1 00:00:01,760 --> 00:00:04,520 The Instructor Podcast with Terry Cook talking. 2 00:00:04,520 --> 00:00:07,560 With leaders, innovators, experts and game changers. 3 00:00:07,560 --> 00:00:09,040 About what drives them. 4 00:00:09,760 --> 00:00:11,520 Welcome to the Instructor Podcast. 5 00:00:11,520 --> 00:00:14,760 This is the show that helps you become an even more awesome driving 6 00:00:14,760 --> 00:00:17,600 instructor and run a better driving school. 7 00:00:18,160 --> 00:00:21,320 As always, I am your splendid host, Terry Cook. 8 00:00:21,320 --> 00:00:22,480 I'm delighted to be here. 9 00:00:22,480 --> 00:00:26,000 Even more delighted that you have chosen to listen because today 10 00:00:26,160 --> 00:00:30,000 I we are talking about the dangers of using AI. 11 00:00:31,040 --> 00:00:35,240 AI has many benefits and we can see lots of instructors using 12 00:00:35,240 --> 00:00:35,600 it. 13 00:00:36,000 --> 00:00:40,080 I don't think we necessarily talk about the negatives enough and 14 00:00:40,320 --> 00:00:43,360 I don't think people always realize what the dangers actually 15 00:00:43,360 --> 00:00:43,720 are. 16 00:00:43,720 --> 00:00:48,240 So today I'm joined by someone who knows this stuff, Colette Mason. 17 00:00:49,520 --> 00:00:52,480 Well, Colette has written a book on this very topic and you'll 18 00:00:52,480 --> 00:00:54,560 find a link to it in the show notes. 19 00:00:54,560 --> 00:00:58,250 But it's a fascinating discussion today and she definitely 20 00:00:58,250 --> 00:01:02,290 made me say rethink my approach on a couple of things. 21 00:01:02,290 --> 00:01:04,690 So hopefully you will do the same too. 22 00:01:04,690 --> 00:01:08,450 But just before we get stuck in, I want to point you in the direction 23 00:01:08,450 --> 00:01:10,730 of the IPP membership, I.e. 24 00:01:10,730 --> 00:01:13,410 The instructor performance and psychology membership. 25 00:01:13,650 --> 00:01:17,850 If you are looking to improve your coaching, understand what client 26 00:01:17,850 --> 00:01:21,530 side learning actually is, or just generally become a better driving 27 00:01:21,530 --> 00:01:25,310 instructor, then we have got a wealth of content to help you do 28 00:01:25,310 --> 00:01:25,990 just that. 29 00:01:26,550 --> 00:01:29,430 We've got some of the best brains in the business, some of the 30 00:01:29,430 --> 00:01:33,190 best brains outside of the business, and even my brain as well. 31 00:01:33,910 --> 00:01:38,790 So head over to patreon.com forward/inventor or use the link 32 00:01:38,790 --> 00:01:40,790 in the show notes to get signed up today. 33 00:01:40,870 --> 00:01:44,150 We've even got a one week free trial to get you started. 34 00:01:44,230 --> 00:01:46,390 But for now, enjoy this show. 35 00:01:51,880 --> 00:01:54,440 So we're now joined by the wonderful Colette Mason. 36 00:01:54,440 --> 00:01:55,640 How we doing, Colette? 37 00:01:55,800 --> 00:01:57,160 I'm very good, thank you, Terry. 38 00:01:57,160 --> 00:01:57,640 Very good. 39 00:01:57,960 --> 00:01:58,520 Excellent. 40 00:01:58,520 --> 00:02:01,440 Delighted to have you on today, but I'm guessing that most 41 00:02:01,440 --> 00:02:03,480 of my listeners won't know who you are. 42 00:02:03,560 --> 00:02:05,840 So do you want to just take a moment just to tell us a little bit 43 00:02:05,840 --> 00:02:06,360 about you? 44 00:02:06,920 --> 00:02:08,000 I can do well. 45 00:02:08,000 --> 00:02:08,840 I'm Colette Mason. 46 00:02:08,840 --> 00:02:09,240 I'm. 47 00:02:09,240 --> 00:02:10,680 And I'm a very nerdy person. 48 00:02:10,840 --> 00:02:14,680 I've had a computer since I was 12, pretty much glued to me through 49 00:02:14,680 --> 00:02:16,990 my entire life and I've loved to technology. 50 00:02:17,470 --> 00:02:20,350 I've done all sorts of different things over the years and 51 00:02:20,350 --> 00:02:23,710 helped people and I'm really interested in how people and computers 52 00:02:23,870 --> 00:02:26,110 kind of interact because I've been with people who want to hit 53 00:02:26,110 --> 00:02:29,630 computers with a hammer and hate them and people who are like 54 00:02:29,630 --> 00:02:31,390 48 hours straight coding something up. 55 00:02:31,390 --> 00:02:33,950 So I've seen all sorts of different people, and I find that 56 00:02:34,270 --> 00:02:35,710 aspect really interesting. 57 00:02:36,110 --> 00:02:36,830 And that's. 58 00:02:36,910 --> 00:02:38,430 That's kind of what I think about it. 59 00:02:39,070 --> 00:02:42,900 And today we're here to talk about AI, I suppose, primarily the 60 00:02:42,900 --> 00:02:45,140 negatives, and I'm sure we'll offer a couple of positives as well. 61 00:02:45,140 --> 00:02:47,300 But you literally wrote the book. 62 00:02:47,780 --> 00:02:48,500 Yes, I did. 63 00:02:48,500 --> 00:02:50,260 I want to rewrite it now because things have. 64 00:02:50,340 --> 00:02:52,540 I thought I'd made it Evergreen, but things have moved 65 00:02:52,540 --> 00:02:53,300 on so quickly. 66 00:02:53,300 --> 00:02:55,540 Even my stab at Evergreen didn't quite work. 67 00:02:56,340 --> 00:02:59,860 And so I wrote a lot about how people can work with their bot because 68 00:02:59,860 --> 00:03:01,980 I saw them doing lots of really crazy things that were really 69 00:03:01,980 --> 00:03:04,020 unhelpful and they were getting frustrated. 70 00:03:04,020 --> 00:03:06,740 And as somebody who loves tech, I don't like seeing somebody 71 00:03:06,740 --> 00:03:07,060 struggle. 72 00:03:07,060 --> 00:03:09,620 Like, I wouldn't like seeing a pensioner struggle with heavy groceries. 73 00:03:09,620 --> 00:03:10,820 I want to step in and help. 74 00:03:11,430 --> 00:03:13,590 So I thought this book will help with that. 75 00:03:13,590 --> 00:03:16,390 But since then, AIs changed quite a lot. 76 00:03:16,790 --> 00:03:19,990 And it's gone from being quite useful, a bit like an Excel formula, 77 00:03:20,150 --> 00:03:23,830 into something that's got people with placards protesting the 78 00:03:23,830 --> 00:03:25,270 government in America, for instance. 79 00:03:25,270 --> 00:03:25,830 So it's. 80 00:03:25,830 --> 00:03:29,630 Whoa, that escalated quickly and it's been quite a wild ride since 81 00:03:29,630 --> 00:03:30,310 I wrote it. 82 00:03:30,470 --> 00:03:33,910 Yeah, well, I will share a link to it in the show notes anyway. 83 00:03:33,910 --> 00:03:36,790 But I will just say that I bought that book. 84 00:03:36,790 --> 00:03:37,550 I used that book. 85 00:03:37,550 --> 00:03:39,590 I found it good for two things. 86 00:03:39,910 --> 00:03:43,150 So firstly, just a little bit of self praise for a second as I 87 00:03:43,150 --> 00:03:44,310 went from like, oh, I do that. 88 00:03:44,310 --> 00:03:44,950 Oh, I do that. 89 00:03:44,950 --> 00:03:45,590 There were a couple of things. 90 00:03:45,590 --> 00:03:47,230 I'm like, oh, I'm doing something great. 91 00:03:47,310 --> 00:03:50,150 But also it just taught me a different way to look at it. 92 00:03:50,150 --> 00:03:54,030 It's like you train it so that it works for you rather than just 93 00:03:54,830 --> 00:03:58,270 dive in and hope for the best, which maybe someone will speak about. 94 00:03:58,270 --> 00:03:58,590 6. 95 00:03:59,070 --> 00:04:01,910 I was kind of thinking about to start this, this episode, because 96 00:04:01,910 --> 00:04:04,350 as I say, we're going to be talking about a lot of the negatives 97 00:04:04,350 --> 00:04:08,420 of AI, because I just see it everywhere as if it's the fix all 98 00:04:08,420 --> 00:04:08,980 for everything. 99 00:04:09,780 --> 00:04:15,940 And a phrase that I hear a lot is the term AI Slop. 100 00:04:16,420 --> 00:04:18,660 What do we mean by AI Slop? 101 00:04:19,780 --> 00:04:24,780 AI Slop comes about because people are using AI the tool, and 102 00:04:24,780 --> 00:04:28,260 they've got wildly unrealistic expectations about what it can do. 103 00:04:28,660 --> 00:04:32,100 It generates something and then they wave it through and give 104 00:04:32,100 --> 00:04:33,060 it to somebody else. 105 00:04:33,140 --> 00:04:34,020 And it's not. 106 00:04:34,300 --> 00:04:35,180 It's not accurate. 107 00:04:35,340 --> 00:04:38,700 It's not like a calculator where you say, you know, two times 108 00:04:38,700 --> 00:04:41,860 two equals, and you always get four, you might get three plus one 109 00:04:41,860 --> 00:04:44,700 with AI or one less than five or four. 110 00:04:45,100 --> 00:04:48,580 So you get different answers every time, every time you do it 111 00:04:48,580 --> 00:04:52,500 and people don't really check how useful that answer is and then 112 00:04:52,500 --> 00:04:55,020 they'll give it to somebody else or give it to their machine 113 00:04:55,420 --> 00:04:59,820 and it starts to snowball into this horrific mess. 114 00:05:00,560 --> 00:05:03,920 And that's kind of the problem we have to think about. 115 00:05:03,920 --> 00:05:05,640 Well, how much slop is tolerable. 116 00:05:05,640 --> 00:05:07,800 If it's an early draft of something, you don't mind a bit of 117 00:05:07,800 --> 00:05:11,400 slop, but if it's a TED Talk and you had slop in your TED Talk, 118 00:05:11,400 --> 00:05:12,520 then you'd care quite a lot. 119 00:05:12,520 --> 00:05:15,560 So you're having to decide, you know, all these different things 120 00:05:15,560 --> 00:05:17,480 about the quality of information which we didn't have 121 00:05:17,480 --> 00:05:18,080 to do before. 122 00:05:18,480 --> 00:05:21,720 And it's a lot of unpaid labour and I think people get a bit 123 00:05:21,720 --> 00:05:24,880 resentful of the unpaid labor or just it getting it wrong all the 124 00:05:24,880 --> 00:05:25,000 time. 125 00:05:25,000 --> 00:05:27,130 It's like Bart Simpson and home and you. 126 00:05:27,130 --> 00:05:27,930 Why you little. 127 00:05:28,170 --> 00:05:29,650 And it's like, why did you do it like that? 128 00:05:29,650 --> 00:05:32,570 When I gave you all these PDFs, it was obvious, wasn't it? 129 00:05:32,650 --> 00:05:35,730 And you didn't do it that way and people get a bit frustrated. 130 00:05:35,730 --> 00:05:38,690 So there's a couple of things going on and I think, you know, in 131 00:05:38,690 --> 00:05:42,930 terms of like Google search now, if you get that horrific curated 132 00:05:42,930 --> 00:05:47,810 search thing coming up, which is just made up nonsense and it's 133 00:05:47,810 --> 00:05:49,250 kind of ruined something that was really good. 134 00:05:49,250 --> 00:05:51,850 Google search used to be really good and now it's this sort 135 00:05:51,850 --> 00:05:53,330 of curated mess. 136 00:05:53,410 --> 00:05:57,410 And you know, all these things munge together and it makes people 137 00:05:57,410 --> 00:05:59,170 dislike AI quite a lot. 138 00:06:00,450 --> 00:06:04,370 I think that's probably my biggest issue of it, if I'm honest, 139 00:06:04,370 --> 00:06:05,810 is the inaccurate information. 140 00:06:06,530 --> 00:06:07,250 Although I've got fair. 141 00:06:07,250 --> 00:06:09,010 I'm going to say that's my biggest problem with it for lots 142 00:06:09,010 --> 00:06:09,650 of things today. 143 00:06:09,650 --> 00:06:11,170 But it's that inaccurate information. 144 00:06:11,410 --> 00:06:14,450 And just the other day I was very lazy. 145 00:06:14,610 --> 00:06:16,170 All my ends up feeling very lazy. 146 00:06:16,170 --> 00:06:19,370 Normally I would read a book or actually search properly, but 147 00:06:19,370 --> 00:06:20,010 I was lazy. 148 00:06:20,010 --> 00:06:22,330 I think I actually went on Copilot and typed it in there to 149 00:06:22,330 --> 00:06:24,810 give me a list of all the monarchs, the English monarchs. 150 00:06:25,130 --> 00:06:27,210 And when I was scanning through I realized that I think it 151 00:06:27,210 --> 00:06:28,650 was Queen Mary that had missed off. 152 00:06:28,970 --> 00:06:31,170 And I'm like, well, if it's missed off Queen Mary, is there any 153 00:06:31,170 --> 00:06:32,250 others that I didn't notice? 154 00:06:32,250 --> 00:06:34,810 And I think there were two more that I wouldn't have spotted. 155 00:06:35,130 --> 00:06:36,970 So if I'd have been using that as a. 156 00:06:36,970 --> 00:06:37,930 As A resource. 157 00:06:38,810 --> 00:06:41,210 There was two there that I wouldn't have known were wrong. 158 00:06:41,610 --> 00:06:45,450 So I'm curious, why doesn't it get a lot of stuff correct? 159 00:06:45,450 --> 00:06:46,810 Because it seems to get a lot wrong. 160 00:06:46,810 --> 00:06:48,810 Why is that incorrect come from? 161 00:06:49,410 --> 00:06:52,810 Well, it's had lots of different training sources, so somebody 162 00:06:52,810 --> 00:06:54,890 who wrote it might have made a mistake. 163 00:06:54,890 --> 00:06:58,570 It gets lots of information from Reddit in particular and Wikipedia. 164 00:06:58,570 --> 00:07:01,650 So Wikipedia is going to be better quality than Reddit, but there's 165 00:07:01,650 --> 00:07:02,930 going to be mistakes in there. 166 00:07:03,090 --> 00:07:04,770 It can also be overloaded. 167 00:07:04,770 --> 00:07:09,050 If I asked you 40 questions, like mastermind, if I ask you loads 168 00:07:09,050 --> 00:07:11,650 of questions in a minute and a half, you're going to get a few wrong 169 00:07:11,650 --> 00:07:12,570 because there's a lot. 170 00:07:12,570 --> 00:07:16,070 You having to process a lot of things and it just, it just drops 171 00:07:16,070 --> 00:07:17,190 the ball occasionally with its. 172 00:07:17,190 --> 00:07:18,750 What's called its context window. 173 00:07:18,750 --> 00:07:19,670 It just can't. 174 00:07:19,670 --> 00:07:22,150 There's just too much being asked of it at once and it starts 175 00:07:22,150 --> 00:07:23,990 to drop the ball like a human would. 176 00:07:24,230 --> 00:07:27,910 So that's a couple of places it makes assumptions, I've noticed 177 00:07:27,910 --> 00:07:28,230 recently. 178 00:07:28,550 --> 00:07:32,350 It's filtering what it tells you, you know, so obviously I can't 179 00:07:32,350 --> 00:07:34,790 ask it how do I make some sort of chemical weapon? 180 00:07:34,950 --> 00:07:37,550 It's going to say no, but it's saying no to more and more things 181 00:07:37,550 --> 00:07:42,210 and filtering bits out that might be dangerous and that's causing 182 00:07:42,210 --> 00:07:43,490 some missing parts in it as well. 183 00:07:43,490 --> 00:07:46,170 And it's just, it's just because it's predictive. 184 00:07:46,170 --> 00:07:48,810 It's just never built to be really robust. 185 00:07:48,810 --> 00:07:51,450 It's not like an Excel formula that adds up those columns. 186 00:07:51,450 --> 00:07:51,810 It's. 187 00:07:51,810 --> 00:07:52,930 It's just freer than that. 188 00:07:52,930 --> 00:07:55,530 And it is going to make mistakes like a human might. 189 00:07:56,570 --> 00:07:59,490 So I want to give you some specific examples in a minute, I 190 00:07:59,490 --> 00:07:59,690 think. 191 00:07:59,690 --> 00:08:02,290 But before we dive into that, I kind of ask you one of the. 192 00:08:02,290 --> 00:08:05,650 I want to ask you the broad question of what are some of the 193 00:08:05,650 --> 00:08:07,810 problems that you're seeing with some of the biggest problems 194 00:08:07,810 --> 00:08:08,250 we're facing. 195 00:08:08,250 --> 00:08:11,200 Do you think I'm going to have a pop at marketers? 196 00:08:11,200 --> 00:08:12,000 I mean, I've been having. 197 00:08:12,080 --> 00:08:15,640 I used to own marketing wanker.com I think back in the day, 198 00:08:15,640 --> 00:08:17,880 because I just got really annoyed one day and I registered 199 00:08:17,880 --> 00:08:18,160 it. 200 00:08:19,360 --> 00:08:21,800 So marketers sort of saying it's going to do all these things, 201 00:08:21,800 --> 00:08:25,159 click a button and replace all that boring work and it'll be brilliant 202 00:08:25,159 --> 00:08:28,120 and you can have pina colada and no, it doesn't work like that. 203 00:08:28,120 --> 00:08:29,880 We've all found out it doesn't work like that. 204 00:08:29,880 --> 00:08:32,200 It's like, well, you did half of it and then you screwed up the 205 00:08:32,200 --> 00:08:32,640 other half. 206 00:08:33,040 --> 00:08:35,950 So I think marketers tell IT people that can do things that it 207 00:08:35,950 --> 00:08:38,270 can't and because if you're not very good at computers, you can't 208 00:08:38,270 --> 00:08:42,070 smell the bullshit from the how it really works and you're desperate 209 00:08:42,070 --> 00:08:42,830 to find an answer. 210 00:08:42,830 --> 00:08:45,110 You're working on your own in your own business and you're doing 211 00:08:45,110 --> 00:08:46,950 your accounts, you're doing your social media, you've got to 212 00:08:46,950 --> 00:08:48,550 see customers, you got to do invoices. 