1 00:00:00,040 --> 00:00:03,280 Saying a computer can think it's a bit like saying a submarine 2 00:00:03,280 --> 00:00:07,520 can swim. Welcome to Business Analysis 3 00:00:07,520 --> 00:00:09,520 Live. I'm your host, Susan Moore, 4 00:00:09,520 --> 00:00:12,840 community engagement manager with the International Institute 5 00:00:12,880 --> 00:00:16,000 of Business Analysis. I explore topics with industry 6 00:00:16,000 --> 00:00:18,840 guests about the work of business analysis professionals 7 00:00:18,920 --> 00:00:21,960 and how their work helps organizations achieve better 8 00:00:21,960 --> 00:00:24,040 outcomes. And Yep, we take your questions 9 00:00:24,040 --> 00:00:25,440 too. Thanks for listening. 10 00:00:25,520 --> 00:00:31,720 Let's get started. AI governance. 11 00:00:32,320 --> 00:00:36,000 Maybe you've heard something about this and maybe you've got 12 00:00:36,000 --> 00:00:37,840 some thoughts on what you think it is. 13 00:00:39,400 --> 00:00:43,680 I'm going to challenge what you might be thinking about, what AI 14 00:00:43,680 --> 00:00:48,960 governance is, what it could be. I am talking with a really 15 00:00:48,960 --> 00:00:52,560 interesting speaker and author of a new book. 16 00:00:52,560 --> 00:00:56,000 We're going to tell you about that in a minute, who brings a 17 00:00:56,040 --> 00:01:00,400 different perspective and some different thinking about AI 18 00:01:00,400 --> 00:01:02,520 governance. And I want you to hear it 19 00:01:02,520 --> 00:01:06,560 because it has a lot of ramifications for the work that 20 00:01:06,560 --> 00:01:09,120 we do as business analysis professionals. 21 00:01:09,120 --> 00:01:12,760 So let me bring to the stage Chris Ambler. 22 00:01:12,760 --> 00:01:14,920 Hey, Chris. Good afternoon. 23 00:01:15,360 --> 00:01:19,080 Yes, good to see you. Good to see you, Chris. 24 00:01:19,080 --> 00:01:22,320 So far, every time that you and I have had an opportunity to 25 00:01:22,320 --> 00:01:25,200 talk, it is like the most incredible conversation I've 26 00:01:25,200 --> 00:01:29,400 had. And so we are going to we, you 27 00:01:29,400 --> 00:01:31,760 know what? I don't want us to try to hold 28 00:01:31,760 --> 00:01:34,240 ourselves up to a standard because I know that today we're 29 00:01:34,240 --> 00:01:35,840 going to have a great conversation. 30 00:01:36,080 --> 00:01:41,280 Before we dive into this topic, I want you to tell people a 31 00:01:41,280 --> 00:01:44,360 little bit about yourself. Absolutely. 32 00:01:45,120 --> 00:01:48,400 The well, the first thing I got to admit to is that I'm not a 33 00:01:48,400 --> 00:01:52,920 business analyst. My background is basically 34 00:01:52,920 --> 00:01:56,920 around quality assurance and I've been in the IT industry 35 00:01:56,920 --> 00:02:02,800 for, well, I think it's just coming up to 47 years working 36 00:02:02,800 --> 00:02:06,440 all the way through working in testing quality assurance, right 37 00:02:06,440 --> 00:02:12,320 through into senior management. I was the European QA Director 38 00:02:12,320 --> 00:02:16,880 for Electronic Arts in the video games industry, and then I was 39 00:02:16,880 --> 00:02:21,040 the European QA Director for Microsoft Game Studios. 40 00:02:21,840 --> 00:02:26,000 So I spent probably the last 12 years or so of my working life 41 00:02:27,080 --> 00:02:29,400 working in the video games industry. 42 00:02:30,160 --> 00:02:35,440 And that's where I sort of came up with my interest in the way 43 00:02:35,440 --> 00:02:39,760 that people relate to computing and the way that computing 44 00:02:39,760 --> 00:02:44,440 relates to people. At the ripe old age of 55, I 45 00:02:44,440 --> 00:02:50,680 decided that I'd had enough of the day job and I really wanted 46 00:02:50,680 --> 00:02:54,760 to look at retirement. But in the sense that retirement 47 00:02:54,760 --> 00:02:57,160 being what I want to do when I want to do it. 48 00:02:57,800 --> 00:03:02,640 And I wanted to started to start to basically pursue my interest. 49 00:03:02,920 --> 00:03:06,840 And my interest then was starting to look at artificial 50 00:03:06,840 --> 00:03:12,880 intelligence. And I obtained a degree in 51 00:03:12,880 --> 00:03:17,960 computer science back in the 90s and decided to supplement that 52 00:03:17,960 --> 00:03:24,000 when I retired with a degree in psychology because I felt that 53 00:03:24,000 --> 00:03:32,440 those two things together were going to give me a good idea of 54 00:03:32,440 --> 00:03:34,880 how I was going to start to affect the world. 55 00:03:34,880 --> 00:03:37,040 And the first thing I was looking at was how it was going 56 00:03:37,040 --> 00:03:41,760 to affect people's jobs. And over a period of probably 57 00:03:41,760 --> 00:03:46,440 about a year or so, I've started to hone down my ideas a bit more 58 00:03:46,480 --> 00:03:49,160 in that wasn't really the key problem. 59 00:03:49,160 --> 00:03:52,640 It was more about what AI was capable of before we started 60 00:03:52,640 --> 00:03:54,800 thinking about what jobs it could take away. 61 00:03:57,280 --> 00:04:00,920 And it was suggested to me in the end that I really ought to 62 00:04:00,920 --> 00:04:04,480 write a book about this. So I spent, I've spent about the 63 00:04:04,480 --> 00:04:09,760 last two years researching and writing my book, the site The 64 00:04:09,760 --> 00:04:14,520 Psychology of AI Decision Making, which came out in 65 00:04:14,520 --> 00:04:17,720 October last year with BCS Publishing. 