --> 01:20:52,400 We definitely need to do a Part 2. 1557 01:20:52,400 --> 01:20:53,560 This has been. We'll do a Part 2. 1558 01:20:53,560 --> 01:20:56,360 This has. Been wonderful, fantastic. 1559 01:20:56,440 --> 01:20:59,720 But as a parting message for the audience, we'll give you the mic 1560 01:21:00,880 --> 01:21:04,840 for your book and what you've got going on. 1561 01:21:04,840 --> 01:21:07,200 It'll be great for you to tell the last word. 1562 01:21:07,200 --> 01:21:10,560 Yeah, what only to say, yeah, I've got so got this new book 1563 01:21:10,560 --> 01:21:13,840 out how to talk to AI and how not to. 1564 01:21:13,840 --> 01:21:17,040 And the how not to is probably more important than the how to 1565 01:21:17,040 --> 01:21:20,400 at the moment, but it won't always be so yeah, check that 1566 01:21:20,400 --> 01:21:21,920 out. Available everywhere including 1567 01:21:21,920 --> 01:21:27,000 Amazon and WH with travel and even Tesco. 1568 01:21:27,240 --> 01:21:30,280 That's how mainstream it's got. That book is now being sold next 1569 01:21:30,280 --> 01:21:34,080 to like 50 Shades of Grey. So, you know, this thing has 1570 01:21:34,080 --> 01:21:36,320 gone big. Like everyone's thinking about 1571 01:21:36,320 --> 01:21:39,440 AI, wondering how to talk to it. And it is about talking and 1572 01:21:39,440 --> 01:21:41,320 typing. It's, you know, it's not about 1573 01:21:41,320 --> 01:21:44,080 machines and being a coder. It's like language. 1574 01:21:44,080 --> 01:21:46,920 It's how do we communicate our ideas. 1575 01:21:47,200 --> 01:21:50,560 And so that I think means it really is quite accessible to a 1576 01:21:50,560 --> 01:21:53,440 lot of people, but people are afraid of it because it's called 1577 01:21:53,520 --> 01:21:55,880 AI. So it sounds very, very complex 1578 01:21:56,440 --> 01:21:59,040 in your use of it. It's not, and most people do 1579 01:21:59,640 --> 01:22:03,240 have quite a lot of skills here that they probably don't realise 1580 01:22:03,240 --> 01:22:05,080 they do. So yeah, check that one out. 1581 01:22:05,720 --> 01:22:06,600 Amazing. Fantastic. 1582 01:22:06,720 --> 01:22:08,640 Thank you. I'm a Scorpio by the way. 1583 01:22:09,240 --> 01:22:09,680 There you go.