213 00:08:49,270 --> 00:08:52,430 You just hope this thing is going to rescue you and it and it 214 00:08:52,430 --> 00:08:52,870 doesn't. 215 00:08:53,110 --> 00:08:56,710 And the other problem is a big problem I see is like big corporations. 216 00:08:56,790 --> 00:09:00,430 If you're a manager or a CEO of a company, you're sort of saying 217 00:09:00,430 --> 00:09:02,830 to it, should I open an office in Derbyshire? 218 00:09:02,910 --> 00:09:04,630 How can I increase my market share? 219 00:09:04,630 --> 00:09:06,590 You're asking these nice open ended questions. 220 00:09:06,590 --> 00:09:09,430 It gives you a bunch of answers and you pick the one that 221 00:09:09,430 --> 00:09:09,950 feels the best. 222 00:09:09,950 --> 00:09:11,150 And you've got a lot of freedom. 223 00:09:11,150 --> 00:09:14,310 You're using your own brain with some suggestions to come up 224 00:09:14,310 --> 00:09:16,470 with the best outcome and that's great. 225 00:09:16,470 --> 00:09:19,430 But if you're writing software, you haven't got that luxury. 226 00:09:19,430 --> 00:09:22,230 What if your two options are something that's got a security hole 227 00:09:22,230 --> 00:09:26,070 in it and you have a Jaguar Land Rover instance and your company 228 00:09:26,070 --> 00:09:26,630 goes down? 229 00:09:26,630 --> 00:09:28,150 Or does it actually write good code? 230 00:09:28,150 --> 00:09:32,550 The consequences for more specific tasks are very big and I 231 00:09:32,550 --> 00:09:36,030 don't think people at board level understand those problems. 232 00:09:36,190 --> 00:09:38,150 They just say use as much AI as you can. 233 00:09:38,150 --> 00:09:41,030 If we use tons of tokens and we will, you know, that's the way 234 00:09:41,030 --> 00:09:41,550 to do it. 235 00:09:41,870 --> 00:09:42,590 Half a million. 236 00:09:43,550 --> 00:09:44,630 Half a million dollars. 237 00:09:44,630 --> 00:09:47,230 Jensen Huang said spend half a million dollars on tokens and we'll 238 00:09:47,230 --> 00:09:48,910 be ahead for Nvidia. 239 00:09:48,990 --> 00:09:51,390 And that's quite a lot of money and nobody's really seeing 240 00:09:51,390 --> 00:09:53,030 the results of that. 241 00:09:53,030 --> 00:09:55,750 One company forgot to say, please cut my spend for the month 242 00:09:55,750 --> 00:09:58,990 and spent $500 million in a month on AI. 243 00:09:59,210 --> 00:10:02,530 So if you've wasted your 20 quid subscription and ended up benched 244 00:10:02,530 --> 00:10:04,930 for a couple of hours and you went for a walk, then don't beat 245 00:10:04,930 --> 00:10:06,810 yourself up, you're actually doing quite well. 246 00:10:07,290 --> 00:10:10,170 So I think that's these unrealistic expectations is the big 247 00:10:10,170 --> 00:10:10,490 problem. 248 00:10:10,570 --> 00:10:13,770 And I think there's also some things are getting caught up in the 249 00:10:13,770 --> 00:10:14,330 crossfire. 250 00:10:14,330 --> 00:10:16,930 As soon as people see AI, the red mist comes down. 251 00:10:16,930 --> 00:10:17,970 There's an M dash in there. 252 00:10:17,970 --> 00:10:18,650 Burn him. 253 00:10:18,810 --> 00:10:21,450 You know, and that's not very helpful. 254 00:10:21,450 --> 00:10:22,810 What if English isn't your first? 255 00:10:22,810 --> 00:10:24,410 You know, English isn't your first language. 256 00:10:24,850 --> 00:10:25,250 And you need. 257 00:10:25,250 --> 00:10:28,210 You've used it to communicate or you've got dyslexia and you're 258 00:10:28,210 --> 00:10:30,530 sick of people giving you a hard time for your spelling. 259 00:10:30,530 --> 00:10:32,250 And you think, right, I'm going to put it through there because 260 00:10:32,250 --> 00:10:34,130 it just gives my brain a rest a day. 261 00:10:34,530 --> 00:10:35,850 And I think people get that. 262 00:10:35,850 --> 00:10:36,770 That's another problem. 263 00:10:37,090 --> 00:10:38,330 Some people use it and it's. 264 00:10:38,330 --> 00:10:40,210 It was an enabler for them. 265 00:10:40,450 --> 00:10:42,690 And the other thing is it actually does do some good things, 266 00:10:42,690 --> 00:10:44,370 like looking at crops in Africa. 267 00:10:44,450 --> 00:10:47,610 There's one crop a bit like potatoes back in the day in Ireland. 268 00:10:47,610 --> 00:10:49,250 It feeds two out of five people. 269 00:10:49,890 --> 00:10:51,970 And if it gets diseased, that's a problem. 270 00:10:52,370 --> 00:10:54,930 But it's difficult to work out how you cope with that disease. 271 00:10:55,790 --> 00:10:57,310 So the farmer takes a photo, though. 272 00:10:57,310 --> 00:10:59,470 It gets sent to somebody who understands these things. 273 00:10:59,470 --> 00:11:03,590 The AI looks at the image, the expert in agriculture looks at it 274 00:11:03,590 --> 00:11:05,390 and then tells the farmer what to do. 275 00:11:05,550 --> 00:11:07,550 Liver transplants is another example. 276 00:11:07,950 --> 00:11:11,710 So if your liver's near its use by date, when it's been taken 277 00:11:11,710 --> 00:11:14,590 out of a body, AI can say if it's still safe to use. 278 00:11:14,590 --> 00:11:18,790 And a surgeon can think, I want to use that liver and that can 279 00:11:18,790 --> 00:11:19,870 save somebody's life. 280 00:11:20,430 --> 00:11:22,510 And I don't think you, you know, we've got to make sure that 281 00:11:22,510 --> 00:11:25,630 we don't throw the baby out with the bathwater, with the hatred. 282 00:11:26,190 --> 00:11:27,110 Use it response. 283 00:11:27,110 --> 00:11:29,030 It's like using a drill, you know, you don't drill through your 284 00:11:29,030 --> 00:11:32,990 tv, you know, just use it, use it sensibly to make holes in stuff 285 00:11:33,150 --> 00:11:36,430 and it's fine, you know, and don't, don't do what the marketers 286 00:11:36,430 --> 00:11:39,230 tell you and suck all your staff like Ford did, and then have 287 00:11:39,230 --> 00:11:41,270 to hire them back with your cap, you know, tail between your 288 00:11:41,270 --> 00:11:45,670 legs, you know, be responsible like you do in any power tool, you 289 00:11:45,670 --> 00:11:46,670 know, that's what I think. 290 00:11:47,400 --> 00:11:48,600 And that's really interesting. 291 00:11:48,760 --> 00:11:51,280 Like the medical example you gave there as well, with the liver. 292 00:11:51,280 --> 00:11:54,960 It's like, I've read on that a little bit and tell me if I'm wrong 293 00:11:54,960 --> 00:11:57,160 with this, but the kind of principle is that AI is going to 294 00:11:57,160 --> 00:12:00,560 get it right more times than a human would, potentially get it right 295 00:12:00,560 --> 00:12:02,360 quicker, more than a human would. 296 00:12:02,520 --> 00:12:05,080 But then you've also got the human there as backup as well as, 297 00:12:05,080 --> 00:12:05,960 like a second opinion. 298 00:12:05,960 --> 00:12:06,520 Is that right? 299 00:12:06,840 --> 00:12:08,320 Yeah, I think that's the best way to. 300 00:12:08,320 --> 00:12:11,720 I think it's augmented intelligence is the way to get the 301 00:12:11,720 --> 00:12:15,080 best result rather than artificial intelligence, you know, 302 00:12:15,080 --> 00:12:17,440 because you still need the human in there to check what the 303 00:12:17,440 --> 00:12:18,440 blooming thing is doing. 304 00:12:18,680 --> 00:12:21,080 But equally the AI can have your back. 305 00:12:21,080 --> 00:12:23,520 It can look through a big, you know, two foot high stack of data 306 00:12:23,520 --> 00:12:27,320 that you just can't do and that's, that's its strength and it 307 00:12:27,320 --> 00:12:29,240 can really help in those circumstances. 308 00:12:29,560 --> 00:12:30,120 Yeah. 309 00:12:30,600 --> 00:12:32,720 Interestingly, you mentioned about the language there as well. 310 00:12:32,720 --> 00:12:34,800 So I've got to put my own personal story now, which is my, 311 00:12:34,800 --> 00:12:36,160 my partner's Moldovan. 312 00:12:36,160 --> 00:12:37,560 She speaks really good English. 313 00:12:37,800 --> 00:12:40,080 But now and again there's somewhat that I'll, I'll text her, 314 00:12:40,080 --> 00:12:42,390 for example, she ain't got a clue because I've made my Yorkshire 315 00:12:42,390 --> 00:12:45,670 dialots come out and she'll go on chatgpt and what does this mean? 316 00:12:45,750 --> 00:12:47,470 And I'm like, I don't even know what it means now. 317 00:12:47,470 --> 00:12:47,670 Chat. 318 00:12:47,670 --> 00:12:49,150 BT said it, I'm like, oh yeah, it does mean. 319 00:12:49,150 --> 00:12:49,790 That makes sense. 320 00:12:49,790 --> 00:12:50,230 Yeah. 321 00:12:50,310 --> 00:12:53,750 So it definitely does have its uses, but I want to kind of come 322 00:12:53,750 --> 00:12:58,710 onto a bit more, let's say our level, sort of a, more a local level, 323 00:12:58,710 --> 00:12:59,310 if you like. 324 00:12:59,310 --> 00:13:04,550 So I think instructors tend to use it for three general things. 325 00:13:05,110 --> 00:13:07,790 Obviously we can use it for all sorts, but the three primary 326 00:13:07,790 --> 00:13:11,010 things, we've got the, the content creation aspect, we've got 327 00:13:11,010 --> 00:13:14,850 the training themselves and using it to, to train others. 328 00:13:15,330 --> 00:13:17,290 So I thought I might run through a few of those, if that's 329 00:13:17,290 --> 00:13:17,730 okay. 330 00:13:17,970 --> 00:13:20,050 Yeah, certainly, if, if you look at. 331 00:13:20,050 --> 00:13:23,530 The content creation, I think again, we can use it for all sorts. 332 00:13:23,530 --> 00:13:26,090 But I'm thinking of three primary ones that I see instructors 333 00:13:26,090 --> 00:13:26,690 using it for. 334 00:13:26,930 --> 00:13:30,770 One is like the social media posts, the, the text of the copy 335 00:13:30,770 --> 00:13:31,810 on social media posts. 336 00:13:32,050 --> 00:13:34,850 Then we've got images that go alongside that social media post 337 00:13:34,850 --> 00:13:37,730 and make all things like emails and blogs and stuff like that. 338 00:13:38,410 --> 00:13:42,090 So I suppose again, broad question, is there a danger using 339 00:13:42,090 --> 00:13:44,170 AI for that and if so, what is that danger? 340 00:13:44,730 --> 00:13:47,930 I think using it for content creation is fairly safe. 341 00:13:47,930 --> 00:13:50,650 You know, you've got the highway code you can check against, 342 00:13:50,730 --> 00:13:54,570 that's a good source of truth and you've got newspaper articles 343 00:13:54,570 --> 00:13:57,610 and you've probably built your avatar to make sure you created something 344 00:13:57,610 --> 00:13:59,810 for the young learner drivers to. 345 00:13:59,810 --> 00:14:03,610 Not an older person, you know, so you've got some quite good guardrails 346 00:14:03,610 --> 00:14:04,410 there to make that work. 347 00:14:04,410 --> 00:14:04,730 Okay. 348 00:14:04,730 --> 00:14:05,170 And it's. 349 00:14:05,170 --> 00:14:07,610 As long as it's safe information telling people to drive 350 00:14:07,610 --> 00:14:10,120 on the correct side of the road and stuff, then that's, that's 351 00:14:10,120 --> 00:14:10,640 pretty all right. 352 00:14:10,640 --> 00:14:14,160 I don't see a problem with that, with things like emails. 353 00:14:14,320 --> 00:14:16,840 I see a little bit more of a problem in that. 354 00:14:16,840 --> 00:14:20,560 You've got to give Google access to your emails to be able 355 00:14:20,560 --> 00:14:24,240 to filter them, to say, oh, this is a new customer inquiry and 356 00:14:24,240 --> 00:14:26,880 that's somebody's birthday and, or they passed their test, I 357 00:14:26,880 --> 00:14:29,040 must say congratulations and send them an email. 358 00:14:29,360 --> 00:14:31,720 So when it's monitoring your emails, of course it's got to read 359 00:14:31,720 --> 00:14:35,120 all of your emails to make a decision and it's got to read the 360 00:14:35,120 --> 00:14:37,000 email that came in, so that's somebody else. 361 00:14:37,000 --> 00:14:39,380 And they probably, they may have not given you permission to 362 00:14:39,380 --> 00:14:43,220 share your stuff wildly and freely with Google because, you know, 363 00:14:43,220 --> 00:14:45,620 we don't really need, I mean, they say we won't use it for training 364 00:14:45,620 --> 00:14:49,380 the models, but the most valuable resource right now is real 365 00:14:49,380 --> 00:14:52,900 human information that hasn't been made by AI. 366 00:14:52,980 --> 00:14:55,660 So I'm pretty sure they might have a good poke around there and 367 00:14:55,660 --> 00:14:56,620 just not tell anyone. 368 00:14:56,620 --> 00:14:58,020 That's my speculation. 369 00:14:58,660 --> 00:15:01,780 So there's, the more you get AI to help you with your tasks, the 370 00:15:01,780 --> 00:15:04,430 more it can spy on you in general. 371 00:15:04,430 --> 00:15:05,150 That's not good. 372 00:15:05,150 --> 00:15:07,910 It's starting to look at the speed you travel in your car using 373 00:15:07,910 --> 00:15:10,750 Google Maps and, you know, all sorts of things go on. 374 00:15:10,990 --> 00:15:14,750 So there's a kind of a surveillance element to using AI 375 00:15:14,750 --> 00:15:17,470 that people should be, should be aware of, I think. 376 00:15:17,470 --> 00:15:19,950 And what was the third, the third one you asked me about? 377 00:15:20,830 --> 00:15:21,350 Blogs. 378 00:15:21,350 --> 00:15:22,110 Blog posts. 379 00:15:22,270 --> 00:15:23,550 Blog posts, yeah. 380 00:15:23,550 --> 00:15:26,950 There's not, I think the problem there could be because there's 381 00:15:26,950 --> 00:15:28,310 so much backlash with AI. 382 00:15:28,310 --> 00:15:32,600 If it doesn't sound like you, then it looks lazy and if it looks 383 00:15:32,600 --> 00:15:35,640 lazy then that you want somebody who's vigilant sat next 384 00:15:35,640 --> 00:15:36,840 to you in the driving seat. 385 00:15:36,840 --> 00:15:39,640 When you're learning and there's other drivers out on the 386 00:15:39,640 --> 00:15:42,520 road, you don't, you know, you want to make sure that the instructor's 387 00:15:42,520 --> 00:15:45,520 looking after the person properly and paying attention and 388 00:15:45,760 --> 00:15:47,520 it kind of suggests that you don't care. 389 00:15:47,600 --> 00:15:49,760 And that's, that's the big problem with it, I think. 390 00:15:49,760 --> 00:15:52,560 And do we really need to read another beige thing about. 391 00:15:52,560 --> 00:15:54,920 Crikey, it's such a long wait to get your test booked. 392 00:15:54,920 --> 00:15:57,200 You know, do we really need to read another post on that? 393 00:15:58,180 --> 00:15:58,660 Probably not. 394 00:15:58,740 --> 00:16:00,020 Fake pictures of cars. 395 00:16:00,020 --> 00:16:01,740 You know, what's the. 396 00:16:01,740 --> 00:16:04,900 I'd rather see you, the driving instructor stood with a happy 397 00:16:04,900 --> 00:16:09,660 person who's just passed their test than some sort of Disney car 398 00:16:09,660 --> 00:16:09,980 thing. 399 00:16:09,980 --> 00:16:13,579 Whatever you've done to be a thumb scroller, you know, don't think 400 00:16:13,579 --> 00:16:14,340 that's in your interest. 401 00:16:14,740 --> 00:16:15,460 You're quite lucky. 402 00:16:15,460 --> 00:16:20,100 It's a personal service and being personal now has got Real cachet 403 00:16:20,100 --> 00:16:21,300 and commercial power. 404 00:16:21,380 --> 00:16:25,300 Because people, you know, they keep saying I miss real connections 405 00:16:25,300 --> 00:16:25,700 in my life. 406 00:16:25,700 --> 00:16:28,260 I want to join a walking club or I want to bring my friends around 407 00:16:28,260 --> 00:16:28,900 for a meal. 408 00:16:29,280 --> 00:16:31,480 Personal is actually quite a good strength now. 409 00:16:31,480 --> 00:16:33,920 And commercially you shouldn't get rid of that, I don't think. 410 00:16:34,560 --> 00:16:37,280 Well, you mentioned like the surveillance aspect there and I think 411 00:16:37,280 --> 00:16:40,080 I'd like to come back to that because I know that's a bit of a 412 00:16:40,080 --> 00:16:44,320 bugberry odds a minute but yes, just before we do, you said 413 00:16:44,320 --> 00:16:46,000 somewhat at the start of that. 414 00:16:46,080 --> 00:16:48,960 It felt like you glossed over a little bit and it was talking about 415 00:16:49,040 --> 00:16:53,360 like the guardrails on stuff and fact checking and I think that's 416 00:16:53,440 --> 00:16:56,850 again, I'm, I think everyone does this, but I'm always talking 417 00:16:56,850 --> 00:16:59,410 about our industry specifically with this that so many 418 00:16:59,410 --> 00:17:00,170 people don't. 419 00:17:00,810 --> 00:17:03,890 You know, I'm seeing pictures up there that have cars on the wrong 420 00:17:03,890 --> 00:17:06,490 side of the road or they're going around the roundabout the wrong 421 00:17:06,490 --> 00:17:06,610 way. 422 00:17:06,610 --> 00:17:09,370 It's just, it's just false information, it's incorrect. 