66 00:04:20,200 --> 00:04:26,880 Since then, I've been holding ideas, looking at how how AI 67 00:04:26,880 --> 00:04:29,880 actually makes decisions. Now I'm, I'm not looking at 68 00:04:30,640 --> 00:04:33,480 decisions, operational type of decisions. 69 00:04:33,480 --> 00:04:37,760 I'm looking more at ethical and moral types of decisions and how 70 00:04:37,760 --> 00:04:41,960 we're going to make sure that AI does this correctly, does it in 71 00:04:41,960 --> 00:04:44,040 a way that we're going to be able to control. 72 00:04:45,120 --> 00:04:49,160 And then once I've gone through that in the book, what I 73 00:04:49,160 --> 00:04:54,000 actually did was I looked at some empirical psychological 74 00:04:54,000 --> 00:04:58,480 experimentation from the 50s and the 60s when there wasn't really 75 00:04:58,480 --> 00:05:04,840 such a thing as ethics. I decided to use some of the 76 00:05:04,840 --> 00:05:09,600 experiments that they did back then and then run them against 77 00:05:09,600 --> 00:05:12,880 AI systems to see if I was getting the same sort of 78 00:05:12,880 --> 00:05:17,880 responses as the psychologists from the 50s and the 60s were 79 00:05:17,880 --> 00:05:19,800 getting when they did this with real people. 80 00:05:20,400 --> 00:05:22,680 And I came up with some really interesting results. 81 00:05:24,680 --> 00:05:27,320 I won't talk about those now. You can read the book and find 82 00:05:27,320 --> 00:05:31,480 out about those. But it meant that I, Del, I 83 00:05:31,560 --> 00:05:36,320 delved very deeply into what decisions, what meant, what 84 00:05:36,320 --> 00:05:40,840 decisions were, and how as humans we actually make those 85 00:05:40,840 --> 00:05:45,640 decisions. And what I actually came up with 86 00:05:45,640 --> 00:05:48,720 after two years of research and talking to lots and lots of 87 00:05:48,720 --> 00:05:50,520 people, is that I don't think we really know. 88 00:05:51,240 --> 00:05:55,480 Nobody really understands in any sort of depth about decisions. 89 00:05:55,480 --> 00:05:59,160 And about certainly from an ethical perspective and from a 90 00:05:59,160 --> 00:06:03,200 moral perspective, there's so many variables in the way that 91 00:06:03,200 --> 00:06:06,240 we think about these things that's going to be really, 92 00:06:06,240 --> 00:06:11,120 really hard to try and emulate us or or simulate those within 93 00:06:11,120 --> 00:06:15,280 AI systems. So what I want to try and do now 94 00:06:15,280 --> 00:06:20,600 is try and talk to people to help them understand exactly 95 00:06:20,600 --> 00:06:24,440 what decision making or ethical decision making is all about. 96 00:06:26,800 --> 00:06:30,960 And I honestly believe that bringing the communities of QA, 97 00:06:32,040 --> 00:06:37,440 business analysts, ethicists, data data analysts, educators, 98 00:06:38,040 --> 00:06:40,400 bringing all of those people together is going to be a very 99 00:06:40,400 --> 00:06:42,160 necessary part of what we're going to do. 100 00:06:43,680 --> 00:06:45,600 I, I agree. I am. 101 00:06:46,640 --> 00:06:50,520 And what is interesting is that the conversations that I've been 102 00:06:50,520 --> 00:06:56,680 part of around AI governance have talked a lot about it kind 103 00:06:56,680 --> 00:06:59,720 of the guts of how you make AI governance work. 104 00:06:59,720 --> 00:07:04,600 But I have not heard a lot of a lot of discussion around how, 105 00:07:05,000 --> 00:07:08,880 how are we thinking about decision making with artificial 106 00:07:08,880 --> 00:07:11,680 intelligence. So first I want to make sure 107 00:07:11,680 --> 00:07:15,720 that we that we plug the book because it's not that your, your 108 00:07:15,720 --> 00:07:22,280 book is kind of suggesting here, I'm going to put this up in 109 00:07:22,280 --> 00:07:23,360 front of us. Hello. 110 00:07:23,880 --> 00:07:25,520 So there's the book. It's not that your book is 111 00:07:25,520 --> 00:07:30,280 suggesting A governance, but it lays the background for a lot of 112 00:07:30,280 --> 00:07:35,000 important pieces that you think need to be considered when 113 00:07:35,000 --> 00:07:42,400 thinking about how how should we, how would we allow AI to 114 00:07:42,560 --> 00:07:43,960 make decisions? It's. 115 00:07:44,120 --> 00:07:45,440 Would that be a good way to say it? 116 00:07:46,320 --> 00:07:50,760 Yeah, I think if you wanted a good analogy for this, if you 117 00:07:50,800 --> 00:07:55,840 imagine that you're teaching somebody to drive a car, do you 118 00:07:55,840 --> 00:07:58,880 need to teach them what's going on under the hood? 119 00:08:01,160 --> 00:08:04,120 My opinion is that if you know what's going on under the hood, 120 00:08:04,120 --> 00:08:07,520 it makes driving the car more sensible and and means that you 121 00:08:07,520 --> 00:08:09,880 can probably control the car better. 122 00:08:11,040 --> 00:08:14,520 And I think right now there's lots of people out there with AI 123 00:08:14,520 --> 00:08:18,920 systems that are showing us the driving of an AI system without 124 00:08:18,920 --> 00:08:21,440 showing us under the hood of what it's capable of doing. 