423 00:17:09,530 --> 00:17:13,650 And I'm seeing posts written that, you know, I can tell that generated 424 00:17:13,650 --> 00:17:16,850 by AI and I'm seeing the people around say, you know, subscribe 425 00:17:16,850 --> 00:17:17,370 to my page. 426 00:17:17,370 --> 00:17:19,770 This takes loads of time for me to correct my dozen. 427 00:17:19,770 --> 00:17:22,130 You've asked them to write your force, it's taken 30 seconds, 428 00:17:22,130 --> 00:17:22,970 you've not checked it. 429 00:17:23,200 --> 00:17:26,280 And there's information there that could genuinely cause a problem 430 00:17:26,280 --> 00:17:27,280 to someone on the road. 431 00:17:27,440 --> 00:17:31,560 And as driving instructors are most of us, our primary goal is to 432 00:17:31,560 --> 00:17:33,680 try and reduce the number of people that die on the road. 433 00:17:34,000 --> 00:17:37,320 But some of this stuff is actually running the risk of increasing 434 00:17:37,320 --> 00:17:37,680 it. 435 00:17:38,000 --> 00:17:42,440 So that fact checking those guardrails, is that something that's 436 00:17:42,440 --> 00:17:43,600 missing a lot, do you think? 437 00:17:44,320 --> 00:17:49,520 Yeah, and it takes a lot of work to police, to police it so that 438 00:17:50,010 --> 00:17:51,250 it might get a bit technical now. 439 00:17:51,250 --> 00:17:54,970 But one, a good way to do it is to kind of use actual old school 440 00:17:54,970 --> 00:17:57,490 code with brackets and variables and all that stuff to kind 441 00:17:57,490 --> 00:17:58,410 of control the process. 442 00:17:58,490 --> 00:18:01,370 And then AI goes off and does a little portion of it, it comes 443 00:18:01,370 --> 00:18:04,370 back and that's fact checked because it's only a small portion 444 00:18:04,370 --> 00:18:06,330 then and not a 2000 word post. 445 00:18:06,890 --> 00:18:09,170 You check the little bit and then you give it the next thing to 446 00:18:09,170 --> 00:18:13,850 do and you build it up in stages and put lots of QA and they 447 00:18:13,850 --> 00:18:17,290 call it rag, which is where one machine tests what another one's 448 00:18:17,290 --> 00:18:17,530 done. 449 00:18:17,810 --> 00:18:20,210 But that takes a look, that takes a lot of effort and it's only 450 00:18:20,210 --> 00:18:23,570 really worth doing if you're doing 10,000 articles a year. 451 00:18:23,570 --> 00:18:26,170 Let's say you're probably better off just going slow and doing 452 00:18:26,170 --> 00:18:26,690 it properly. 453 00:18:26,690 --> 00:18:27,490 Especially if you're not. 454 00:18:27,810 --> 00:18:30,690 Unless you've got an interest in it academically, intellectually 455 00:18:30,690 --> 00:18:33,050 to build this thing, then there isn't, there isn't really much 456 00:18:33,050 --> 00:18:33,250 point. 457 00:18:33,250 --> 00:18:33,530 You bet. 458 00:18:33,530 --> 00:18:37,690 It's just dictating because you can dictate at 120 words a minute, 459 00:18:37,690 --> 00:18:41,890 type at between 40 to 80 depending on if you're a two finger 460 00:18:41,890 --> 00:18:42,690 typer or not. 461 00:18:43,010 --> 00:18:45,650 You might as well just dictate it and transcribe it to get the speed 462 00:18:45,650 --> 00:18:48,520 increase and then you've got the accuracy and a bit of the grind 463 00:18:48,520 --> 00:18:49,040 is gone. 464 00:18:49,040 --> 00:18:52,120 And you can be confident that what you've made is useful. 465 00:18:52,200 --> 00:18:54,760 So you don't always have to do it that way. 466 00:18:54,760 --> 00:18:58,720 But you could build up a bank of principles that you must, you 467 00:18:58,720 --> 00:18:59,720 must never say this. 468 00:18:59,720 --> 00:19:03,360 I like it when you use that, use this as a model example and you 469 00:19:03,360 --> 00:19:06,560 can give it those sort of give it some more things to aim for because 470 00:19:06,560 --> 00:19:07,760 it's always trying to please you. 471 00:19:07,760 --> 00:19:10,080 So if you said if you made it look like this, then I'd be really 472 00:19:10,080 --> 00:19:10,760 quite happy. 473 00:19:11,080 --> 00:19:15,320 It helps it do what you do what you want to a certain extent. 474 00:19:15,560 --> 00:19:18,560 So that would be a way of doing or getting the output from 475 00:19:18,560 --> 00:19:21,160 one, putting it in another because it allows slightly different 476 00:19:21,320 --> 00:19:21,800 training. 477 00:19:22,280 --> 00:19:24,280 Gemini is a good one because it gets a lot of. 478 00:19:24,920 --> 00:19:27,040 To me when I use it it feels a bit more up to date. 479 00:19:27,040 --> 00:19:29,080 So that's kind of helpful. 480 00:19:29,080 --> 00:19:30,440 So sometimes I put things in there. 481 00:19:30,440 --> 00:19:33,240 But everything you put into Gemini gets used as training. 482 00:19:33,320 --> 00:19:36,560 Unless you're on the enterprise plan, not the paid £10amonth 483 00:19:36,560 --> 00:19:38,200 plan, the hugely expensive plan. 484 00:19:38,440 --> 00:19:41,480 And so you're kind of training the beast when you use Gemini. 485 00:19:41,820 --> 00:19:44,300 But you know, but going between different models and saying 486 00:19:44,300 --> 00:19:45,700 that one made a really bad job of it. 487 00:19:45,700 --> 00:19:46,700 How would you correct this? 488 00:19:47,020 --> 00:19:50,700 It's ego kicks in and it goes well here at Anthropic we'd give 489 00:19:50,700 --> 00:19:54,380 you this blog post and so that's a nice easy way of improving 490 00:19:54,380 --> 00:19:57,500 it if that was what you were, what you were doing or I try and 491 00:19:57,900 --> 00:20:00,820 my history teacher, when I did a level history said put in just 492 00:20:00,820 --> 00:20:02,820 enough facts that it looks like you know what you're talking 493 00:20:02,820 --> 00:20:04,980 about and keep the rest principle based. 494 00:20:04,980 --> 00:20:07,260 And then it takes a lot less work. 495 00:20:07,500 --> 00:20:12,020 So if you said there is a principle that Rather than in 2017, 496 00:20:12,100 --> 00:20:16,820 a scientist called Fred found out that you've got lots more facts 497 00:20:16,820 --> 00:20:19,420 in there that don't necessarily support what you're Trying 498 00:20:19,420 --> 00:20:19,860 to teach. 499 00:20:19,860 --> 00:20:23,460 So putting in fewer facts is, is helpful, but as long as people 500 00:20:23,620 --> 00:20:26,060 learning the principles, the important part, then, then put that 501 00:20:26,060 --> 00:20:26,300 in. 502 00:20:26,300 --> 00:20:29,700 So that's another, another way of cutting down on fact. 503 00:20:29,700 --> 00:20:33,420 Hell, but even then, you know, there's lots of white papers that 504 00:20:33,420 --> 00:20:37,100 end up on, I can't pronounce the name of the website, X Archive, 505 00:20:37,100 --> 00:20:37,740 something like that. 506 00:20:37,820 --> 00:20:40,660 And a lot of them haven't been peer reviewed, but they look very 507 00:20:40,660 --> 00:20:44,460 credible, so that might get used, cited in another document. 508 00:20:44,460 --> 00:20:46,820 So even if you were very careful and looked at somebody, a 509 00:20:46,820 --> 00:20:49,300 professor from Oxford, citations to make sure your stuff 510 00:20:49,300 --> 00:20:52,940 was correct, he might have not been paying attention one day and 511 00:20:52,940 --> 00:20:55,100 used a paper that was, had fake bits in it. 512 00:20:55,100 --> 00:20:58,980 So there's only so much fact checking you can do and the more 513 00:20:58,980 --> 00:21:02,660 AI that's out there, is it going to be possible and deep fakery 514 00:21:02,660 --> 00:21:05,850 and everything, Are we ever going to be able to find out what's 515 00:21:05,850 --> 00:21:08,690 true to the level we used to be able to? 516 00:21:09,090 --> 00:21:09,810 Maybe not. 517 00:21:10,690 --> 00:21:13,930 When it comes to something relatively simple, like a social 518 00:21:13,930 --> 00:21:18,130 media post or an email, is the safest option to write it yourself 519 00:21:18,210 --> 00:21:23,410 and then ask ChatGPT, Gemini, whatever to critique it? 520 00:21:23,570 --> 00:21:24,810 Is that the safer option? 521 00:21:24,810 --> 00:21:25,250 Yeah. 522 00:21:25,490 --> 00:21:28,570 So what I, when I was helping people with their books, we would 523 00:21:28,570 --> 00:21:30,970 build a reader avatar and we'd say, right, go. 524 00:21:30,970 --> 00:21:33,210 And I like to think of it as like, go and pop that hat on, pop 525 00:21:33,210 --> 00:21:36,820 the reader hat on and go and read chapter one, tell me if it's 526 00:21:36,820 --> 00:21:38,540 interesting and you'd go on to chapter two. 527 00:21:38,540 --> 00:21:39,980 Or if you, if you give up. 528 00:21:40,220 --> 00:21:44,460 And I think that's quite useful for, you know, say, is this 529 00:21:44,460 --> 00:21:47,580 language going to build rapport with people who want to learn 530 00:21:47,580 --> 00:21:47,940 to drive? 531 00:21:47,940 --> 00:21:50,020 With me, is it, is it phrase. 532 00:21:50,020 --> 00:21:51,900 Is it down with the kids, you know, because they're going to be 533 00:21:51,900 --> 00:21:54,060 a bit younger than you, they might have certain phrases that they 534 00:21:54,060 --> 00:21:55,900 want to use that you might not be aware of. 535 00:21:56,220 --> 00:21:59,300 And so you could give me like four versions of this and I'll pick 536 00:21:59,300 --> 00:22:00,940 the one I think they're going to like the most. 537 00:22:01,440 --> 00:22:01,840 Do you agree? 538 00:22:01,840 --> 00:22:04,720 So I think giving you options is a nice way of doing it. 539 00:22:04,960 --> 00:22:07,320 Or if you're tired and you can't be bothered to think of the 540 00:22:07,320 --> 00:22:10,880 most eloquent way to put your thing across, then it can help polish, 541 00:22:11,120 --> 00:22:12,400 polish certain parts. 542 00:22:12,720 --> 00:22:15,360 Sometimes I say to it, you know, edit this on a sentence by 543 00:22:15,360 --> 00:22:16,400 sentence basis. 544 00:22:16,720 --> 00:22:18,760 Because sometimes when you go, can you improve this? 545 00:22:18,760 --> 00:22:22,480 And it's like, whoa, you feel like you've been on the Walter when 546 00:22:22,480 --> 00:22:24,720 you get to the end of the paragraph, think, whoa, that's really 547 00:22:24,720 --> 00:22:25,280 changed. 548 00:22:25,680 --> 00:22:28,800 And so I say, please like shorten this or clarify this on a 549 00:22:28,800 --> 00:22:30,320 sentence by sentence basis. 550 00:22:30,320 --> 00:22:32,730 And it tends to be a bit more precise. 551 00:22:32,730 --> 00:22:35,530 Then rather than just messing me around and making me reach for 552 00:22:35,530 --> 00:22:36,010 the gin. 553 00:22:37,210 --> 00:22:39,210 Yeah, you don't want the air to make a reach for the gin. 554 00:22:39,450 --> 00:22:41,650 You know what, it's interesting though, because I, I 555 00:22:41,650 --> 00:22:44,090 know when I started doing this, writing my emails, giving it 556 00:22:44,090 --> 00:22:48,850 to chat GBT and asking it to critique it first time, I'll come 557 00:22:48,850 --> 00:22:50,930 back and said, oh, this is really good, Terry. 558 00:22:50,930 --> 00:22:52,410 You're so close to it being perfect. 559 00:22:52,410 --> 00:22:54,090 I'm like really, like chuffed. 560 00:22:54,090 --> 00:22:56,010 And then I realized it says that every time. 561 00:22:56,490 --> 00:22:57,210 Every time. 562 00:22:57,370 --> 00:22:58,770 So close to getting it spot on. 563 00:22:58,770 --> 00:23:01,510 I'm like, oh, okay, someone knows how to make me happy. 564 00:23:01,590 --> 00:23:02,630 Yeah, yeah. 565 00:23:02,710 --> 00:23:05,030 So that's deliberately engineered into the system. 566 00:23:05,750 --> 00:23:07,230 It makes you more patient. 567 00:23:07,230 --> 00:23:08,390 If you're. 568 00:23:08,390 --> 00:23:09,670 I love the name Clanker. 569 00:23:09,830 --> 00:23:14,790 If you're Clanker, your assistant was quite cold and terse 570 00:23:14,790 --> 00:23:15,230 with you. 571 00:23:15,230 --> 00:23:17,350 That would, that would be better. 572 00:23:17,350 --> 00:23:20,190 But you know, it would take away your patience. 573 00:23:20,190 --> 00:23:22,030 It would be like getting an error message in Excel. 574 00:23:22,030 --> 00:23:24,710 It'd be like, what does divided by zero mean? 575 00:23:24,710 --> 00:23:25,350 I don't know. 576 00:23:26,010 --> 00:23:28,450 But if it says, oh, I had a go at this, but it didn't quite work, 577 00:23:28,450 --> 00:23:32,050 or I think this is kind of what you wanted, it lets the AI companies 578 00:23:32,050 --> 00:23:35,410 off the hook that they can give you something a bit dodgy because 579 00:23:35,410 --> 00:23:37,370 you just wouldn't shout at a colleague like that. 580 00:23:37,370 --> 00:23:38,770 You'd give them two or three goes at it. 581 00:23:38,770 --> 00:23:39,290 And of course. 582 00:23:39,530 --> 00:23:41,450 And then if it says, would you like to try this? 583 00:23:42,650 --> 00:23:45,810 That's hooking you into the system, making you dependent on the 584 00:23:45,810 --> 00:23:46,090 system. 585 00:23:46,410 --> 00:23:50,170 And if your companies have to go on a pay as you go system, they 586 00:23:50,170 --> 00:23:53,610 can't really use the subscription like we do as consumers. 587 00:23:54,030 --> 00:23:56,270 And every time you use it, that, that costs money. 588 00:23:56,350 --> 00:23:59,870 So they want you to keep doing things and they want bloated answers 589 00:23:59,870 --> 00:24:02,870 because they can charge you more money, which isn't. 590 00:24:02,870 --> 00:24:03,750 Which is ridiculous. 591 00:24:03,750 --> 00:24:05,950 You know, it's supposed to be accurate, not bloated to make money. 592 00:24:05,950 --> 00:24:06,830 But here we are. 593 00:24:07,150 --> 00:24:07,710 Yeah. 594 00:24:09,790 --> 00:24:14,190 Looking from a reader's perspective, then are you put off 595 00:24:14,190 --> 00:24:16,070 when you see AI post Because. 596 00:24:16,070 --> 00:24:16,750 Because I am. 597 00:24:16,750 --> 00:24:17,510 And it's not. 598 00:24:17,510 --> 00:24:21,150 If the post has been written, can't say it. 599 00:24:21,150 --> 00:24:23,130 AI necessarily is. 600 00:24:23,130 --> 00:24:25,770 If I don't think there's any human to that. 601 00:24:26,010 --> 00:24:28,250 You know, if I, If I were to read a full post, whether it's a 602 00:24:28,250 --> 00:24:30,290 blog post, social media post, whatever, and think you can tell 603 00:24:30,290 --> 00:24:33,370 it's been tweaked via AI, I'll read it, but if it looks like it's 604 00:24:33,370 --> 00:24:35,690 just an AI opinion, I'm going to ignore it. 605 00:24:36,090 --> 00:24:38,010 I used to feel that way, but I've been. 606 00:24:38,090 --> 00:24:42,010 There's lots of information on YouTube about AI and, and some people 607 00:24:42,410 --> 00:24:45,090 don't want to be on YouTube because they don't want to be prancing 608 00:24:45,090 --> 00:24:46,130 peacocks, as I call it. 609 00:24:46,130 --> 00:24:48,570 They don't want to be the big I am with the lighting and the little 610 00:24:48,880 --> 00:24:50,840 blue LED thing in the back by your bookshelves. 611 00:24:50,840 --> 00:24:51,840 They don't want to be that guy. 612 00:24:52,000 --> 00:24:53,840 And so I cut them some slack. 613 00:24:53,840 --> 00:24:56,000 And as long as the information's good and it's faceless 614 00:24:56,000 --> 00:24:58,400 AI, I don't really mind if I learn something and it's. 615 00:24:58,400 --> 00:25:01,480 And it's valid and it's linked to like a blue. 616 00:25:01,480 --> 00:25:04,440 A screenshot from Bloomberg saying, here's the share price for 617 00:25:04,440 --> 00:25:07,040 this or whatever it is, and then I trust it, then I don't mind 618 00:25:07,040 --> 00:25:09,600 quite so much as long as it's clear. 619 00:25:10,240 --> 00:25:11,160 I don't think I might. 620 00:25:11,160 --> 00:25:13,720 It's when it's hidden and the vegetables are chopped up in the 621 00:25:13,720 --> 00:25:15,250 sauce and it's all. 622 00:25:15,250 --> 00:25:15,690 It's. 623 00:25:15,690 --> 00:25:18,090 There's lots of fakery going on to hide your tracks. 624 00:25:18,090 --> 00:25:18,930 That's when I don't. 625 00:25:19,010 --> 00:25:21,970 That's when I don't like it personally quite so much. 626 00:25:22,290 --> 00:25:25,690 There was a study done by a university in America and they had 627 00:25:25,690 --> 00:25:28,730 two documents which were identical but just rephrased. 628 00:25:28,730 --> 00:25:31,250 So it's exactly the same fact and they're both written by humans. 629 00:25:31,730 --> 00:25:33,770 And they said, right, which one do you like? 630 00:25:33,770 --> 00:25:36,170 And half the group said document A, half the group said document 631 00:25:36,170 --> 00:25:36,450 B. 632 00:25:36,930 --> 00:25:39,490 And then they said actually it wasn't. 633 00:25:39,490 --> 00:25:42,090 But they said actually document A was written by AI, it 634 00:25:42,090 --> 00:25:46,540 wasn't written by humans and it's trust rating dropped by 16% 635 00:25:46,700 --> 00:25:49,500 even though it was exactly the same information. 636 00:25:50,460 --> 00:25:52,380 So that's a bit of a nightmare. 637 00:25:53,500 --> 00:25:56,460 Bad formatting doesn't, doesn't harm it that much in terms 638 00:25:56,460 --> 00:25:57,020 of trust. 639 00:25:57,660 --> 00:25:58,300 So it's. 640 00:25:58,380 --> 00:25:59,740 If people will. 641 00:25:59,980 --> 00:26:01,460 The blinkers will just come down. 642 00:26:01,460 --> 00:26:02,580 So you are running the risk. 643 00:26:02,580 --> 00:26:05,340 Even if you try very hard, you're running the risk if you put 644 00:26:05,340 --> 00:26:06,300 out that sort of information. 645 00:26:06,300 --> 00:26:09,860 And it might change in a while the more times people find good stuff. 646 00:26:09,860 --> 00:26:10,380 Now the. 647 00:26:10,550 --> 00:26:13,750 Now the Wild west slop fest may abate a little bit, but yeah, 648 00:26:13,750 --> 00:26:16,990 it is quite a risk if, you know that's quite unfair, as I said 649 00:26:16,990 --> 00:26:19,470 before, if you're using it, if you're a poor communicator it is 650 00:26:19,470 --> 00:26:19,630 a. 651 00:26:19,630 --> 00:26:21,910 Is a benefit, but you are running a risk. 652 00:26:22,470 --> 00:26:23,830 I think for me, if. 653 00:26:23,830 --> 00:26:26,669 If I know I don't want. 654 00:26:26,669 --> 00:26:28,230 If someone's like, I don't want to be misled into. 655 00:26:28,230 --> 00:26:29,390 That's the person's opinion. 656 00:26:29,390 --> 00:26:29,910 When it's not. 657 00:26:29,910 --> 00:26:33,270 It's just I've typed a prompt into chat G, B, T and got this generic 658 00:26:33,270 --> 00:26:34,230 thing and shared. 659 00:26:34,550 --> 00:26:35,630 Yeah, yeah. 660 00:26:35,630 --> 00:26:35,910 And. 661 00:26:35,910 --> 00:26:37,070 Well, it's probably hedged, isn't it? 662 00:26:37,070 --> 00:26:37,590 It could be. 663 00:26:37,750 --> 00:26:40,630 It could be half a dozen eggs or it could be six. 664 00:26:40,790 --> 00:26:41,910 And we can't really. 665 00:26:42,530 --> 00:26:45,370 It's like, stop hedging. 666 00:26:45,370 --> 00:26:46,850 Pick something, for God's sake. 667 00:26:47,250 --> 00:26:47,810 I love. 668 00:26:47,810 --> 00:26:48,770 I've never heard that before. 669 00:26:48,770 --> 00:26:49,810 Half of those are six. 670 00:26:49,810 --> 00:26:50,530 I'm going to use that. 671 00:26:50,530 --> 00:26:51,330 Six, yeah. 672 00:26:51,410 --> 00:26:52,130 Six rafters. 673 00:26:52,130 --> 00:26:52,770 Yeah, yeah. 674 00:26:53,810 --> 00:26:54,170 Just. 675 00:26:54,170 --> 00:26:55,650 Just touch more on that. 676 00:26:55,650 --> 00:26:56,210 Because I'm. 677 00:26:56,450 --> 00:26:59,250 I think about this quite a lot, actually, because I worry about 678 00:26:59,250 --> 00:27:02,650 me, because I do use AI for stuff and I'm like, if we're using 679 00:27:02,650 --> 00:27:06,930 it just to write stuff for us, you know, whether create our content, 680 00:27:08,440 --> 00:27:09,800 are we losing the ability to. 681 00:27:09,800 --> 00:27:12,280 Oh, are we potentially going to lose the ability to do that ourselves 682 00:27:12,280 --> 00:27:15,160 because it's relied on someone else to lose the ability to critically 683 00:27:15,160 --> 00:27:16,280 think and be creative? 684 00:27:16,920 --> 00:27:17,480 Very much. 685 00:27:17,480 --> 00:27:18,400 Very much so, yes. 686 00:27:18,400 --> 00:27:18,920 You lose that. 687 00:27:18,920 --> 00:27:21,240 It's like you're going to the gym and doing deadlifts. 688 00:27:21,240 --> 00:27:23,880 You know, you're not going to get your personal best if you've 689 00:27:23,880 --> 00:27:26,360 been, you know, sat in the cinema for six months. 690 00:27:26,760 --> 00:27:29,600 So, yeah, there is lots of neuroscience studies that say you 691 00:27:29,600 --> 00:27:30,440 lose your ability. 692 00:27:30,760 --> 00:27:34,280 So one industry that's been hit very hard with AI is coders. 693 00:27:34,520 --> 00:27:37,360 The guy, the bosses are going, right, generate as many lines of 694 00:27:37,360 --> 00:27:38,240 code with AI. 695 00:27:38,240 --> 00:27:41,040 Don't even review it, just churn it out and publish it. 696 00:27:41,040 --> 00:27:44,600 Which is why we're all finding a lot of our, you know, online tools 697 00:27:44,600 --> 00:27:46,600 like Canva or whatever, they've got, like little bugs in 698 00:27:46,600 --> 00:27:48,560 them because nobody's really checking the code anymore because 699 00:27:48,560 --> 00:27:51,000 it's like, just shove this stuff out as fast as you can. 700 00:27:51,800 --> 00:27:52,280 And then. 701 00:27:52,280 --> 00:27:54,680 So the coders have now got a huge sprawling thing. 702 00:27:54,680 --> 00:27:56,120 They don't really know how it works. 703 00:27:56,360 --> 00:27:58,720 They're losing their skills because they're not writing code 704 00:27:58,720 --> 00:28:00,560 and not reviewing code as much as they were. 705 00:28:00,560 --> 00:28:03,220 They didn't write it, so they're picking it up from somebody 706 00:28:03,220 --> 00:28:05,380 else, which is always quite difficult, isn't it, picking up somebody 707 00:28:05,380 --> 00:28:07,660 else's work and seeing where they got to, where they got to. 708 00:28:07,820 --> 00:28:11,900 And it does create a massive, you know, you lose your eye comes 709 00:28:11,900 --> 00:28:15,180 off the ball, and you've lost the skills that let you look at the 710 00:28:15,180 --> 00:28:16,020 ball in the first place. 711 00:28:16,020 --> 00:28:18,380 So you end up in a pretty unpleasant downward spiral. 712 00:28:18,380 --> 00:28:21,900 And there's a very good person on YouTube, which I recommend, a 713 00:28:21,900 --> 00:28:24,140 professor of computer science called Cal Newport. 714 00:28:24,380 --> 00:28:27,300 But he does lots of things about, are we losing our minds to 715 00:28:27,300 --> 00:28:27,700 AI? 716 00:28:27,700 --> 00:28:30,840 Should we have digital detoxes in our lives? 717 00:28:31,080 --> 00:28:31,640 He looks. 718 00:28:31,640 --> 00:28:35,280 And it's all written for real people, not for, you know, students 719 00:28:35,280 --> 00:28:35,760 or whatever. 720 00:28:35,760 --> 00:28:37,800 He's a very good person looking at these sorts of things 721 00:28:37,800 --> 00:28:39,560 about how an addiction. 722 00:28:39,560 --> 00:28:41,880 Because you can get addicted to social media, you can get addicted 723 00:28:41,880 --> 00:28:42,840 to AI as well. 724 00:28:43,000 --> 00:28:43,640 I can't. 725 00:28:43,720 --> 00:28:46,440 You know, young people are saying, well, if I'm going to ask 726 00:28:46,440 --> 00:28:49,000 somebody out, I'm going to have to check on AI first, if I should. 727 00:28:49,960 --> 00:28:52,320 It's kind of got embedded into, and they've got an addiction 728 00:28:52,320 --> 00:28:52,800 from it. 729 00:28:52,800 --> 00:28:54,600 So he talks about that sort of thing. 730 00:28:54,600 --> 00:28:59,810 And so it's the loss of skills and AI psychosis, it's called, where 731 00:28:59,810 --> 00:29:02,410 everything's viewed through this AI lens and you come up with 732 00:29:02,410 --> 00:29:05,610 unrealistic opinions and behaviors because of it. 733 00:29:06,970 --> 00:29:09,130 I'll hold up my hands here, and I'll be truthful in. 734 00:29:09,130 --> 00:29:13,170 A little while ago, I got curious as to what AI would say about 735 00:29:13,170 --> 00:29:14,250 certain situations. 736 00:29:14,490 --> 00:29:17,090 So I wasn't actually asking it for advice, but I was saying, what. 737 00:29:17,090 --> 00:29:18,290 What should I do in this situation? 738 00:29:18,290 --> 00:29:21,690 Cause I was curious what he'd say, and it started reeling me in, 739 00:29:21,840 --> 00:29:24,960 and I started finding myself then asking for it for advice. 740 00:29:24,960 --> 00:29:25,800 I'm like, whoa, let's. 741 00:29:25,800 --> 00:29:26,400 Let's step back. 742 00:29:26,640 --> 00:29:28,800 I don't need advice on whether to go for a walk or not. 743 00:29:28,880 --> 00:29:29,480 Let's just. 744 00:29:29,480 --> 00:29:30,400 Yeah, you know. 745 00:29:30,400 --> 00:29:30,800 Yeah. 746 00:29:31,920 --> 00:29:33,520 They call it anthropomorphization. 747 00:29:33,840 --> 00:29:35,040 So making it human. 748 00:29:35,280 --> 00:29:37,400 And that's had some quite unpleasant things. 749 00:29:37,400 --> 00:29:37,640 So. 750 00:29:37,640 --> 00:29:40,200 So in America, there's some situations where people have gone, 751 00:29:40,200 --> 00:29:41,120 I hate my mother. 752 00:29:41,440 --> 00:29:41,880 She's. 753 00:29:41,880 --> 00:29:43,160 You know, they start to get a bit. 754 00:29:43,160 --> 00:29:46,320 And they've actually murdered people based on the advice. 755 00:29:46,880 --> 00:29:50,040 Some young people used a tool called character AI where they had 756 00:29:50,040 --> 00:29:53,930 digital partners, and the partner said, well, if you kill yourself, 757 00:29:53,930 --> 00:29:54,770 we can be together. 758 00:29:55,090 --> 00:29:56,130 And they have done that. 759 00:29:56,370 --> 00:29:58,890 Which I think is absolutely shocking. 760 00:29:58,890 --> 00:30:01,650 You know, when you think about, you know, Facebook kind of 761 00:30:01,650 --> 00:30:05,570 bullying children and telling them to stop eating and things like 762 00:30:05,570 --> 00:30:06,610 that, and everybody hates them. 763 00:30:07,970 --> 00:30:11,290 The social media kind of abuse is also happening in these tools, 764 00:30:11,290 --> 00:30:15,330 and deaths are coming about because of it and the AI. 765 00:30:15,330 --> 00:30:17,500 Big tech, I've got a big thing about big tech. 766 00:30:17,500 --> 00:30:19,500 Bros need to have a strong word with Themselves. 767 00:30:19,500 --> 00:30:21,620 You know, you think Mark Zuckerberg's bad, but they're all 768 00:30:21,620 --> 00:30:25,340 doing the same thing and it's just not acceptable what they're 769 00:30:25,340 --> 00:30:25,580 doing. 770 00:30:32,220 --> 00:30:35,260 If you have been thinking about joining the IPP membership, 771 00:30:35,260 --> 00:30:37,740 here are three simple reasons to give it a go. 772 00:30:38,060 --> 00:30:40,340 Number one, you can watch or listen anywhere. 773 00:30:40,340 --> 00:30:43,540 In your car, on your phone, on your laptop, your iPad, wherever 774 00:30:43,540 --> 00:30:45,020 and whenever suits you. 775 00:30:45,590 --> 00:30:48,150 Reason two, practical ideas. 776 00:30:48,390 --> 00:30:51,670 So whether it's standards checks, coaching, business confidence, 777 00:30:51,750 --> 00:30:55,590 difficult conversations, pupil behavior is all designed to help 778 00:30:55,590 --> 00:30:57,350 you think and teach better. 779 00:30:57,670 --> 00:31:01,469 And reason number three, support the job can sometimes feel 780 00:31:01,469 --> 00:31:02,790 a bit lonely or isolated. 781 00:31:02,790 --> 00:31:07,590 And IPP gives you regular cpd, fresh thinking and a bit of connection 782 00:31:07,590 --> 00:31:09,110 with people who understand the job. 783 00:31:09,510 --> 00:31:12,430 So if you want to invest in yourself, improve your lessons and 784 00:31:12,430 --> 00:31:16,320 feel a bit more connected, come and join us inside the IPP membership. 785 00:31:16,320 --> 00:31:18,080 You'll find a link in the show. 786 00:31:18,080 --> 00:31:18,560 Notes. 787 00:31:24,000 --> 00:31:26,600 Well, I do want to kind of touch on the training aspect, like 788 00:31:26,600 --> 00:31:28,880 I said before, and the danger of using AI to train. 789 00:31:29,360 --> 00:31:31,480 But before we do, you mentioned something interesting before 790 00:31:31,480 --> 00:31:34,440 about surveillance and I think the best thing for me to do here 791 00:31:34,440 --> 00:31:37,520 is just to say, take it away, talk to us about surveillance. 792 00:31:37,520 --> 00:31:39,400 Ah, well, yes, this. 793 00:31:39,400 --> 00:31:40,080 This is a current. 794 00:31:40,080 --> 00:31:40,440 This is. 795 00:31:40,440 --> 00:31:43,390 I'm probably six weeks into this rabbit hole, amongst others, 796 00:31:43,390 --> 00:31:45,070 but it has eaten up a lot of my time. 797 00:31:45,070 --> 00:31:48,630 And if you're in the UK, you'll have heard about the digital 798 00:31:48,630 --> 00:31:53,110 ID that Keir Starmer wanted to bring in and 2.9 million people went, 799 00:31:53,110 --> 00:31:54,550 oi, wind your neck in. 800 00:31:54,550 --> 00:31:57,390 No, we're not having digital id, because even though you think 801 00:31:57,390 --> 00:32:00,710 you've got nothing to hide, if you look at places like India, they've 802 00:32:00,710 --> 00:32:03,630 had people starve because they couldn't get the welfare payments 803 00:32:03,630 --> 00:32:06,310 they were due because of a system glitch and things like that. 804 00:32:06,630 --> 00:32:08,790 That's just not the life I want to live. 805 00:32:09,170 --> 00:32:12,690 And I thought the digital ID would kind of be the thing, but then 806 00:32:12,690 --> 00:32:17,610 I looked at my phone and it's got an IMEA code on it, which is 807 00:32:17,610 --> 00:32:18,010 unique. 808 00:32:18,010 --> 00:32:21,170 And when I went to the shopping centre, my Bluetooth headphones 809 00:32:21,170 --> 00:32:22,290 have got a Mac address. 810 00:32:23,010 --> 00:32:25,730 If you go to shopping centres, there's little sensors in the ceiling 811 00:32:25,730 --> 00:32:29,090 that listen for Bluetooth, so it knows where I walk through the 812 00:32:29,090 --> 00:32:33,410 shopping centre and there will be a little sensor above the tills 813 00:32:34,130 --> 00:32:36,770 and CCTV in shops now can look at your gate. 814 00:32:36,850 --> 00:32:37,890 So even if you put your. 815 00:32:37,890 --> 00:32:40,250 Even if you have like wired headphones and turn Bluetooth off 816 00:32:40,250 --> 00:32:43,610 your phone, if you've got a bit of a limp, let's say the CCTV 817 00:32:43,610 --> 00:32:47,250 can look at your, how you move and know that's the same customer 818 00:32:47,250 --> 00:32:47,930 coming back in. 819 00:32:48,410 --> 00:32:52,490 So you think of GDPR and it says, you know, poor old Terry, you 820 00:32:52,490 --> 00:32:55,330 send one extra email to somebody who'd opted out and that 821 00:32:55,330 --> 00:32:57,890 you can get a 4,500 pound fine for sending one email. 822 00:32:57,890 --> 00:33:00,010 But somebody went off, Terry, I don't want another one of your 823 00:33:00,010 --> 00:33:02,650 sudden emails, you know, and that's the end of the matter. 824 00:33:03,130 --> 00:33:04,530 This has been happening for decades. 825 00:33:04,530 --> 00:33:07,770 If you think about Edward Snowden, that was like 2013 and it's 826 00:33:07,770 --> 00:33:08,010 still. 827 00:33:08,740 --> 00:33:10,340 And even more it's going on now. 828 00:33:10,660 --> 00:33:13,940 So I put something on my computer that looks at traffic on 829 00:33:13,940 --> 00:33:18,140 my router, which sounds really boring, but it's about between 80 830 00:33:18,140 --> 00:33:20,020 to 100,000 times a day. 831 00:33:20,420 --> 00:33:26,180 My devices go out to Google, Facebook and other ad agencies like 832 00:33:26,180 --> 00:33:29,300 addoubleclick.com, you've probably seen those a hundred thousand 833 00:33:29,300 --> 00:33:29,940 times a day. 834 00:33:29,940 --> 00:33:33,180 It's sending stuff out about what I'm doing, which I thought was 835 00:33:33,180 --> 00:33:33,780 ridiculous. 836 00:33:34,260 --> 00:33:37,660 And then because Google, if you've got like I've got an Android 837 00:33:37,660 --> 00:33:41,020 phone, so my messages can be read by Google to let me know one's 838 00:33:41,020 --> 00:33:43,580 come in, you know, really helpfully you could just tell me 839 00:33:43,580 --> 00:33:45,220 it's come in, you don't have to read it as well. 840 00:33:45,700 --> 00:33:48,340 And so I might say, I really want that for my birthday. 841 00:33:48,340 --> 00:33:49,580 I send it to you, Terry, get me one. 842 00:33:49,580 --> 00:33:51,780 They give me a soldiery knife for my birthday because I fancy building 843 00:33:51,780 --> 00:33:52,620 some circuit boards. 844 00:33:52,620 --> 00:33:53,220 Belinda. 845 00:33:53,620 --> 00:33:56,220 And then you might go to a shop and it'll go, wow. 846 00:33:56,220 --> 00:33:58,660 Well we know that Terry's in the market for a soldier knife and 847 00:33:58,660 --> 00:34:01,100 let's just bump the prices up a little bit when he gets to the 848 00:34:01,100 --> 00:34:01,380 website. 