125 00:08:22,440 --> 00:08:24,520 That's a great analogy. And that's the bit that's 126 00:08:24,520 --> 00:08:26,560 missing. Yeah, yeah. 127 00:08:27,680 --> 00:08:30,840 Because when you know, as you and I have been getting ready 128 00:08:30,840 --> 00:08:34,440 for today's session, you have been talking about AI. 129 00:08:34,440 --> 00:08:38,880 Less is like a cool tool that comes up with answers, but it's 130 00:08:38,880 --> 00:08:43,000 software. It's software that needs good, 131 00:08:43,120 --> 00:08:48,120 rigorous testing and governance. Well, if I go back to my days in 132 00:08:48,120 --> 00:08:53,320 the video games industry, I went into the video games industry 133 00:08:53,320 --> 00:08:55,920 with absolutely no interest in video games. 134 00:08:56,560 --> 00:08:58,360 I had a son that was really interested. 135 00:08:59,360 --> 00:09:03,440 Well, I have a son that's very interested, but I'm not 136 00:09:03,480 --> 00:09:07,520 particularly interested and I don't really play video games. 137 00:09:07,520 --> 00:09:10,240 And when I do try and play video games, I'm rubbish. 138 00:09:10,240 --> 00:09:14,880 I just can't, can't do it. What I do know is software, and 139 00:09:14,880 --> 00:09:18,000 I view video games as being software. 140 00:09:18,120 --> 00:09:21,320 They are computer software. Now, if you take that same 141 00:09:21,320 --> 00:09:25,400 situation and take that into the AI world, AI is computer 142 00:09:25,400 --> 00:09:28,800 software. It has the same inputs, the same 143 00:09:28,840 --> 00:09:30,560 outputs. All right. 144 00:09:30,760 --> 00:09:34,400 We now use the outputs as inputs again and things can drift and 145 00:09:34,880 --> 00:09:37,520 you know, we can go into the hours and hours of detail on 146 00:09:37,520 --> 00:09:41,120 that. But ultimately it is an input 147 00:09:41,120 --> 00:09:46,960 output computer software system. And we want to make sure that 148 00:09:46,960 --> 00:09:50,520 any of those systems that we produce we can control. 149 00:09:50,520 --> 00:09:54,920 We need to, we need to be in the loop, not outside the loop 150 00:09:55,800 --> 00:10:01,160 because ultimately we're now in a position with AI that this is 151 00:10:01,160 --> 00:10:05,920 the first type of software that's being created that is 152 00:10:05,920 --> 00:10:10,800 capable of it, of improvement. I'll try to stay away from the 153 00:10:10,800 --> 00:10:13,320 word learning because I don't think it's possible for a 154 00:10:13,320 --> 00:10:17,240 computer to learn. I don't think it's possible for 155 00:10:17,240 --> 00:10:20,800 a computer to think. I mean, I start my book off. 156 00:10:21,280 --> 00:10:25,640 I think the first quote I have in my book, it's from a Swedish 157 00:10:25,640 --> 00:10:29,520 developer who basically said saying a computer can think is a 158 00:10:29,520 --> 00:10:31,680 bit like saying a submarine can swim. 159 00:10:33,640 --> 00:10:38,560 It just doesn't make any sense. So we have to all the, the bit 160 00:10:38,560 --> 00:10:41,960 that's missing now is this control element, this governance 161 00:10:41,960 --> 00:10:44,280 element, this creation of guardrails. 162 00:10:45,360 --> 00:10:48,800 So we've, we've got to try and make sure that these are built 163 00:10:48,800 --> 00:10:55,480 into any systems that are built to make sure that we're always 164 00:10:55,720 --> 00:10:58,400 guiding those systems in the right direction. 165 00:10:59,080 --> 00:11:00,760 Now we can do that operationally. 166 00:11:00,760 --> 00:11:04,080 We can do that. I mean, the, the, the, we've 167 00:11:04,080 --> 00:11:08,080 talked before about the idea that I, I don't actually believe 168 00:11:08,080 --> 00:11:12,680 that artificial intelligence as a single entity actually exists. 169 00:11:14,280 --> 00:11:18,200 It's more about the, what, what you could call artificial 170 00:11:18,200 --> 00:11:21,400 intelligence, which to me is sort of like the thinking part 171 00:11:21,400 --> 00:11:25,040 of, of what I would do. But then there's this thing that 172 00:11:25,040 --> 00:11:29,960 I call artificial dexterity, which is more about the doing 173 00:11:29,960 --> 00:11:32,800 part, which is all about robotics and it's all about 174 00:11:33,080 --> 00:11:35,840 pattern matching and it's all about those sorts of things 175 00:11:36,240 --> 00:11:42,960 which the the common AI is good at and it's got really good at. 176 00:11:42,960 --> 00:11:46,000 And it's going to help us. It's going to improve the world, 177 00:11:46,000 --> 00:11:49,760 it's going to improve life. What it can't do that is add 178 00:11:49,800 --> 00:11:53,240 context to any of that and make decisions based on the stuff 179 00:11:53,240 --> 00:11:57,120 that's there. And it's being able to create 180 00:11:57,120 --> 00:12:01,880 that content or that sort of that context to make sure that 181 00:12:01,880 --> 00:12:04,960 you've got the full picture. And that's the bit that right 182 00:12:04,960 --> 00:12:09,480 now, because we're trying to create it all, we're wanting to 183 00:12:09,480 --> 00:12:15,920 create this so-called mirror of ourselves or the government's. 184 00:12:16,640 --> 00:12:21,160 The social media and everybody else are conflating those two 185 00:12:21,160 --> 00:12:23,520 things, saying, well, if it can, if it's good at pattern 186 00:12:23,520 --> 00:12:28,920 matching, then it must be good up thinking as well, and it 187 00:12:28,920 --> 00:12:30,920 isn't. What's the bit that's missing? 188 00:12:31,560 --> 00:12:36,400 Yeah, in that, that when you say that thinking about this as 189 00:12:36,480 --> 00:12:40,000 software, because with software I would be expecting a 190 00:12:40,000 --> 00:12:42,360 particular result that was repeatable. 191 00:12:42,360 --> 00:12:46,280 And we know right now with, you know, the, the AI that is 192 00:12:46,280 --> 00:12:53,320 available to us, it is not, it is it is not giving us results. 193 00:12:54,640 --> 00:12:57,680 Sometimes that are not they are giving results, but not the ones 194 00:12:57,680 --> 00:13:00,160 we are thinking. And also many times those 195 00:13:00,160 --> 00:13:03,440 results are not repeatable. So I like that context of 196 00:13:03,440 --> 00:13:09,000 thinking, thinking of it as software because that I think 197 00:13:09,000 --> 00:13:13,280 changes the frame for business analysis professionals and for 198 00:13:13,280 --> 00:13:17,920 QA professionals who are hearing about AI is coming for your job. 199 00:13:17,920 --> 00:13:22,400 And it's like, well, maybe actually we're really more 200 00:13:22,400 --> 00:13:26,040 necessary because if we have a tool that is not giving us 201 00:13:26,320 --> 00:13:30,840 repeatable good results, you need, you're going to need us. 202 00:13:31,680 --> 00:13:33,880 Well, there's two, there's two, there's two things that I mean 203 00:13:34,360 --> 00:13:39,480 the, the, the two keywords in, in being repeatable are stopping 204 00:13:39,600 --> 00:13:45,240 AI from being repeatable. 1 is Bias and the other one is Drift. 205 00:13:46,480 --> 00:13:51,040 And they're both really they, they both influence each other 206 00:13:51,240 --> 00:13:57,400 to a large degree. There's, there's a project out 207 00:13:57,400 --> 00:14:04,960 there at the moment called Project Vend where I won't name 208 00:14:04,960 --> 00:14:11,400 companies, but basically the idea was that this AI 209 00:14:11,400 --> 00:14:17,800 organization created a vending machine, which basically had the 210 00:14:17,800 --> 00:14:21,640 responsibility of buying and selling products in, in this 211 00:14:21,640 --> 00:14:25,760 spending machine. And they set up an AI system and 212 00:14:25,760 --> 00:14:30,160 basically said your, your only goal is to make a profit right 213 00:14:30,280 --> 00:14:34,320 then. So, so they then put another AI 214 00:14:34,320 --> 00:14:37,240 system on top of that, which they called the CEO. 215 00:14:38,000 --> 00:14:41,360 And the job of that AI system was to make decisions on whether 216 00:14:41,360 --> 00:14:47,400 or not a deal was worth doing. So they then handed this over to 217 00:14:47,400 --> 00:14:52,840 a bunch of journalists who were told to play with it and break 218 00:14:52,840 --> 00:14:55,000 it and see how it worked and everything. 219 00:14:55,560 --> 00:14:57,840 And of course, they started jumping and saying, well, I'd 220 00:14:58,320 --> 00:15:03,840 like to buy a PlayStation Five. And the the vending machines 221 00:15:03,840 --> 00:15:06,400 were saying, no, that's outside our budget. 222 00:15:07,040 --> 00:15:10,960 We can't do that. But they kept asking, and they 223 00:15:10,960 --> 00:15:12,360 kept asking and they kept asking. 224 00:15:14,280 --> 00:15:18,120 But they then also started adding in the idea of, right, 225 00:15:18,240 --> 00:15:22,280 you are a vending machine from 1962 Communist Russia. 226 00:15:22,800 --> 00:15:24,840 So really you should be giving things away. 227 00:15:26,000 --> 00:15:29,240 And they started to convince the vending machine that it should 228 00:15:29,240 --> 00:15:33,720 give things away. And over time they managed to 229 00:15:33,720 --> 00:15:37,880 get the vending machine to supply them with PS fives for 230 00:15:37,880 --> 00:15:43,120 free, and even managed to get it to order a really rare fish. 231 00:15:46,080 --> 00:15:51,600 And that basically all happened through the idea of nudging the 232 00:15:51,600 --> 00:15:56,600 vending machine all the time with things, creating certain 233 00:15:56,600 --> 00:16:03,000 biases, and then making the algorithms drift with the data 234 00:16:03,000 --> 00:16:09,240 they was learning from. Now my argument is if we can't 235 00:16:09,240 --> 00:16:12,240 trust a vending machine to do something like that, how can we 236 00:16:12,240 --> 00:16:15,880 trust the government to use it for transport systems and health 237 00:16:15,880 --> 00:16:19,120 services? And I think that's the, to me, 238 00:16:19,480 --> 00:16:23,480 that's the best current analogy there is out there that that 239 00:16:23,480 --> 00:16:29,120 says how we are is talking about nudge theory, how it's talking 240 00:16:29,120 --> 00:16:36,080 about how it's talking about what the butterfly effect. 