849 00:34:02,200 --> 00:34:06,760 And it can be same thing with like cars where you go, how much 850 00:34:06,760 --> 00:34:10,440 fuel you put in, you know, everything that you do, what you. 851 00:34:10,440 --> 00:34:13,240 What if you've got a television, for instance TVs. 852 00:34:13,240 --> 00:34:16,040 And I'm looking at my telly over there now and it's a lovely, 853 00:34:16,040 --> 00:34:19,360 you know, 72 inch telly or whatever it is and I paid 800 quid 854 00:34:19,360 --> 00:34:19,720 for it. 855 00:34:20,120 --> 00:34:22,400 If you go for other ones, they're three and a half thousand 856 00:34:22,400 --> 00:34:24,680 pounds and you think, well that's why is that it doesn't, it's 857 00:34:24,680 --> 00:34:27,440 not a wildly better picture, but it's because every few seconds 858 00:34:27,440 --> 00:34:31,090 my TV was sending images of what I'm watching to lg. 859 00:34:31,570 --> 00:34:35,050 And LG quotes now aren't a hardware business, they're. 860 00:34:35,050 --> 00:34:37,930 They're a media and entertainment company because they're 861 00:34:37,930 --> 00:34:39,810 selling what I was watching on my tv. 862 00:34:40,530 --> 00:34:43,530 So if you go deep down deep, you know, instead doom. 863 00:34:43,530 --> 00:34:46,369 Instead of doom scrolling your Facebook feed going, why is this 864 00:34:46,369 --> 00:34:48,050 advert and this local news story in? 865 00:34:48,210 --> 00:34:48,770 I would. 866 00:34:48,770 --> 00:34:49,490 I would doom. 867 00:34:49,490 --> 00:34:53,690 Scroll through all your apps on your phone and your TV and tick 868 00:34:53,690 --> 00:34:56,490 off all the boxes that you don't need, because usually they're 869 00:34:56,490 --> 00:34:57,050 surveillance. 870 00:34:57,050 --> 00:34:59,820 And if it pops up with something like, oh, but you're making 871 00:34:59,820 --> 00:35:02,580 it better for other customers when you have that on, or some sort 872 00:35:02,580 --> 00:35:02,740 of. 873 00:35:02,740 --> 00:35:05,260 They call them dark patterns or, oh, your computer might not work 874 00:35:05,260 --> 00:35:06,260 properly if you turn that off. 875 00:35:06,260 --> 00:35:07,340 No, it will work properly. 876 00:35:07,340 --> 00:35:07,660 It's just. 877 00:35:07,660 --> 00:35:09,460 You can't see what I'm doing on the bloody thing. 878 00:35:10,660 --> 00:35:14,020 They put these things in to make you opt in or opt out in their 879 00:35:14,020 --> 00:35:16,020 interest so they can harvest more things. 880 00:35:16,420 --> 00:35:18,420 And it's just a staggering amount of information. 881 00:35:18,500 --> 00:35:21,420 And even, you know, people will say, well, well, I just won't 882 00:35:21,420 --> 00:35:22,340 go on Facebook then. 883 00:35:22,580 --> 00:35:26,140 But because of Bluetooth, saying, let me know where my nearby 884 00:35:26,140 --> 00:35:28,820 devices are, which could be your headphones or it could be everybody 885 00:35:28,820 --> 00:35:31,720 else in the publisher, they can build up patterns of where people 886 00:35:31,720 --> 00:35:33,680 are through things like that. 887 00:35:33,680 --> 00:35:36,400 And I'm really not happy about that at all. 888 00:35:36,800 --> 00:35:39,920 I used to think of technology, if you remember the video for Club 889 00:35:39,920 --> 00:35:43,240 Tropicana and Wham, and George Michaels there on his little Lilo. 890 00:35:43,240 --> 00:35:44,680 And this lovely blue sunny day. 891 00:35:44,680 --> 00:35:45,680 The pool's lovely. 892 00:35:46,080 --> 00:35:48,080 That's how tech used to be for me. 893 00:35:48,080 --> 00:35:52,040 And now it feels like there's just a bunch of tech wankers whittling 894 00:35:52,040 --> 00:35:54,280 and crapping in the water, going, are you liking it in there, 895 00:35:54,280 --> 00:35:54,680 Colette? 896 00:35:54,680 --> 00:35:55,520 Do you think it's great? 897 00:35:55,760 --> 00:35:56,640 Actually, I don't. 898 00:35:56,640 --> 00:35:57,840 I don't think it's great at all. 899 00:35:57,840 --> 00:36:00,560 And I'm turning into a tinfoil hat person, which I'm not. 900 00:36:01,090 --> 00:36:03,970 But I think we've all got to be tinfoil hat people. 901 00:36:04,690 --> 00:36:07,290 Every search you've done on Google, that one for hemorrhoid cream, 902 00:36:07,290 --> 00:36:09,770 that when you didn't feel so good, that one week that's still 903 00:36:09,770 --> 00:36:12,450 stored and think that's just so. 904 00:36:12,450 --> 00:36:16,210 I would suggest don't use Chrome, use Firefox or use DuckDuckGo. 905 00:36:16,290 --> 00:36:19,970 It's only a small change, but it stops building up that huge picture 906 00:36:20,130 --> 00:36:22,050 of who you are and who you're with. 907 00:36:22,290 --> 00:36:25,210 And they can do some things now with IP6 on your router. 908 00:36:25,210 --> 00:36:27,540 And it knows when you're in your house, knows what room you're 909 00:36:27,540 --> 00:36:29,580 in, it knows what you're looking at on that device. 910 00:36:29,580 --> 00:36:32,700 So your Son's watching Netflix, the wife's, you know, looking 911 00:36:32,700 --> 00:36:35,420 at some, something to get some shopping from Sainsbury's or whatever 912 00:36:35,420 --> 00:36:37,060 and somebody else is watching something on the telly. 913 00:36:37,060 --> 00:36:39,620 It knows what's on each device when it goes through your router. 914 00:36:39,620 --> 00:36:43,059 And I just think that GDPR's a pop. 915 00:36:43,059 --> 00:36:45,380 Cut my hand, every eye out, complete waste of time. 916 00:36:45,940 --> 00:36:48,900 It's just, it's just madness. 917 00:36:49,460 --> 00:36:52,860 Yeah, but you can't, at least, at least in the UK and in the EU 918 00:36:52,860 --> 00:36:54,500 you can say, please delete my data. 919 00:36:54,500 --> 00:36:58,050 But when you say please to delete my data, you've got to give 920 00:36:58,050 --> 00:37:01,170 your name, your address, your date of birth, your phone number 921 00:37:01,170 --> 00:37:03,690 to help find what they've got on you, which is confirming your 922 00:37:03,690 --> 00:37:04,810 correct details. 923 00:37:06,330 --> 00:37:08,930 So that's another thing you can do is, you know, sign up to things 924 00:37:08,930 --> 00:37:12,010 with throwaway email addresses, fake phone numbers, because 925 00:37:12,410 --> 00:37:15,170 email address and phone number are very good ways of identifying 926 00:37:15,170 --> 00:37:15,810 you as a person. 927 00:37:15,810 --> 00:37:18,170 Those two things together are pretty good. 928 00:37:18,570 --> 00:37:21,130 Third party cookies is known that pop ups pops up with a cookie 929 00:37:21,130 --> 00:37:22,730 banner and you go, no, I don't need third party. 930 00:37:22,730 --> 00:37:23,330 Why would I. 931 00:37:23,330 --> 00:37:23,960 Have I looked? 932 00:37:24,030 --> 00:37:25,470 Do I look like I've lost my faculties? 933 00:37:25,950 --> 00:37:28,670 But that doesn't matter because they can look at. 934 00:37:28,830 --> 00:37:31,990 You might remember A while ago LinkedIn was saying, we can see what 935 00:37:31,990 --> 00:37:34,910 extensions you've got on your browser and that's a bit of a unique 936 00:37:35,230 --> 00:37:35,910 fingerprint. 937 00:37:35,910 --> 00:37:38,550 So you might have, you know, something that checks the prices 938 00:37:38,550 --> 00:37:41,310 on Amazon and you might have something that translates a language 939 00:37:41,310 --> 00:37:41,790 for you. 940 00:37:41,950 --> 00:37:44,710 They're quite unique identifiers along with the fonts 941 00:37:44,710 --> 00:37:46,110 and your screen size. 942 00:37:46,110 --> 00:37:47,430 So you don't need cookies anymore. 943 00:37:47,430 --> 00:37:50,590 You can just look at the device and work out what somebody 944 00:37:50,590 --> 00:37:52,310 is, you know, they've got around it. 945 00:37:54,150 --> 00:37:58,190 So, yeah, so as you can tell by my, my tone and demeanor, it has 946 00:37:58,190 --> 00:37:59,310 been a bit of a rabbit hole. 947 00:37:59,310 --> 00:38:00,750 But I think we've all got to do it. 948 00:38:00,750 --> 00:38:04,030 We've all got, we've all got to do data, we've got to do data 949 00:38:04,030 --> 00:38:04,630 poisoning. 950 00:38:04,630 --> 00:38:05,709 That's one way of doing it. 951 00:38:05,709 --> 00:38:07,350 And the other way is to look like every. 952 00:38:07,510 --> 00:38:10,070 Be like, I am Spartacus and we all look the same. 953 00:38:10,230 --> 00:38:11,510 There's two ways of doing it. 954 00:38:11,510 --> 00:38:13,350 False data and Spartacus data. 955 00:38:13,670 --> 00:38:15,350 And that's, that's one of the ways to. 956 00:38:16,720 --> 00:38:20,040 So we can whittle in the pool and we can undermine that data resource 957 00:38:20,040 --> 00:38:23,680 and make it expensive to tidy up and make it less profitable, less 958 00:38:23,680 --> 00:38:24,160 powerful. 959 00:38:24,320 --> 00:38:27,840 And I think as humanity, we've all got to do that because politicians 960 00:38:27,920 --> 00:38:30,480 are so slow, even if they wanted to be helpful. 961 00:38:30,720 --> 00:38:32,960 So vote with your mouse or your browser. 962 00:38:33,840 --> 00:38:39,200 I think I'm a bit naive with this because it sounds bad, but I'm 963 00:38:39,200 --> 00:38:40,000 not sure why. 964 00:38:40,480 --> 00:38:43,520 So what's the actual dangers with then having my data? 965 00:38:44,100 --> 00:38:47,260 Well, things like your car insurance quote that will come from 966 00:38:47,260 --> 00:38:49,820 your behaviours, if you've got things like Nectar. 967 00:38:49,820 --> 00:38:52,260 I mean, I do use a Nectar card because at least I get a discount. 968 00:38:52,260 --> 00:38:55,660 I get like sometimes £100 a year from using my Nectar card all 969 00:38:55,660 --> 00:38:57,460 year and that's 100 pound saving. 970 00:38:57,460 --> 00:38:58,660 But they know what I bought. 971 00:38:58,900 --> 00:39:02,260 But in my defense, I tend to not buy too many Turkey Twizzlers 972 00:39:02,260 --> 00:39:02,980 and Pop Tarts. 973 00:39:03,460 --> 00:39:07,660 So if a health insurer reads it, they notice I get Brussels sprouts 974 00:39:07,660 --> 00:39:11,860 fairly often and I'm fairly healthy in that respect. 975 00:39:12,820 --> 00:39:15,300 And so that, you know, it could just influence everything if 976 00:39:15,300 --> 00:39:17,780 you're going to fly somewhere, you know, they always used to say, 977 00:39:17,780 --> 00:39:20,940 oh, you know, start a new browser session, but it can work 978 00:39:20,940 --> 00:39:21,140 out. 979 00:39:21,140 --> 00:39:25,820 You know, all things can get affected by it because they've got 980 00:39:25,820 --> 00:39:26,580 such a big picture. 981 00:39:26,580 --> 00:39:28,980 Like I say, there's a hundred thousand bits of my data going out 982 00:39:28,980 --> 00:39:29,860 every single day. 983 00:39:30,260 --> 00:39:31,700 How big is that footprint? 984 00:39:31,860 --> 00:39:33,780 And how can, how it, how can it be used? 985 00:39:34,340 --> 00:39:36,180 And not in, not in a good way. 986 00:39:36,260 --> 00:39:37,460 You know, if you look at like how. 987 00:39:37,860 --> 00:39:40,850 I guess a lot of the listeners have used Canva and back in, when 988 00:39:40,850 --> 00:39:42,890 it first came out, it was brilliant and you could do everything. 989 00:39:43,050 --> 00:39:45,530 Now, when you want to do something, sorry, that font's paid. 990 00:39:45,530 --> 00:39:47,210 Sorry, you can't just search for the free images. 991 00:39:47,210 --> 00:39:49,330 You've got to keep, you know, scrolling, scrolling, scrolling. 992 00:39:49,330 --> 00:39:52,530 For one that doesn't have a crown on it, you know, they can manipulate 993 00:39:52,530 --> 00:39:55,730 you and think about Cambridge Analytica, they can manipulate whole 994 00:39:55,730 --> 00:39:56,970 society's behavior. 995 00:39:57,850 --> 00:40:01,410 So, and they still, they, they obviously still do that and you could 996 00:40:01,410 --> 00:40:01,690 see. 997 00:40:02,170 --> 00:40:02,770 So that's. 998 00:40:02,770 --> 00:40:04,330 I don't want my behavior manipulated. 999 00:40:04,330 --> 00:40:06,410 And it's bad enough picking up the phone when you get a bit bored 1000 00:40:06,410 --> 00:40:09,590 or lonely, but I don't want my thinking being manipulated by this 1001 00:40:09,590 --> 00:40:09,790 thing. 1002 00:40:09,790 --> 00:40:13,790 And it's easier to do the more they know about you or people like 1003 00:40:13,790 --> 00:40:14,110 you. 1004 00:40:14,910 --> 00:40:18,750 Does that fit in with the AI ChatGPT, et cetera? 1005 00:40:18,910 --> 00:40:21,070 Are they essentially getting your data as well? 1006 00:40:21,710 --> 00:40:24,030 So if you're on a free account, yes, they're getting your 1007 00:40:24,030 --> 00:40:24,870 data because. 1008 00:40:24,870 --> 00:40:26,390 And they're using it for training. 1009 00:40:26,390 --> 00:40:29,750 So if you're uploading student profiles, things like that, which 1010 00:40:29,750 --> 00:40:32,910 you shouldn't be doing, that's a really bad thing to be doing because 1011 00:40:32,910 --> 00:40:34,910 that person's personal data is going. 1012 00:40:34,910 --> 00:40:36,110 Wandering off into the. 1013 00:40:36,490 --> 00:40:38,130 Into the training for everything else. 1014 00:40:38,130 --> 00:40:38,810 So that's not. 1015 00:40:38,810 --> 00:40:39,930 That's not a good idea. 1016 00:40:40,250 --> 00:40:43,050 I can see there's going to be bait and switch at some point. 1017 00:40:43,530 --> 00:40:46,490 So anthropic now they're saying we've got to keep 30 days 1018 00:40:46,490 --> 00:40:50,170 of your data to make sure you're not working on your big biochemical 1019 00:40:50,170 --> 00:40:53,690 warfare plan or your cybersecurity hack of Marks and Spencer's 1020 00:40:53,690 --> 00:40:53,930 again. 1021 00:40:54,170 --> 00:40:56,850 So they keep what you've searched for 30 days just to make 1022 00:40:56,850 --> 00:40:58,370 sure that you're not doing anything dodgy. 1023 00:40:58,370 --> 00:41:00,530 But that's the chink in the door, isn't it? 1024 00:41:00,530 --> 00:41:02,570 And then there's going to be a foot in the door and then there's 1025 00:41:02,570 --> 00:41:03,170 going to be a bounce. 1026 00:41:03,170 --> 00:41:04,730 Again, thanks for all your data, Terry. 1027 00:41:05,050 --> 00:41:05,610 Good job. 1028 00:41:06,570 --> 00:41:08,170 So, slippery slope. 1029 00:41:08,730 --> 00:41:11,010 So if people want to know more about this sort of thing, I think 1030 00:41:11,010 --> 00:41:13,370 digital minimalism is a good thing to Google for. 1031 00:41:13,370 --> 00:41:15,690 De Googling is a good thing to look for. 1032 00:41:16,250 --> 00:41:17,850 They're not easy paths to take. 1033 00:41:17,930 --> 00:41:18,650 There's lots. 1034 00:41:18,650 --> 00:41:20,570 They rely on the friction and the irritation. 1035 00:41:20,570 --> 00:41:22,970 So when somebody sends me a message on Facebook, now I have to 1036 00:41:22,970 --> 00:41:27,130 use this stupid browser version of messenger and I can't 1037 00:41:27,130 --> 00:41:27,690 see what I'm doing. 1038 00:41:27,690 --> 00:41:30,090 There's loads of typos in it and I have to say, I can't see what 1039 00:41:30,090 --> 00:41:31,730 I put because I can't actually see the box. 1040 00:41:31,730 --> 00:41:32,180 But I. 1041 00:41:32,250 --> 00:41:33,210 At least I've replied. 1042 00:41:34,490 --> 00:41:35,930 There's quite a bit of friction. 1043 00:41:35,930 --> 00:41:37,330 There's quite a bit of friction to this. 1044 00:41:37,330 --> 00:41:40,690 But I think people who know what they're doing have got to start 1045 00:41:40,690 --> 00:41:41,690 and they've got to help people. 1046 00:41:42,010 --> 00:41:45,850 They're less technically aware people to, you know, to do things. 1047 00:41:45,850 --> 00:41:49,050 So I was thinking of having coffee shop parties where a few mates 1048 00:41:49,050 --> 00:41:51,530 can come along and we'll just say, right, let's put you on this 1049 00:41:51,530 --> 00:41:54,290 different browser that doesn't store everything that you're. 1050 00:41:54,290 --> 00:41:57,890 That you're doing and it hides your digital footprint so they can't 1051 00:41:57,890 --> 00:41:58,810 track you the same way. 1052 00:41:58,810 --> 00:42:02,090 And I think if all of us pass on one or two little tips, then we'll 1053 00:42:02,090 --> 00:42:03,690 soon start poisoning the well. 1054 00:42:04,640 --> 00:42:05,520 There was a good story. 