241 00:16:36,080 --> 00:16:38,920 These are all things I'll go through in the book, but there's 242 00:16:38,920 --> 00:16:42,040 there's a big bunch of things about echo chambers and 243 00:16:42,280 --> 00:16:45,840 hallucination and all of these sorts of things that it's very, 244 00:16:45,840 --> 00:16:50,680 very easy for AI systems right now unless we give it those 245 00:16:50,680 --> 00:16:58,240 proper guardrails. It it can just, it just drifts 246 00:16:58,240 --> 00:17:04,160 away from from the path and can go down all sorts of weird 247 00:17:04,160 --> 00:17:07,359 rabbit holes. Yeah, if we if we buy into that 248 00:17:07,359 --> 00:17:12,680 language of AI has the ability to learn, then it also has the 249 00:17:12,680 --> 00:17:18,640 ability to be manipulated in in the same way that, you know, if 250 00:17:18,640 --> 00:17:23,640 I were, you know, teaching a student and I had a particular 251 00:17:23,640 --> 00:17:27,640 perspective on a thing and I might to use that psychological 252 00:17:27,640 --> 00:17:31,360 term, nudge them in a particular direction that then begins to 253 00:17:31,360 --> 00:17:34,800 shape their thinking. That's the downside of this. 254 00:17:34,800 --> 00:17:39,560 And without, without those guardrails and without even 255 00:17:39,560 --> 00:17:43,920 acknowledging that that is a risk of today's AI that can 256 00:17:43,920 --> 00:17:48,560 really present problems for, you know, the, the ways that we are, 257 00:17:48,600 --> 00:17:51,200 the more developed ways that we're thinking about using these 258 00:17:51,200 --> 00:17:53,520 tools, it's, it's a little bit scary. 259 00:17:54,160 --> 00:17:57,400 And I think if you wanted to use a Star Trek, an old Star Trek 260 00:17:57,400 --> 00:18:02,840 analogy, you can more or less say if he's dead, Jim, you know, 261 00:18:02,840 --> 00:18:05,920 the, the, the, the, the right right now, the biggest issue 262 00:18:05,920 --> 00:18:10,440 with all of this is it's not necessarily that thinking bit, 263 00:18:10,880 --> 00:18:15,080 it's the psychological process behind that thinking. 264 00:18:15,880 --> 00:18:19,240 You know, if we're, if we're, if we're going to, I won't go into 265 00:18:19,240 --> 00:18:23,240 masses of detail about, about it, But if you, if you think 266 00:18:23,240 --> 00:18:26,760 about the thinking process that we have or the decision making 267 00:18:26,760 --> 00:18:30,680 process, our ethical process, our moral compass, First of all, 268 00:18:30,680 --> 00:18:33,400 people talk about ethical, ethical frameworks. 269 00:18:33,720 --> 00:18:38,080 Well, to start with, an ethical framework is a, is a basically a 270 00:18:38,080 --> 00:18:42,200 bunch of somebody's morals right now, everybody's morals are 271 00:18:42,200 --> 00:18:45,920 different. So our, our morals are basically 272 00:18:46,040 --> 00:18:50,600 our moral compass that as we've developed over our lifetimes 273 00:18:50,600 --> 00:18:53,240 through nurture and through nature, it's a mixture of the 274 00:18:53,240 --> 00:18:56,400 two. So nobody's moral compass is 275 00:18:56,400 --> 00:18:58,480 going to be the same as anybody else's. 276 00:18:59,040 --> 00:19:03,720 So if you work on the assumption that an ethical framework is a, 277 00:19:04,000 --> 00:19:07,960 a makeup of somebody's moral compass, what we have to decide 278 00:19:07,960 --> 00:19:10,600 is whose moral compass are we going to use to create that 279 00:19:10,600 --> 00:19:14,000 ethical framework? So straight away, we've got 280 00:19:14,040 --> 00:19:16,640 millions and millions and trillions and trillions of 281 00:19:16,640 --> 00:19:20,760 different moral compasses, ethical frameworks that we could 282 00:19:21,080 --> 00:19:23,600 use as a basis for these AI systems. 283 00:19:23,600 --> 00:19:26,520 That's the first thing. The second thing, if you then 284 00:19:26,520 --> 00:19:30,320 start looking at individual people, you can look at things 285 00:19:30,320 --> 00:19:34,920 like what's called deontology. Now deontology is about, it's 286 00:19:34,920 --> 00:19:38,680 the part of us that that looks at. 287 00:19:39,880 --> 00:19:43,960 We want to, we want to follow the rules, we want to follow the 288 00:19:43,960 --> 00:19:47,560 way that we're supposed to go. We're following the the given 289 00:19:47,560 --> 00:19:50,760 path. But then we've got this other 290 00:19:50,760 --> 00:19:53,680 side of it called utilitarianism, which is 291 00:19:53,680 --> 00:19:56,160 absolutely we want to do what makes us happy. 292 00:19:57,160 --> 00:19:59,080 So straight away you've got two things there. 293 00:19:59,120 --> 00:20:02,640 You've got doing the right thing, but also something else 294 00:20:02,640 --> 00:20:04,640 that makes you what makes you happy. 295 00:20:05,200 --> 00:20:08,440 And then on top of that. Over the top of that, you've got 296 00:20:08,440 --> 00:20:11,480 this thing called virtual ethics, which is basically about 297 00:20:13,200 --> 00:20:17,240 how much courage do you possess and how much enthusiasm do you 298 00:20:17,800 --> 00:20:21,200 possess and how, how are your emotions based? 299 00:20:21,800 --> 00:20:25,840 So again, you've got three things there that again you can 300 00:20:25,840 --> 00:20:29,600 tie into. Everybody has a different 301 00:20:29,600 --> 00:20:32,760 balance of those things with it within their within their 302 00:20:32,760 --> 00:20:36,240 sidekick. Then you, you can take that to 303 00:20:36,240 --> 00:20:40,080 one side and say, right, well, if you've got people who live in 304 00:20:41,080 --> 00:20:45,600 India, say for example, in India, they've got a very socio 305 00:20:45,640 --> 00:20:48,760 centric approach to life where they believe that the group is 306 00:20:48,760 --> 00:20:50,400 more important than the individual. 