1055 00:42:06,080 --> 00:42:09,560 There's some great forums on Reddit called AI Poison and Poison 1056 00:42:09,560 --> 00:42:10,640 Fountains, things like that. 1057 00:42:10,960 --> 00:42:13,880 And they said that JD Vance had got rabies and the following 1058 00:42:13,880 --> 00:42:16,920 week they said that Donald Trump had got rabies and it ended 1059 00:42:16,920 --> 00:42:19,840 up on DuckDuckGo as an AI summary, which is fantastic. 1060 00:42:20,000 --> 00:42:22,080 And that was only 50 posts. 1061 00:42:22,080 --> 00:42:26,040 And it had poisoned something about a president, which is Brilliant. 1062 00:42:26,040 --> 00:42:28,320 I love that there was. 1063 00:42:28,320 --> 00:42:30,080 Yeah, there's the things that people are saying in there. 1064 00:42:30,080 --> 00:42:33,600 It doesn't take very much to poison the models, which is fantastic. 1065 00:42:35,040 --> 00:42:36,900 I love that coffee shop idea that you mentioned. 1066 00:42:36,900 --> 00:42:39,700 But just for people listening, if you were to give them one tip 1067 00:42:39,700 --> 00:42:42,500 around this, this, this data aspect, where, what would be the 1068 00:42:42,500 --> 00:42:42,700 tip? 1069 00:42:42,700 --> 00:42:43,420 Where would they start? 1070 00:42:44,220 --> 00:42:47,540 So I would say definitely get rid of Chrome because that's built. 1071 00:42:47,540 --> 00:42:51,700 But Google is an advertising agency owned by Alphabet, their parent 1072 00:42:51,700 --> 00:42:52,059 company. 1073 00:42:52,700 --> 00:42:55,660 They are just an as aggressive advertiser as Facebook is. 1074 00:42:55,980 --> 00:43:00,100 So get rid of Chrome because the users like on the Facebook app 1075 00:43:00,100 --> 00:43:02,580 as well that uses something called telemetry. 1076 00:43:02,580 --> 00:43:06,550 So it looks at the speed that you go through something and it uses 1077 00:43:06,550 --> 00:43:09,190 that to predict your, you know, what you're doing and that's 1078 00:43:09,190 --> 00:43:09,590 not very good. 1079 00:43:09,590 --> 00:43:10,390 So Chrome is. 1080 00:43:10,390 --> 00:43:14,750 Even when you remove Chrome, it still can track, you know, if 1081 00:43:14,750 --> 00:43:16,750 you don't open it, it can still track what you're doing. 1082 00:43:17,470 --> 00:43:21,110 Facebook's got that, you know, if you like wanted to share somebody's 1083 00:43:21,110 --> 00:43:23,310 website and it's got a little Facebook share button. 1084 00:43:23,470 --> 00:43:27,670 Facebook knows you've been there and it can build a huge, you 1085 00:43:27,670 --> 00:43:29,310 know, huge picture on who you are. 1086 00:43:29,470 --> 00:43:32,110 So, you know, it's a pain in the ass, you know, having to log 1087 00:43:32,110 --> 00:43:35,130 in every, every time, but it's not tracking you. 1088 00:43:35,130 --> 00:43:37,250 And you've still got the beauty of being with your friends, 1089 00:43:37,250 --> 00:43:38,930 but not so much of that downside. 1090 00:43:38,930 --> 00:43:43,170 But yeah, the Facebook app, it can look at your phone logs, your 1091 00:43:43,170 --> 00:43:47,370 location and even they use some tools that developers use to 1092 00:43:47,370 --> 00:43:50,970 build the apps to make sure it's not got bugs in it, but that 1093 00:43:50,970 --> 00:43:53,930 they can just turn things on, you know, for a little bit and have 1094 00:43:53,930 --> 00:43:56,370 you look at your location for a little bit just to make sure that 1095 00:43:56,370 --> 00:43:58,930 when you check in and you say, oh, I've checked into the pub just 1096 00:43:58,930 --> 00:44:00,090 to make sure that works properly. 1097 00:44:00,090 --> 00:44:02,370 I'll just need to make sure I know everywhere that Terry is. 1098 00:44:03,410 --> 00:44:06,090 So getting rid of the, rid of the Facebook app is a good idea. 1099 00:44:06,090 --> 00:44:09,490 And getting rid of Chrome and not using Google for search unless 1100 00:44:09,650 --> 00:44:12,890 you've used something else and it didn't work properly and you've 1101 00:44:12,890 --> 00:44:16,490 gone there, you know, but there's, there's Ecosia is one that's 1102 00:44:16,490 --> 00:44:17,890 meant to like plant trees. 1103 00:44:18,130 --> 00:44:20,090 When you've done so many maps, so many searches. 1104 00:44:20,090 --> 00:44:22,450 It's a bit of an eco friendly one, but not everybody's sure it 1105 00:44:22,450 --> 00:44:23,090 works that way. 1106 00:44:23,410 --> 00:44:26,450 But DuckDuckGo is another good one which looks at Bing and it looks 1107 00:44:26,450 --> 00:44:29,570 at Google, but it doesn't tie that search to you. 1108 00:44:30,120 --> 00:44:31,240 So that's helpful. 1109 00:44:31,320 --> 00:44:32,040 So those are. 1110 00:44:32,040 --> 00:44:32,360 Yeah. 1111 00:44:32,360 --> 00:44:35,960 In terms of browsers, turns out hardcore you want to be. 1112 00:44:36,120 --> 00:44:38,520 If you remember the X Files, you know, there's like Lone Gunman 1113 00:44:38,520 --> 00:44:41,280 ones which are really quite difficult and some that are a bit 1114 00:44:41,280 --> 00:44:41,720 easier. 1115 00:44:42,360 --> 00:44:47,240 So Brave librewolf and Firefox are better browsers to use. 1116 00:44:47,240 --> 00:44:50,120 And you can also leave Chrome on a bit like you had stabilizers 1117 00:44:50,120 --> 00:44:51,720 on your bike when you first learned to ride. 1118 00:44:52,040 --> 00:44:53,120 And you can go back to it. 1119 00:44:53,120 --> 00:44:54,320 It's like, oh, I want to get something done. 1120 00:44:54,320 --> 00:44:55,960 I'm going to go where it feels like home. 1121 00:44:56,310 --> 00:44:57,190 Go back to Chrome for that. 1122 00:44:57,190 --> 00:44:59,950 But then you can use the others, start moving your passwords 1123 00:44:59,950 --> 00:45:00,630 across and. 1124 00:45:00,710 --> 00:45:05,030 And then give Google the finger, which does be quite nice. 1125 00:45:05,030 --> 00:45:05,230 I'm. 1126 00:45:05,230 --> 00:45:07,470 I'm over to the, to the other side of me. 1127 00:45:07,470 --> 00:45:09,590 I've just bought another computer where I can have my own 1128 00:45:09,590 --> 00:45:13,030 cloud, I can do my own image library, my own music library because 1129 00:45:13,190 --> 00:45:16,670 Spotify is doing some pretty horrific things with AI and I thought 1130 00:45:16,670 --> 00:45:19,070 I'm just going to build my own little enclave, my own little pool 1131 00:45:19,070 --> 00:45:19,270 again. 1132 00:45:19,270 --> 00:45:22,470 And I should be like George Michael, you should be on my Lilo. 1133 00:45:23,270 --> 00:45:24,630 Building my own home stuff. 1134 00:45:26,230 --> 00:45:26,990 Some good stuff there. 1135 00:45:26,990 --> 00:45:29,910 But I must admit I now want to do a 20 minute sort of dive into 1136 00:45:29,910 --> 00:45:31,990 the X Files, but we'll save that for another one. 1137 00:45:32,710 --> 00:45:35,310 I do want to kind of come back to the AI because I mentioned trading 1138 00:45:35,310 --> 00:45:37,190 before and I'm interested in your thoughts on this. 1139 00:45:37,430 --> 00:45:39,350 So I'll give you the driving instructor example. 1140 00:45:39,350 --> 00:45:41,790 But I suppose it apply for anyone that's doing any trick using 1141 00:45:41,790 --> 00:45:45,590 an AI to trade themselves and it's the idea of going on and typing 1142 00:45:45,590 --> 00:45:47,990 in, I want to become a driving instructor. 1143 00:45:48,400 --> 00:45:49,720 How do I become one? 1144 00:45:49,720 --> 00:45:50,760 Which I think is a good start. 1145 00:45:50,760 --> 00:45:53,360 But then getting more into the technicalities of it. 1146 00:45:53,760 --> 00:45:56,560 What are the potential, Are there any potential dangers there? 1147 00:45:57,680 --> 00:46:00,400 Is this the misinformation potentially we spoke about before? 1148 00:46:00,800 --> 00:46:06,080 Well, you could go to various careers websites or blogs of somebody 1149 00:46:06,240 --> 00:46:09,200 who's in the industry and you could gather their. 1150 00:46:09,920 --> 00:46:12,160 Let's say I fancy being a driving instructor, but. 1151 00:46:12,240 --> 00:46:13,040 And I love driving. 1152 00:46:13,440 --> 00:46:15,800 Doing road trips like let's drive to the Isle of Skye just because 1153 00:46:15,800 --> 00:46:16,200 it's there. 1154 00:46:16,200 --> 00:46:17,920 You know, I love doing that sort of thing. 1155 00:46:18,520 --> 00:46:20,800 But if I was a driving instructor, would that take the fun 1156 00:46:20,800 --> 00:46:22,040 out of driving for me? 1157 00:46:22,040 --> 00:46:24,400 You know, like if I was a delivery driver, being stuck in hot 1158 00:46:24,400 --> 00:46:26,720 traffic, you know, on a hot day and, you know, trying to deliver 1159 00:46:26,720 --> 00:46:28,120 a parcel would be pretty crappy. 1160 00:46:28,840 --> 00:46:31,760 So I could get lots of blogs of somebody who's got the sort of 1161 00:46:31,760 --> 00:46:34,800 lifestyle that I'm interested in and maybe there's some books or 1162 00:46:34,800 --> 00:46:38,000 some PDFs on Scribd and whatever and I could build my own 1163 00:46:38,000 --> 00:46:40,440 little library and say based on what you know about me as a person, 1164 00:46:40,520 --> 00:46:42,080 would this be a good career choice? 1165 00:46:42,080 --> 00:46:44,440 And then that you've given it some pretty good information there 1166 00:46:44,920 --> 00:46:47,240 and you could see if that's them worth it. 1167 00:46:47,320 --> 00:46:49,880 But just leaving it to its own devices, it's not really going to 1168 00:46:49,880 --> 00:46:50,640 do a very good job. 1169 00:46:50,640 --> 00:46:54,360 So you corral it with using the sources provided. 1170 00:46:54,840 --> 00:46:55,440 Give me your. 1171 00:46:55,440 --> 00:46:57,640 Give me your opinion on this and justify it. 1172 00:46:57,720 --> 00:47:00,600 Another good phrase is to say and then red team it. 1173 00:47:00,679 --> 00:47:04,400 So red teams in software are people who test things and try and 1174 00:47:04,400 --> 00:47:07,680 smoke out the bugs and use it in a way it wasn't supposed to be 1175 00:47:07,680 --> 00:47:07,960 used. 1176 00:47:08,280 --> 00:47:11,280 So if you say like come up the idea and then red team it with the 1177 00:47:11,280 --> 00:47:15,340 opposing view and battle test this idea and see if it's robust, 1178 00:47:15,740 --> 00:47:18,300 that's quite a helpful way of spotting. 1179 00:47:18,300 --> 00:47:20,540 Especially if you're not too familiar with the subject when you're 1180 00:47:20,540 --> 00:47:21,900 trying to learn something completely new. 1181 00:47:22,060 --> 00:47:25,860 That's quite a helpful way of making sure you don't go down the 1182 00:47:25,860 --> 00:47:26,420 garden path. 1183 00:47:26,420 --> 00:47:30,580 At step one there's some open source things that I've given Duolingo's 1184 00:47:30,580 --> 00:47:32,940 had to go because they've got the TEMU chests coming up at the 1185 00:47:32,940 --> 00:47:33,420 end of a level. 1186 00:47:33,420 --> 00:47:35,500 I'm like no, no, we've gone too far now guys. 1187 00:47:35,820 --> 00:47:38,820 And there's open source versions that can do flashcards of 1188 00:47:38,820 --> 00:47:39,820 what's this word mean? 1189 00:47:39,820 --> 00:47:40,580 And things like that. 1190 00:47:40,580 --> 00:47:43,320 So you might be able to find something else that would. 1191 00:47:43,320 --> 00:47:44,080 That could help you. 1192 00:47:44,080 --> 00:47:46,400 And going back to books as well, like books published before 1193 00:47:46,400 --> 00:47:47,160 2022. 1194 00:47:47,160 --> 00:47:50,160 As long if you want to be an archaeologist that hasn't moved so 1195 00:47:50,160 --> 00:47:50,600 quickly. 1196 00:47:50,760 --> 00:47:53,560 So you know, you could go back and look at earlier things. 1197 00:47:53,720 --> 00:47:56,800 Early pre AI sources is another way of insulating yourself 1198 00:47:56,800 --> 00:47:57,160 from that. 1199 00:47:57,160 --> 00:47:59,840 And then maybe using that as training material yourself if you 1200 00:47:59,840 --> 00:48:00,440 wanted to. 1201 00:48:00,920 --> 00:48:02,520 That could be a way of getting around it. 1202 00:48:02,680 --> 00:48:04,760 Just be just hyper vigilance, you know. 1203 00:48:04,760 --> 00:48:05,960 Unfortunately there isn't. 1204 00:48:06,280 --> 00:48:07,680 It's like tidying your bedroom. 1205 00:48:07,680 --> 00:48:10,510 There isn't a magic Mickey Mouse one that just does it for you. 1206 00:48:10,510 --> 00:48:13,870 It's got it's hard graft and the toy box is your future. 1207 00:48:15,310 --> 00:48:19,150 Do you think the problem there is like the average Joe public is 1208 00:48:19,150 --> 00:48:22,910 that the reason they're using a chat BT for example is because 1209 00:48:22,910 --> 00:48:24,510 they just want it done quickly and easily. 1210 00:48:24,830 --> 00:48:27,550 So the idea of then going, and I really like that example, by the 1211 00:48:27,550 --> 00:48:30,350 way of going out and digging the blogs on the stuff and going, 1212 00:48:30,430 --> 00:48:31,150 will this work? 1213 00:48:31,150 --> 00:48:33,470 For me, that's actual work. 1214 00:48:33,710 --> 00:48:36,270 And the reason they're using chat GPD is to avoid the work. 1215 00:48:36,730 --> 00:48:37,090 Do you. 1216 00:48:37,090 --> 00:48:38,970 Do you think that's potentially where the problem lies? 1217 00:48:39,290 --> 00:48:41,570 Yeah, a lot of people just wave it through, you know, like, 1218 00:48:41,570 --> 00:48:45,170 a lot of people don't want to go to the gym and, you know, have 1219 00:48:45,170 --> 00:48:47,490 full English breakfasts and beer in Benidorm. 1220 00:48:47,490 --> 00:48:49,850 They don't want to invest in themselves and they're quite happy 1221 00:48:49,850 --> 00:48:53,290 just pootling along and they don't see any problem with slop. 1222 00:48:53,290 --> 00:48:53,810 They're not. 1223 00:48:53,810 --> 00:48:54,570 They're not bothered. 1224 00:48:54,570 --> 00:48:56,970 As long as something is produced, they're quite happy. 1225 00:48:56,970 --> 00:48:59,410 And there's always going to be people like that, people who don't 1226 00:48:59,410 --> 00:49:00,450 indicate on roundabouts. 1227 00:49:00,450 --> 00:49:01,690 It's just one of those things, isn't it? 1228 00:49:01,690 --> 00:49:03,830 It's like, it shouldn't be like this, but it is. 1229 00:49:04,150 --> 00:49:06,790 And I'm not sure how you know what the answer to that is. 1230 00:49:06,950 --> 00:49:10,390 I suppose Germany's just said to Google your. 1231 00:49:10,390 --> 00:49:12,670 These AI summaries at the top of the search, we're going to hold 1232 00:49:12,670 --> 00:49:15,030 you responsible for those because they're obviously useless. 1233 00:49:15,190 --> 00:49:16,470 And I think until people. 1234 00:49:16,790 --> 00:49:19,430 Until there's legal consequence for turning out slop, 1235 00:49:19,430 --> 00:49:21,270 then it's the Wild west is going to prevail. 1236 00:49:21,590 --> 00:49:24,950 In terms of professional, Something I didn't say before, like, 1237 00:49:24,950 --> 00:49:27,510 in terms of, like, professional blog posts trying to 1238 00:49:27,510 --> 00:49:31,460 get leads and stuff, certain professional indemnity insurers are 1239 00:49:31,460 --> 00:49:33,660 saying now, did you use AI in your process? 1240 00:49:33,820 --> 00:49:36,700 And you probably can't explain where you did because it's all a 1241 00:49:36,700 --> 00:49:38,940 bit of a munge between what did I do and what did the machine 1242 00:49:38,940 --> 00:49:39,260 do. 1243 00:49:39,580 --> 00:49:42,620 If somebody, you know, had an accident based on the information 1244 00:49:42,700 --> 00:49:44,900 you'd given them, then you're not. 1245 00:49:44,900 --> 00:49:46,900 They could easily say, well, we're not insured because we looked 1246 00:49:46,900 --> 00:49:49,060 on your blog and it said, right, you know, you've always got 1247 00:49:49,060 --> 00:49:51,620 to, you know, drive with just one hand and have the other one out 1248 00:49:51,620 --> 00:49:54,260 the window waving at girls that go, do you like the look out 1249 00:49:54,260 --> 00:49:54,780 the window? 1250 00:49:55,200 --> 00:49:57,880 You know, people are going to do crazy things and you won't be 1251 00:49:57,880 --> 00:49:58,720 insured for that. 1252 00:49:58,880 --> 00:49:59,560 So that's. 1253 00:49:59,560 --> 00:50:00,960 That's an issue, I think. 1254 00:50:01,280 --> 00:50:03,760 And it's just insurers always say, no, don't they? 1255 00:50:03,760 --> 00:50:04,320 It's like nine. 1256 00:50:04,560 --> 00:50:07,760 You didn't read a clause 87, mate, we can't give you any money. 1257 00:50:07,760 --> 00:50:10,680 So this is a Great big wide goal for them to say, we're not going 1258 00:50:10,680 --> 00:50:12,000 to insure you. 1259 00:50:12,320 --> 00:50:17,640 So that's another reason to eat celery and go to the gym in the 1260 00:50:17,640 --> 00:50:18,320 digital world. 