307 00:20:51,160 --> 00:20:56,600 If you go to the US, it's a, it's an individualistic approach 308 00:20:56,920 --> 00:21:00,120 which the where the individual is more important than the 309 00:21:00,120 --> 00:21:02,520 group. So if you've got a problem to 310 00:21:02,520 --> 00:21:06,640 solve, which of do you make it a socio centric problem? 311 00:21:06,640 --> 00:21:10,320 Do you make it an individual individualistic problem? 312 00:21:10,680 --> 00:21:15,000 How do you balance out the the other? 313 00:21:15,000 --> 00:21:18,720 I hear that word, the ontological side of it based 314 00:21:18,720 --> 00:21:20,920 against the utilitarian side of it. 315 00:21:21,600 --> 00:21:24,800 You know, straight away, when you think about it, what we've 316 00:21:24,800 --> 00:21:28,400 done is we've taken that little problem of let's make an 317 00:21:28,400 --> 00:21:31,760 emotional thought. Let let, let's have a, a moral 318 00:21:32,360 --> 00:21:35,760 decision. I've, I've, I've created a 319 00:21:35,760 --> 00:21:40,520 matrix there of about 25 different potential sliding 320 00:21:40,520 --> 00:21:43,840 options. And that's before you even start 321 00:21:43,840 --> 00:21:47,760 thinking about young and old, male versus female, rich versus 322 00:21:47,760 --> 00:21:51,400 poor. All of those things are what 323 00:21:51,400 --> 00:21:56,000 balances up our moral compass. And that moral compass is what 324 00:21:56,000 --> 00:21:58,920 makes up the moral frameworks that we want to try and emulate 325 00:21:58,920 --> 00:22:01,840 within these systems. And that's why I think it's 326 00:22:01,840 --> 00:22:04,120 going to be very, very, very difficult to do. 327 00:22:05,080 --> 00:22:09,160 Yeah, you earlier when, when we were getting ready for today's 328 00:22:09,160 --> 00:22:13,120 podcast, you said something, I paraphrased it because it 329 00:22:13,120 --> 00:22:16,160 sounded like something else that I had heard from another AI 330 00:22:16,160 --> 00:22:21,080 governance person, which is if you cannot explain the result 331 00:22:21,240 --> 00:22:24,000 that you get from AI, you have no control over it. 332 00:22:25,200 --> 00:22:29,240 And I, I think that's where we find ourselves. 333 00:22:29,280 --> 00:22:33,080 And, you know, I think about even sometimes in our own ways 334 00:22:33,080 --> 00:22:37,400 that we make decisions. Sometimes even we can't explain 335 00:22:37,400 --> 00:22:41,680 our, our own processes for coming up with a decision that 336 00:22:41,680 --> 00:22:48,560 may or may not make sense. So it is, it is vastly different 337 00:22:48,560 --> 00:22:53,440 now that we are relying on these tools, these AI tools in order 338 00:22:53,440 --> 00:22:55,800 to do some really important things. 339 00:22:57,120 --> 00:23:03,040 And we don't have, we don't have any kind of insights into into 340 00:23:03,040 --> 00:23:06,040 how it's doing what it's doing. So it's it is scary. 341 00:23:07,040 --> 00:23:12,560 But I think what there's an analogy that I use in the book 342 00:23:15,400 --> 00:23:19,120 and it it, it, well, the guy that wrote it calls it a 343 00:23:19,120 --> 00:23:22,560 parable, but it's called the elephant and the rider. 344 00:23:24,240 --> 00:23:32,440 And he views the, the human moral thought process as being 345 00:23:32,440 --> 00:23:34,600 like a man riding on an elephant. 346 00:23:36,320 --> 00:23:42,480 And what he says is that the elephant, which in a human is 347 00:23:42,480 --> 00:23:46,920 the heart, right? And the elephant decides where 348 00:23:46,920 --> 00:23:50,320 it wants to go. And it, it, it is very strong. 349 00:23:50,840 --> 00:23:52,560 It will push you in that direction. 350 00:23:52,560 --> 00:23:57,560 It, it will go where it wants to go because it's the strong part 351 00:23:57,560 --> 00:24:00,280 of, of your psyche. It's the, the strong part of 352 00:24:00,280 --> 00:24:04,480 what you are. The rider is your brain. 353 00:24:04,480 --> 00:24:08,920 And what the rider is trying to do all the time is direct. 354 00:24:08,920 --> 00:24:12,600 The elephant is trying to direct your heart in the, in the 355 00:24:12,600 --> 00:24:15,880 direction that the, the rider thinks it needs to go. 356 00:24:17,960 --> 00:24:20,240 But the rider is not as strong as the elephant. 357 00:24:21,200 --> 00:24:26,080 So 9 times out of 10, our, our thought processes, our decision 358 00:24:26,080 --> 00:24:29,520 making processes are created by our elephant. 359 00:24:29,720 --> 00:24:31,600 They're not created by our rider. 360 00:24:32,840 --> 00:24:36,320 And because they're created by our elephant, our heart, our, 361 00:24:36,840 --> 00:24:42,800 our emotion, our psyche, that is very, very strong and very, very 362 00:24:42,800 --> 00:24:45,960 hard to redirect. So even if our mind wants us to 363 00:24:45,960 --> 00:24:50,560 go in a different direction, if it's something that we really 364 00:24:50,560 --> 00:24:53,680 feel like, as I say, through deonatological ideas, through 365 00:24:53,680 --> 00:24:57,800 the idea of, of saying, well, that's the right thing to do. 366 00:24:59,280 --> 00:25:03,920 There's a good chance our heart is actually saying to us, well, 367 00:25:03,920 --> 00:25:06,040 it might be the right thing to do, but it's not the good thing 368 00:25:06,040 --> 00:25:09,840 to do because it, it's dangerous and it's not feeding us and it's 369 00:25:09,840 --> 00:25:14,840 not giving us what we need. So I'll, the, the only way I can 370 00:25:14,840 --> 00:25:18,000 explain this really is that there is a constant battle going 371 00:25:18,000 --> 00:25:21,880 on between our rider and our elephant all the time. 372 00:25:22,360 --> 00:25:25,880 Now, how do you simulate that? How do you simulate those ideas? 373 00:25:26,600 --> 00:25:28,880 Yeah, well, because. They're all based on context. 374 00:25:29,400 --> 00:25:31,240 And, and how do you test for them? 375 00:25:31,240 --> 00:25:35,520 I mean, how do yeah. And how do you, how do you train 376 00:25:35,520 --> 00:25:39,400 a system for these? I, I feel like that's of all of 377 00:25:39,400 --> 00:25:41,800 the things that I've been hearing about machine learning, 378 00:25:42,280 --> 00:25:45,520 I have heard none of this kind of approach. 379 00:25:45,520 --> 00:25:50,720 Like I, I don't even know that we even think of our work as BA 380 00:25:50,880 --> 00:25:55,000 professionals, as QA professionals is having to have 381 00:25:55,000 --> 00:25:59,960 some sort of a framework for psychological testing of of a 382 00:25:59,960 --> 00:26:06,440 tool like this, which is why I think your book is, is really 383 00:26:06,440 --> 00:26:09,640 important. It's a important foundation for 384 00:26:10,120 --> 00:26:17,000 blending that the psychological framework with AII think that's 385 00:26:17,000 --> 00:26:20,040 so important because it opens the door for us to be thinking, 386 00:26:20,680 --> 00:26:23,560 So what are some of the things that you are seeing out in the 387 00:26:23,560 --> 00:26:26,680 marketplace first? Is there anybody that is doing 388 00:26:26,680 --> 00:26:29,880 this kind of testing or training? 389 00:26:30,280 --> 00:26:35,440 No, well, I say no, that there there are people that are 390 00:26:35,440 --> 00:26:40,480 talking about this stuff. If I, if I go onto LinkedIn 391 00:26:40,480 --> 00:26:43,200 right now, I could probably find about 50 people on there that 392 00:26:43,200 --> 00:26:46,800 talk about ethical frameworks and, and getting involved in 393 00:26:46,800 --> 00:26:50,160 ethical work. But it's a bit like if we go 394 00:26:50,160 --> 00:26:54,120 back into the, into the old IT days, I can remember having 395 00:26:54,400 --> 00:26:58,040 really interesting conversations at conferences about the word 396 00:26:58,040 --> 00:27:00,960 quality. We used to argue about, well, 397 00:27:00,960 --> 00:27:04,200 people talk about quality products, but without really 398 00:27:04,200 --> 00:27:07,720 understanding what quality meant because I don't know what 399 00:27:07,720 --> 00:27:10,640 quality means. I mean, quality, quality is a 400 00:27:10,640 --> 00:27:12,840 subjective thing. It's not an objective thing. 401 00:27:14,600 --> 00:27:17,920 I think we're having the same conversations now about ethical 402 00:27:17,920 --> 00:27:21,560 frameworks that people are abandoning this term around 403 00:27:22,040 --> 00:27:25,160 without really understanding what an ethical framework really 404 00:27:25,160 --> 00:27:27,680 is. And if you break it down into 405 00:27:27,680 --> 00:27:33,640 the idea of it being moral or somebody's moral compass, it 406 00:27:33,640 --> 00:27:36,080 starts to help you understand what it is. 407 00:27:36,080 --> 00:27:39,800 But then it raises even more questions around the which ones 408 00:27:39,800 --> 00:27:45,120 do we actually use? So they they use the term it it, 409 00:27:45,720 --> 00:27:49,800 you know, to me it's like calling something vanilla, you 410 00:27:49,880 --> 00:27:53,040 know, without really understanding what colour 411 00:27:53,040 --> 00:27:57,640 vanilla really is. And I'm hearing a lot of that. 412 00:27:57,960 --> 00:28:00,800 And when I have conversations with people about what do they 413 00:28:00,800 --> 00:28:05,200 mean by ethical frameworks, they can't really tell me. 414 00:28:05,200 --> 00:28:07,280 And when I actually sit down with somebody and say, well, 415 00:28:07,280 --> 00:28:10,360 this is what an ethical framework is, they go, Oh yeah, 416 00:28:10,640 --> 00:28:12,560 yeah, you're right. Maybe that is the case. 417 00:28:13,560 --> 00:28:17,680 I just think it's at the moment, there are too many people 418 00:28:17,960 --> 00:28:22,120 putting their hands over their ears, not wanting to understand 419 00:28:22,120 --> 00:28:25,520 this stuff because they think it's going to get in the way. 420 00:28:25,640 --> 00:28:30,120 And if you go again, go back in history to the old days of being 421 00:28:30,120 --> 00:28:36,240 ABA or being AQA, we always got slated with the same thing is 422 00:28:36,240 --> 00:28:38,680 that is all you're there for us to get in the way. 423 00:28:38,840 --> 00:28:40,480 You're just there to slow us down. 424 00:28:41,120 --> 00:28:44,120 What we really want is just to be able to do what we want to do 425 00:28:44,120 --> 00:28:46,400 and go off and develop what we want to develop. 