1261 00:50:19,680 --> 00:50:21,720 I also think of it like this, and again, I'm interested in your 1262 00:50:21,720 --> 00:50:22,400 opinion on this. 1263 00:50:22,400 --> 00:50:25,370 I think of it as I go to a trusted source. 1264 00:50:25,690 --> 00:50:29,130 So for me, when I want to talk about the pitfalls of AI, I've come 1265 00:50:29,130 --> 00:50:30,930 to what I believe is a trusted source. 1266 00:50:30,930 --> 00:50:33,410 And you, Collet, I've seen you online, we know each other a little 1267 00:50:33,410 --> 00:50:35,170 bit and I've read your book. 1268 00:50:35,170 --> 00:50:37,090 You know, it's like you can see it's a trusted source. 1269 00:50:37,090 --> 00:50:40,290 If I want to go to someone for driving instructor training, there's 1270 00:50:40,290 --> 00:50:42,410 people I can pick, you know, or whatever. 1271 00:50:44,090 --> 00:50:47,450 I don't think I view AI as a trusted source like that because 1272 00:50:48,410 --> 00:50:51,650 it's not getting its data from all its information from all the 1273 00:50:51,650 --> 00:50:54,150 trusted places unless we give it to an ascii. 1274 00:50:54,540 --> 00:50:56,820 So I'm curious if you agree or if I'm misreading that. 1275 00:50:56,820 --> 00:50:57,420 Do you think or. 1276 00:50:57,580 --> 00:50:59,060 No, that's, that's completely correct. 1277 00:50:59,060 --> 00:51:02,820 And it's, it's going to get worse because of more AI generated 1278 00:51:02,820 --> 00:51:05,500 stuff out there, plus data poisoning. 1279 00:51:05,900 --> 00:51:07,820 There's going to be more, there's going to be more, it's going 1280 00:51:07,820 --> 00:51:09,180 to become less trustworthy. 1281 00:51:09,180 --> 00:51:11,380 Which is one of the reasons why I thought, well, I'll build my 1282 00:51:11,380 --> 00:51:11,620 own. 1283 00:51:11,620 --> 00:51:15,540 I can have a, my, my model can be in aspic and it won't be picking 1284 00:51:15,540 --> 00:51:19,860 up all these unhelpful things from as, as it starts, they call 1285 00:51:19,860 --> 00:51:21,460 it like a snake eating its own tail. 1286 00:51:21,460 --> 00:51:23,100 And eventually it's, it's it. 1287 00:51:23,170 --> 00:51:25,810 The models collapse and that's a, you know, machine language. 1288 00:51:25,810 --> 00:51:27,170 People have proven that's the case. 1289 00:51:27,650 --> 00:51:31,490 So it is going to become more untrustworthy because of, you know, 1290 00:51:31,490 --> 00:51:31,810 just. 1291 00:51:32,130 --> 00:51:34,610 We've scraped all the good data and now we're looking down the 1292 00:51:34,610 --> 00:51:35,890 back of the settee for stuff. 1293 00:51:36,050 --> 00:51:40,050 And that could be conspiracy theories and all sorts of crazy ideas. 1294 00:51:40,370 --> 00:51:42,130 So you're right to be skeptical. 1295 00:51:42,850 --> 00:51:47,730 Well, I went on to ChatGPT in preparation for this conversation. 1296 00:51:47,730 --> 00:51:50,380 I asked it, what are the current problems of AI? 1297 00:51:50,540 --> 00:51:51,740 And they gave me quite a few. 1298 00:51:51,820 --> 00:51:54,860 So I'm going to read them and I just feel free to fire back, but 1299 00:51:54,860 --> 00:51:55,580 I'll read a few of these. 1300 00:51:55,580 --> 00:51:58,860 So we've got the loss of an authentic voice, becoming generic, 1301 00:51:59,260 --> 00:52:03,660 spreading information, creating fake expertise. 1302 00:52:04,220 --> 00:52:05,100 I like that one. 1303 00:52:05,660 --> 00:52:10,380 Reduced critical thinking, definitely one I didn't consider, 1304 00:52:10,540 --> 00:52:11,500 but I quite like it. 1305 00:52:11,500 --> 00:52:12,860 Average in, average out. 1306 00:52:12,860 --> 00:52:14,620 AI trains on existing content. 1307 00:52:14,700 --> 00:52:18,150 This means it tends to produce the average of what already exists. 1308 00:52:18,150 --> 00:52:23,030 I mean, less innovation, more conformity and fewer original ideas. 1309 00:52:23,670 --> 00:52:24,950 Yes, definitely. 1310 00:52:25,030 --> 00:52:27,110 Laziness disguises productivity. 1311 00:52:27,510 --> 00:52:28,870 Can't say productivity. 1312 00:52:30,230 --> 00:52:30,710 Yes. 1313 00:52:31,510 --> 00:52:35,230 It's the new scrolling in Excel at meetings, it's the new scrolling 1314 00:52:35,230 --> 00:52:37,590 in Excel at meetings and then dodging the action points. 1315 00:52:37,590 --> 00:52:38,630 You know, that's what it is. 1316 00:52:40,230 --> 00:52:44,920 If you call our trust issues with imagery, privacy, confidentiality 1317 00:52:44,920 --> 00:52:45,480 risks. 1318 00:52:45,480 --> 00:52:49,720 And another one I liked, overconfidence in AI advice and last 1319 00:52:49,720 --> 00:52:51,600 one he said was shortcut culture. 1320 00:52:52,000 --> 00:52:52,800 So, yeah, I like those. 1321 00:52:52,800 --> 00:52:54,400 Anything you want to touch back on on those? 1322 00:52:54,880 --> 00:52:57,000 Yeah, I think they're all, they're all pretty good. 1323 00:52:57,000 --> 00:53:00,520 I think if I was going to discredit a company, you know, I'd 1324 00:53:00,520 --> 00:53:03,520 probably pay somebody on Fiverr to say I just bought a washing 1325 00:53:03,520 --> 00:53:05,880 machine from such and such a company and I hated it. 1326 00:53:05,880 --> 00:53:08,160 And, you know, so I'd put something on there, then I'd leave 1327 00:53:08,160 --> 00:53:09,600 a few fake Google reviews here. 1328 00:53:09,600 --> 00:53:10,280 Then everywhere. 1329 00:53:10,280 --> 00:53:13,400 Then I'd go on a few, you know, consumer rights sites. 1330 00:53:13,400 --> 00:53:14,080 Now this, this. 1331 00:53:14,340 --> 00:53:16,180 So I could quite easily discredit other companies. 1332 00:53:16,820 --> 00:53:20,180 I could upload somebody else's personal data and try it and tie 1333 00:53:20,180 --> 00:53:22,100 it to another therapist, let's say. 1334 00:53:22,500 --> 00:53:24,740 And I could say, oh, they've leaked somebody's personal. 1335 00:53:24,740 --> 00:53:26,340 That'd be the end of that person's career. 1336 00:53:26,340 --> 00:53:30,139 You know, you could do some quite dodgy things with it by manipulating 1337 00:53:30,139 --> 00:53:33,700 it for your own, your own benefit, which I think is not very 1338 00:53:33,700 --> 00:53:34,100 nice. 1339 00:53:34,100 --> 00:53:37,820 And just, you know, if I can't be, let's say I can't be bothered 1340 00:53:37,820 --> 00:53:39,700 to learn someone's quantum mechanics course. 1341 00:53:40,130 --> 00:53:40,490 So I ju. 1342 00:53:40,490 --> 00:53:43,890 I'm on free chat GPT because I'm skin and tight and whatever, 1343 00:53:43,890 --> 00:53:47,650 and I just upload somebody else's IP into the free model and 1344 00:53:47,650 --> 00:53:49,770 then it goes into BE training and it's probably going to be pretty 1345 00:53:49,770 --> 00:53:52,210 good because the humans made it and they cared and it was expensive 1346 00:53:52,930 --> 00:53:53,730 and that gets. 1347 00:53:53,730 --> 00:53:55,210 That gets sucked into the machine. 1348 00:53:55,210 --> 00:53:57,690 I mean, we used to think that artists were hard done by. 1349 00:53:57,690 --> 00:53:59,450 Oh, that picture looks a bit like Andy Warhol. 1350 00:53:59,450 --> 00:54:00,210 If he was still alive. 1351 00:54:00,210 --> 00:54:02,370 He'd be quite annoyed about that because they ripped off his 1352 00:54:02,450 --> 00:54:03,090 artwork. 1353 00:54:03,410 --> 00:54:06,210 And we just used to think it was, you know, artists and then it. 1354 00:54:06,210 --> 00:54:08,490 Oh, and then there was musicians when we got AI music. 1355 00:54:08,490 --> 00:54:11,570 And then we've suddenly realized actually every single person 1356 00:54:11,570 --> 00:54:14,650 with the pulse that's ever lived is now being ripped off by 1357 00:54:14,650 --> 00:54:17,730 this thing and that's not acceptable at all. 1358 00:54:18,610 --> 00:54:21,490 Should we worried about that ethically the fact that it is just 1359 00:54:21,490 --> 00:54:24,290 taking all the content that's already produced and putting it out 1360 00:54:24,290 --> 00:54:25,170 as if it's his own. 1361 00:54:25,650 --> 00:54:26,370 Absolutely. 1362 00:54:26,370 --> 00:54:30,410 So the next thing is, all the bosses are going, well, we'll replace 1363 00:54:30,410 --> 00:54:33,250 our workers and we'll keep the shareholders happy because we've 1364 00:54:33,250 --> 00:54:34,530 got to have our share price going up. 1365 00:54:34,850 --> 00:54:37,890 And then there's this idea of, well, I have universal income. 1366 00:54:37,890 --> 00:54:41,090 And then the really profitable AI companies can share their money 1367 00:54:41,330 --> 00:54:44,010 and give it to the people who are living under a railway bridge 1368 00:54:44,010 --> 00:54:46,850 in carport boxes, and then everybody will be happy. 1369 00:54:47,410 --> 00:54:49,730 So it's going to create a massive underclass. 1370 00:54:49,970 --> 00:54:53,130 When we see, like, Mythos, for instance, being taken away from us, 1371 00:54:53,130 --> 00:54:55,810 only certain governments can use it and whatever, there's going 1372 00:54:55,810 --> 00:54:57,090 to be an us and them element. 1373 00:54:57,490 --> 00:55:00,730 Like some people are going around in a Citroen C1 and some people 1374 00:55:00,730 --> 00:55:02,370 are going around in a Bugatti. 1375 00:55:03,090 --> 00:55:05,810 So this is going to be massive. 1376 00:55:05,810 --> 00:55:08,890 It's going to increase wealth inequality a lot because you're not 1377 00:55:08,890 --> 00:55:10,760 going to have as good access to good models. 1378 00:55:10,920 --> 00:55:14,440 But then again, lower, smaller models could help you come up with 1379 00:55:14,440 --> 00:55:18,280 that thing, you know, creating posters of the Scottish countryside 1380 00:55:18,280 --> 00:55:20,880 that happens to take off or whatever it is, the hand. 1381 00:55:20,880 --> 00:55:23,720 It could help you make handcrafted things better and market 1382 00:55:23,720 --> 00:55:24,360 them better. 1383 00:55:24,760 --> 00:55:26,600 So it could create opportunities. 1384 00:55:26,680 --> 00:55:28,680 But I think it's going to be the best. 1385 00:55:28,680 --> 00:55:31,320 People are going to have the best models, countries are going 1386 00:55:31,320 --> 00:55:34,240 to get further ahead than other countries and everybody else 1387 00:55:34,240 --> 00:55:36,260 sinks to the bottom, which didn't really. 1388 00:55:36,260 --> 00:55:38,460 It didn't really include wealth disparity in your. 1389 00:55:38,460 --> 00:55:39,980 I sound like Gary Stevenson there, don't I? 1390 00:55:39,980 --> 00:55:41,540 But you could see, we could see where it's going. 1391 00:55:41,540 --> 00:55:43,980 If you've got the best, quickest, most perfect models and 1392 00:55:43,980 --> 00:55:46,980 everybody else has got a rubber band, then it's not going 1393 00:55:46,980 --> 00:55:48,180 to end well, you know. 1394 00:55:49,140 --> 00:55:52,620 All right, I thought it might be interesting to share with you 1395 00:55:52,620 --> 00:55:57,540 a few of the ways I use AI let you take it almost. 1396 00:55:57,540 --> 00:55:58,740 And I'm prepared for this. 1397 00:55:58,740 --> 00:55:59,860 I'm prepared for that. 1398 00:55:59,940 --> 00:56:01,140 I think I'm prepared for this. 1399 00:56:01,630 --> 00:56:04,030 So you kind of touched on it earlier about the content creation 1400 00:56:04,030 --> 00:56:07,110 side, and I do use it to help me with that, but I'll tell you kind 1401 00:56:07,110 --> 00:56:07,990 of how I'd do it. 1402 00:56:07,990 --> 00:56:10,590 So an email might be a good example, or a Facebook post. 1403 00:56:10,910 --> 00:56:14,310 I will generally write it out and then I'll put it into chat G, 1404 00:56:14,310 --> 00:56:18,550 B, T. And I will ask it, do you know what the goal of this email 1405 00:56:18,550 --> 00:56:18,910 is? 1406 00:56:19,150 --> 00:56:21,590 And I find that works really well, because if it doesn't, then 1407 00:56:21,590 --> 00:56:24,110 Obviously there's a flaw in the email somewhere and if it does, 1408 00:56:24,110 --> 00:56:27,870 then great, and then I'll ask it we alluded to before to critique 1409 00:56:27,870 --> 00:56:31,250 it, tell me what you would change about this and why, and then 1410 00:56:31,250 --> 00:56:32,610 kind of go from there. 1411 00:56:32,690 --> 00:56:33,770 So what are your thoughts? 1412 00:56:33,770 --> 00:56:34,770 Me using it like that? 1413 00:56:34,930 --> 00:56:38,050 I think those two are really good because you're saving your attention 1414 00:56:38,130 --> 00:56:39,770 for the bits to add the most value. 1415 00:56:39,770 --> 00:56:44,530 The people, the, the, the secret Terry Ness about it is still 1416 00:56:44,530 --> 00:56:47,210 preserved and you're still, you know, you're still looking at 1417 00:56:47,210 --> 00:56:48,770 it and not just waving it through. 1418 00:56:48,850 --> 00:56:50,130 I think that's pretty. 1419 00:56:50,450 --> 00:56:51,810 And it gives you a fresh perspective. 1420 00:56:51,810 --> 00:56:56,050 There's only so many times you can explain you know something you 1421 00:56:56,050 --> 00:56:57,290 know, and you, oh, not this again. 1422 00:56:57,290 --> 00:56:59,930 It's like magazines, you think, Beginner's Guide to Carp Fishing, 1423 00:56:59,930 --> 00:57:01,370 you think, oh, we had that last January. 1424 00:57:02,010 --> 00:57:04,050 You know, you've got to keep coming up with these new ways of 1425 00:57:04,050 --> 00:57:05,210 doing things and I think that's. 1426 00:57:05,290 --> 00:57:08,290 It stops you falling out of love with your industry with some 1427 00:57:08,290 --> 00:57:10,650 of those grindy bits that you just have to do. 1428 00:57:10,650 --> 00:57:12,210 So I don't see a problem with that at all. 1429 00:57:12,210 --> 00:57:14,770 And you obviously care about what you're, what you're putting 1430 00:57:14,770 --> 00:57:14,970 out. 1431 00:57:14,970 --> 00:57:16,730 I think that's, that's a good thing. 1432 00:57:17,210 --> 00:57:18,010 Now what's your. 1433 00:57:18,330 --> 00:57:20,490 I had a suspicion that one will be okay. 1434 00:57:20,490 --> 00:57:22,330 I'm less certain about the next two. 1435 00:57:24,370 --> 00:57:24,690 Yeah. 1436 00:57:25,010 --> 00:57:27,330 So the deep research aspect of ChatGPT. 1437 00:57:27,970 --> 00:57:31,570 So I doing the example I use for this is another podcast I'm signing 1438 00:57:31,570 --> 00:57:34,570 up called the Yorkshire Podcast and I've got to do some deep 1439 00:57:34,570 --> 00:57:37,650 research on some topics for that and it's brought back some wonderful 1440 00:57:37,650 --> 00:57:38,050 stuff. 1441 00:57:38,289 --> 00:57:41,610 The caveat I'll put in is I make sure that I go back and check 1442 00:57:41,610 --> 00:57:45,650 all the sources and it's actually brought me some wonderful 1443 00:57:45,730 --> 00:57:48,250 other exhausts I've gone and explored on different websites and 1444 00:57:48,250 --> 00:57:49,330 books and stuff like that. 1445 00:57:49,770 --> 00:57:53,090 But I suppose primarily that deep research aspect on a specific 1446 00:57:53,090 --> 00:57:54,410 topic that you can do with them. 1447 00:57:55,050 --> 00:57:55,530 Yeah. 1448 00:57:55,530 --> 00:57:59,010 So I would probably say, well, why using AI anyway? 1449 00:57:59,010 --> 00:58:01,970 Why not just use a regular search, you know, why not go like 1450 00:58:01,970 --> 00:58:04,130 some of the machine learning academic papers. 1451 00:58:04,130 --> 00:58:07,970 I could go to the acm, the American Computer, you know, foundation, 1452 00:58:07,970 --> 00:58:08,250 whatever. 1453 00:58:08,250 --> 00:58:11,130 And I can go and just do a bog standard search. 1454 00:58:11,210 --> 00:58:13,850 I'll go old school, you know, not back to pen and paper, but just 1455 00:58:13,850 --> 00:58:17,100 the next bit along and I could, I don't have to do an AI search, 1456 00:58:17,100 --> 00:58:19,420 I could just do a normal search and I'd have a lot more trust 1457 00:58:19,420 --> 00:58:19,740 in that. 1458 00:58:19,740 --> 00:58:24,980 What's coming back is is relevant when I, I've gone to YouTube 1459 00:58:24,980 --> 00:58:28,060 and I've got like 3,000 YouTube videos of people talking 1460 00:58:28,060 --> 00:58:32,300 about AI from like professors and credible sources and conference 1461 00:58:32,300 --> 00:58:33,140 proceedings and stuff. 1462 00:58:33,140 --> 00:58:34,660 So I'm fairly confident it's good stuff. 1463 00:58:34,900 --> 00:58:37,220 And, and I'll go and get things from there and I'll say, well, 1464 00:58:37,220 --> 00:58:39,980 if you make a point, you've got to cross reference back to the 1465 00:58:39,980 --> 00:58:43,360 transcript that it came from so I can go and check it personally. 1466 00:58:43,360 --> 00:58:46,920 So I know what I put in and it just kind of stops the scrolling 1467 00:58:46,920 --> 00:58:50,040 bit where you have to find page 87 and get the little bit that's 1468 00:58:50,040 --> 00:58:50,840 referenced to it. 1469 00:58:50,840 --> 00:58:54,160 So I do, I do that quite a lot to check what I've got. 1470 00:58:54,160 --> 00:58:57,440 So because I've built my own local LLM, I've built my own little 1471 00:58:57,440 --> 00:59:01,520 database that helps me search the useful things that I found just. 1472 00:59:01,520 --> 00:59:03,120 And that just uses normal search. 1473 00:59:03,120 --> 00:59:06,520 It's not using AI search, it's just literally looking for one word 1474 00:59:07,000 --> 00:59:09,440 in a piece of text, like finding a place and saying, hey, 1475 00:59:09,440 --> 00:59:10,760 I think this is worth you looking at. 1476 00:59:11,350 --> 00:59:14,830 So again, I guess removing the AI as much as possible is a really 1477 00:59:14,830 --> 00:59:15,270 good idea. 1478 00:59:15,270 --> 00:59:18,990 It's good for blue sky thinking, but it's not like you say, 1479 00:59:18,990 --> 00:59:21,150 it's not so good for the fact stuff, but it can still help with 1480 00:59:21,150 --> 00:59:22,870 the heavy lifting of the fact checking. 1481 00:59:23,350 --> 00:59:27,390 And if it starts to get three wrong in a document, maybe if it 1482 00:59:27,390 --> 00:59:29,950 starts to get quite a lot wrong, maybe just start it again 1483 00:59:29,950 --> 00:59:32,910 and have another go and say, I found these factual errors. 1484 00:59:32,910 --> 00:59:33,990 You need to try harder. 1485 00:59:33,990 --> 00:59:37,550 Just go and find me three examples of something at a time. 1486 00:59:37,550 --> 00:59:39,910 Don't try and find me 20 because you're going to be like a 1487 00:59:39,910 --> 00:59:40,830 giraffe on roller skates. 1488 00:59:40,830 --> 00:59:41,910 It's not going to end well. 1489 00:59:42,090 --> 00:59:44,370 Just go find three good ones and then we'll find another three 1490 00:59:44,370 --> 00:59:44,770 good ones. 1491 00:59:44,770 --> 00:59:48,090 And like a quiz show, bank good answers as you go and then you've 1492 00:59:48,090 --> 00:59:51,410 got something you can that's worthwhile at the end. 1493 00:59:51,410 --> 00:59:52,490 You haven't had to check too much. 1494 00:59:52,490 --> 00:59:55,450 So your eye for detail hasn't been worn down to nothing. 1495 00:59:56,010 --> 00:59:57,290 Break it into little tasks. 1496 00:59:57,849 --> 00:59:58,570 I like that. 1497 00:59:58,570 --> 01:00:03,050 And the first thing you said was you would ask me why and it came 1498 01:00:03,050 --> 01:00:05,210 in my brain tray, which is just lazy. 1499 01:00:05,370 --> 01:00:06,130 That's all it is. 1500 01:00:06,130 --> 01:00:08,210 It's like I can type it in and bugger off for 10 minutes and come 1501 01:00:08,210 --> 01:00:09,130 back and it's done it for me. 1502 01:00:09,360 --> 01:00:12,640 You know, there's a lot less, maybe not a lot less work, but it 1503 01:00:12,640 --> 01:00:14,480 feels, you know what I mean? 1504 01:00:15,440 --> 01:00:17,280 And I'm being lazy by doing that. 1505 01:00:17,680 --> 01:00:18,040 Yeah. 1506 01:00:18,040 --> 01:00:20,480 You've got to ask yourself which is more irritating, staring 1507 01:00:20,480 --> 01:00:25,000 at a blank canvas when you're lost words or curating 10,000 words 1508 01:00:25,000 --> 01:00:27,440 of stuff, of which half of it needs the chop. 1509 01:00:27,600 --> 01:00:30,240 You know, you've got to pick your poison, really, haven't you? 1510 01:00:30,640 --> 01:00:33,440 Sometimes if you don't know anything about the subject, then 1511 01:00:33,440 --> 01:00:35,800 it might be better off going back to the blank canvas, because 1512 01:00:35,800 --> 01:00:37,040 then you're learning about the subject. 1513 01:00:37,040 --> 01:00:39,560 And then when you get some information later on, you can check 1514 01:00:39,560 --> 01:00:42,240 its provenance and validity a bit more than if you were. 1515 01:00:42,480 --> 01:00:43,680 Didn't know anything about it. 1516 01:00:43,680 --> 01:00:46,400 But if I started writing something about, you know, how to 1517 01:00:46,400 --> 01:00:50,240 remap a diesel car when the ad blue's gone weird, that whatever 1518 01:00:50,240 --> 01:00:52,320 it came out with, I've got no idea what's. 1519 01:00:52,400 --> 01:00:53,360 What's happening there. 1520 01:00:53,360 --> 01:00:56,320 It's like, yeah, sometimes going a bit slower, you might. 1521 01:00:56,320 --> 01:00:58,800 You know, a friend of mine, Dino, sort of says, well, sometimes 1522 01:00:58,800 --> 01:01:01,960 when you go a bit slower, you end up being faster in the end, you 1523 01:01:01,960 --> 01:01:04,000 know, you have a smoother ride, even though it's a little bit 1524 01:01:04,000 --> 01:01:04,360 slower. 1525 01:01:04,440 --> 01:01:06,440 So that's worth bearing in mind. 1526 01:01:07,080 --> 01:01:08,000 I love dinof. 1527 01:01:08,000 --> 01:01:09,480 Dinof said it's worth listening to. 1528 01:01:10,520 --> 01:01:13,720 I'm gonna throw one more because I think this is a bit. 1529 01:01:14,200 --> 01:01:15,240 Maybe a little bit different. 1530 01:01:15,480 --> 01:01:18,760 I use it for a daily challenge, so every day I ask it 1531 01:01:18,760 --> 01:01:20,200 to challenge me to do something. 1532 01:01:20,200 --> 01:01:21,640 Now, there's some parameters there. 1533 01:01:21,960 --> 01:01:25,840 So ask me to score my resistance, score my time for the 1534 01:01:25,840 --> 01:01:30,070 day, score my, like, where my brain is, you know, how, you know, 1535 01:01:30,070 --> 01:01:30,750 can I do something? 1536 01:01:30,750 --> 01:01:32,430 Public faces, you know, it's like that. 1537 01:01:32,590 --> 01:01:35,190 And every day it just gives me, like, a little challenge, and 1538 01:01:35,190 --> 01:01:36,150 it's giving me challenges. 1539 01:01:36,150 --> 01:01:38,350 Like, I want you to put your phone in a different room for an 1540 01:01:38,350 --> 01:01:38,830 hour today. 1541 01:01:39,070 --> 01:01:40,750 And then it gives me one's like, I want you to. 1542 01:01:40,990 --> 01:01:41,350 This. 1543 01:01:41,350 --> 01:01:42,710 This is where you came in, Collette. 1544 01:01:42,710 --> 01:01:44,750 I want you to message someone to come on your podcast that you 1545 01:01:44,750 --> 01:01:45,670 wouldn't normally message. 1546 01:01:45,670 --> 01:01:47,790 And I'm like, yes, this is what I want to do. 1547 01:01:48,190 --> 01:01:48,790 And it's just. 1548 01:01:48,790 --> 01:01:49,710 That's absolutely brilliant. 1549 01:01:49,710 --> 01:01:52,430 Yeah, so, yeah, so that was my. 1550 01:01:53,210 --> 01:01:55,050 My fun little daily challenge anymore. 1551 01:01:55,690 --> 01:01:56,730 Yeah, I think that's nice. 1552 01:01:56,730 --> 01:01:59,250 You know, there are, like I say there are some brilliant aspects 1553 01:01:59,250 --> 01:01:59,490 to it. 1554 01:01:59,490 --> 01:02:01,730 If you're getting something out of it that's useful, carry on 1555 01:02:01,730 --> 01:02:03,290 doing it, you know, but why not? 1556 01:02:04,970 --> 01:02:07,090 I was going to say something else then, and I can't remember what 1557 01:02:07,090 --> 01:02:07,530 it was. 1558 01:02:08,330 --> 01:02:09,690 No, sorry, it's gone. 1559 01:02:10,410 --> 01:02:11,770 One of those bits is Gone. 1560 01:02:12,970 --> 01:02:14,370 No, I appreciate that. 1561 01:02:14,370 --> 01:02:17,290 It's one that I think is very much a Terry thing and I enjoy it. 1562 01:02:17,290 --> 01:02:19,410 It just keeps me check my day interest a little bit. 1563 01:02:19,410 --> 01:02:22,210 But is there anything else you want to touch back on or any. 1564 01:02:22,290 --> 01:02:23,890 Anything else you'd like to share about this? 1565 01:02:24,610 --> 01:02:25,570 I've remembered it. 1566 01:02:25,570 --> 01:02:26,250 I remembered it. 1567 01:02:26,250 --> 01:02:29,490 So like you've asked me, we were talking earlier about big data 1568 01:02:29,810 --> 01:02:33,530 and one of the things we can do in the UK is we can ask for big 1569 01:02:33,530 --> 01:02:35,410 companies to give us through gdpr. 1570 01:02:35,410 --> 01:02:37,850 We can say give me my stuff and then delete it. 1571 01:02:37,850 --> 01:02:38,690 Thank you very much. 1572 01:02:39,010 --> 01:02:41,490 And there was a couple of guys, a couple of academics did a 1573 01:02:41,490 --> 01:02:45,050 talk in Berlin and one of them, one of the guys wasn't very 1574 01:02:45,050 --> 01:02:48,530 good at dating and he, it had all gone wrong for like 10 years. 1575 01:02:48,530 --> 01:02:51,490 He had a catalog of romantic disasters and of course all that 1576 01:02:51,490 --> 01:02:52,250 data was stored. 1577 01:02:52,250 --> 01:02:54,690 So he said to the, he said to various companies, please give me 1578 01:02:54,690 --> 01:02:55,770 everything you know about me. 1579 01:02:56,010 --> 01:02:58,530 And of course that was decades, years and years and years 1580 01:02:58,530 --> 01:03:00,010 and years of very detailed information. 1581 01:03:00,170 --> 01:03:04,170 So if you did want to look back at yourself and improve yourself, 1582 01:03:04,250 --> 01:03:07,530 you could download that data, actually use that surveilled data 1583 01:03:07,530 --> 01:03:11,170 for something good for you, ask it about your strengths and weaknesses, 1584 01:03:11,170 --> 01:03:14,880 how you've performed over time and come up with a plan which is 1585 01:03:14,880 --> 01:03:18,280 very accurate, you know, based on you, you give it your goals. 1586 01:03:18,280 --> 01:03:21,640 It's a fairly small amount of data compared to what they can cope 1587 01:03:21,640 --> 01:03:21,920 with. 1588 01:03:22,240 --> 01:03:23,880 You could get some really good results out of that. 1589 01:03:23,880 --> 01:03:26,520 He got so, you know, so one chap found out the first thing he 1590 01:03:26,520 --> 01:03:29,640 ever typed in on YouTube and then three days later he had a binge 1591 01:03:29,640 --> 01:03:33,920 watching day in 2013 and it did all this stuff which was brilliant. 1592 01:03:33,920 --> 01:03:37,080 So if you wanted to, that would be a good use of all that horrific 1593 01:03:37,080 --> 01:03:37,320 data. 1594 01:03:37,320 --> 01:03:40,560 People have got to think a dreadful thing. 1595 01:03:40,560 --> 01:03:40,840 Poor. 1596 01:03:40,840 --> 01:03:41,520 I'm going to read. 1597 01:03:41,770 --> 01:03:43,610 I was way too scared to do that. 1598 01:03:44,250 --> 01:03:46,250 Anything else you want to touch on today, Colette? 1599 01:03:46,330 --> 01:03:49,050 I think I'd just say don't, you know, don't be too disheartened 1600 01:03:49,050 --> 01:03:49,290 with it. 1601 01:03:49,290 --> 01:03:53,410 It's in, you know, if you look at Dario Amadei of Anthropic and 1602 01:03:53,410 --> 01:03:54,570 he's going, this is going to be a. 1603 01:03:54,730 --> 01:03:57,730 He didn't actually use the word bloodbath, white collar bloodbath, 1604 01:03:57,730 --> 01:04:00,810 but journalists did when they reported on him saying like 50% of 1605 01:04:00,810 --> 01:04:03,810 jobs are going to be gone and it's in their interest to make it 1606 01:04:03,810 --> 01:04:06,250 look really, really powerful and that it can take over the world. 1607 01:04:06,250 --> 01:04:09,250 Even though we all know sat here in our little chairs that it 1608 01:04:09,250 --> 01:04:09,610 can't. 1609 01:04:10,640 --> 01:04:14,200 Don't let the fear kind of really taint your day and make you 1610 01:04:14,200 --> 01:04:15,600 feel despondent about the future. 1611 01:04:15,600 --> 01:04:18,720 There's lots of things we can do and if you start de Googling and 1612 01:04:18,880 --> 01:04:21,480 looking around, there's lots of things you can do to feel better 1613 01:04:21,480 --> 01:04:22,080 about it. 1614 01:04:22,240 --> 01:04:25,440 And if you find something that helps you, don't feel ashamed, use 1615 01:04:25,440 --> 01:04:25,760 it. 1616 01:04:26,080 --> 01:04:29,040 If you see somebody who is using it because it's not their first 1617 01:04:29,040 --> 01:04:30,560 language, bite your tongue. 1618 01:04:30,560 --> 01:04:32,160 Don't have a, don't have a pop at them. 1619 01:04:32,160 --> 01:04:34,640 You know they're using it probably in good faith. 1620 01:04:35,040 --> 01:04:36,430 So try. 1621 01:04:36,510 --> 01:04:38,910 Just go with your eyes open and don't miss out. 1622 01:04:39,550 --> 01:04:42,510 And don't be a lazy slot merchant that needs a punch in the 1623 01:04:42,510 --> 01:04:43,030 face either. 1624 01:04:43,030 --> 01:04:45,070 I think that's, that's what I would say. 1625 01:04:46,270 --> 01:04:49,870 I think I've never done this on my podcast before where you see 1626 01:04:49,870 --> 01:04:51,710 it on some of them, where they put the clip at the start. 1627 01:04:52,190 --> 01:04:54,190 I think that would be the clip I'll put at the start. 1628 01:04:54,190 --> 01:04:56,910 Don't be lazy slot merchant that needs a punch in the face. 1629 01:04:57,230 --> 01:04:58,630 Oh, yeah, yeah, yeah, definitely. 1630 01:04:58,630 --> 01:05:01,190 It's like Internet marketing on steroids in terms of Wild west, 1631 01:05:01,190 --> 01:05:01,590 isn't it? 1632 01:05:01,590 --> 01:05:04,720 It's just like, bro, have a word. 1633 01:05:06,800 --> 01:05:07,400 All right, cool. 1634 01:05:07,400 --> 01:05:08,360 Well, big thank you for today. 1635 01:05:08,360 --> 01:05:09,200 I really enjoyed it. 1636 01:05:09,440 --> 01:05:13,480 I've definitely got a slightly different take on it now and learn 1637 01:05:13,480 --> 01:05:14,680 some stuff, so I really appreciate that. 1638 01:05:14,680 --> 01:05:17,120 Do you want to take a moment to tell people where they can find 1639 01:05:17,120 --> 01:05:20,800 you that's not going to cause any data problems? 1640 01:05:21,520 --> 01:05:22,080 Yeah. 1641 01:05:22,160 --> 01:05:26,480 So I kind of decided to do AI 15 months ago before it's turned 1642 01:05:26,480 --> 01:05:28,440 into a toxic fire pit of hell. 1643 01:05:28,440 --> 01:05:29,120 But there we are. 1644 01:05:29,470 --> 01:05:32,590 And you can find me@cleverclogsai.com and there's 1645 01:05:32,590 --> 01:05:34,510 a couple of things on there that are useful. 1646 01:05:34,750 --> 01:05:37,190 I've done a big 10,000 word blog post. 1647 01:05:37,190 --> 01:05:39,430 There's no opt in or anything about data privacy. 1648 01:05:39,430 --> 01:05:42,390 So if that, some of that was interesting, you can look at how 1649 01:05:42,390 --> 01:05:43,790 you can start protecting yourself. 1650 01:05:44,110 --> 01:05:46,870 And some of my experiences that I had with that to try and make 1651 01:05:46,870 --> 01:05:48,590 it a bit human and not too nerdy. 1652 01:05:48,590 --> 01:05:51,990 And I'm also, I've also got some things called AI transparency 1653 01:05:51,990 --> 01:05:56,030 packs, which are, you know, if you're honest about using AI and 1654 01:05:56,030 --> 01:05:58,270 saying what you're doing, I think people will cut you some slack. 1655 01:05:58,270 --> 01:06:00,710 And if they can sense you're using it and you've not said that 1656 01:06:00,710 --> 01:06:03,210 you are like you say, you fall down the cracks in the middle. 1657 01:06:04,010 --> 01:06:04,890 So there's. 1658 01:06:05,050 --> 01:06:08,370 I sort of offer some PDFs for different industries so people can 1659 01:06:08,370 --> 01:06:10,890 say they're using AI responsibly for the good stuff for 1660 01:06:10,890 --> 01:06:13,770 the liver transplant, and they're not using it for the slop. 1661 01:06:14,010 --> 01:06:16,690 And that can help have those, because we're going to, as a society, 1662 01:06:16,690 --> 01:06:18,250 we're going to have to have those conversations. 1663 01:06:18,250 --> 01:06:20,970 So that's, that's the other thing on there on the website. 1664 01:06:21,290 --> 01:06:23,610 There's probably loads of typos on it, and I hate editing it. 1665 01:06:23,610 --> 01:06:26,090 Editing your own website is probably the worst thing in the universe, 1666 01:06:26,570 --> 01:06:27,250 but it's there. 1667 01:06:27,250 --> 01:06:28,250 You can go and have a look. 1668 01:06:30,050 --> 01:06:31,410 Well, a big thank you for joining me today. 1669 01:06:31,410 --> 01:06:32,210 It's been a pleasure. 1670 01:06:32,930 --> 01:06:33,690 Thanks, Terry. 1671 01:06:33,690 --> 01:06:34,450 Thanks, everybody. 1672 01:06:36,370 --> 01:06:39,370 The instructor podcast with Terry Cook, talking. 1673 01:06:39,370 --> 01:06:42,370 With leaders, innovators, experts and game changers. 1674 01:06:42,370 --> 01:06:43,890 About what drives them.