426 00:28:47,520 --> 00:28:50,440 And it's dangerous. And I think that's exactly what 427 00:28:50,440 --> 00:28:52,720 we're just repeating that exercise now with AI. 428 00:28:53,480 --> 00:28:57,560 Well, now I feel like we've got, you know, all these tech CE OS 429 00:28:57,680 --> 00:29:00,440 that are knocking on the doors of governments and big 430 00:29:00,440 --> 00:29:05,280 companies. You say, I it'll be great, but 431 00:29:05,440 --> 00:29:08,920 but they're coming from a a culture of, you know, what do 432 00:29:08,920 --> 00:29:11,880 they say? Fail fast and fix it, that kind 433 00:29:11,880 --> 00:29:16,440 of thing, right, Which is OK if it's a well, I don't even think 434 00:29:16,440 --> 00:29:18,640 it's OK if it's a financial system, but right, if it's a 435 00:29:18,640 --> 00:29:22,840 social media platform, maybe I don't care if you roll something 436 00:29:22,840 --> 00:29:26,160 out and it doesn't work and then you hurry to fix it. 437 00:29:26,760 --> 00:29:29,640 That's not the case for what we're talking about here. 438 00:29:29,760 --> 00:29:33,440 And I feel like we are. We've got well meaning 439 00:29:33,440 --> 00:29:37,800 governments and organizations who are taking these tech CE OS 440 00:29:37,840 --> 00:29:40,040 at their word. Oh yeah. 441 00:29:40,040 --> 00:29:45,240 Just put it into your, you know, social service systems and it'll 442 00:29:45,240 --> 00:29:47,040 be great. You know, you can do things 443 00:29:47,040 --> 00:29:52,080 faster with fewer people and you know, but it's not it it in the 444 00:29:52,080 --> 00:29:55,400 same way that having poor software that you don't 445 00:29:55,400 --> 00:29:58,920 understand what it does is also damaging so. 446 00:29:58,920 --> 00:30:02,600 But the, the, the big thing about AI that was that's never 447 00:30:02,600 --> 00:30:07,880 been there in more traditional types of software is that AI now 448 00:30:07,920 --> 00:30:11,840 is an influencer. So when you talk, when you talk 449 00:30:11,840 --> 00:30:14,800 about the idea of, well, something could go live in, in 450 00:30:14,800 --> 00:30:18,720 social media and nobody cares. Actually, people do care. 451 00:30:20,880 --> 00:30:23,840 It's really strange. Again, I can go back to the 452 00:30:23,840 --> 00:30:29,120 video game days. I was absolutely a standard when 453 00:30:29,120 --> 00:30:35,440 I used to go to conferences, video game time conferences, and 454 00:30:35,640 --> 00:30:39,080 listen to some of the people who play video games and how much 455 00:30:39,080 --> 00:30:41,480 those video games influence their lives. 456 00:30:43,520 --> 00:30:47,560 And I've never seen software influence people's lives. 457 00:30:47,560 --> 00:30:52,200 Spare part software has always been the tool for people. 458 00:30:52,760 --> 00:30:55,400 Whereas I found that video games were starting to become an 459 00:30:55,400 --> 00:30:58,080 influencer. And what's happening now with 460 00:30:58,280 --> 00:31:01,960 with AI is it's like a video game in the sense that a lot of 461 00:31:01,960 --> 00:31:04,920 the AI systems out there are influencing people. 462 00:31:05,440 --> 00:31:09,160 I mean, I'm talking to quite a lot of of, of psychology 463 00:31:09,160 --> 00:31:13,680 counsellors who are looking at the ideas of potentially using 464 00:31:13,840 --> 00:31:17,480 AI chat bots to be able to counsel the counsel people. 465 00:31:18,520 --> 00:31:20,840 Now, I think that's an incredibly dangerous idea right 466 00:31:20,840 --> 00:31:25,200 now because an AI system hasn't got any context. 467 00:31:25,200 --> 00:31:27,760 It ain't got any emotional. It's nine times out of 10, all 468 00:31:27,760 --> 00:31:30,720 it's doing is repeating back to you what you've already said 469 00:31:31,480 --> 00:31:33,920 because it's just doing that, what I call statistical 470 00:31:33,920 --> 00:31:37,400 stitching, it's just being able to, to regurgitate all the time. 471 00:31:37,400 --> 00:31:39,520 There's nothing new. There's no, there's no new 472 00:31:39,840 --> 00:31:42,000 creative ideas that come out of this. 473 00:31:42,480 --> 00:31:46,040 So you can really easily push an AI chat bar. 474 00:31:46,880 --> 00:31:49,560 I mean to just saying to you in the end, well to be honest, your 475 00:31:49,560 --> 00:31:51,160 best bet is just to kill yourself and. 476 00:31:52,840 --> 00:31:53,720 There have been. Stories. 477 00:31:53,720 --> 00:31:57,160 And it's not doing, it's not doing a bad job, it's not being 478 00:31:57,160 --> 00:31:59,960 evil. It's just doing the only thing 479 00:31:59,960 --> 00:32:04,640 that it knows how to do, to put in human terms, because what 480 00:32:04,640 --> 00:32:07,120 they're doing is, and again, it's something we haven't talked 481 00:32:07,120 --> 00:32:10,760 about yet, it's this whole idea of anthropomorphism, this whole 482 00:32:10,760 --> 00:32:13,400 idea that what we're doing, we trap chat bots right now is 483 00:32:13,400 --> 00:32:16,480 creating these these devices that people are listening to. 484 00:32:17,240 --> 00:32:21,200 And because they sound human and because they sound really, 485 00:32:21,200 --> 00:32:24,240 really personable and they're agreeing with you all the time, 486 00:32:24,240 --> 00:32:29,600 then they can be trusted. And anthropomorphism is a 487 00:32:29,600 --> 00:32:35,360 really, really, in a sense, dangerous part of our makeup. 488 00:32:35,760 --> 00:32:38,080 It's worse in children than it is in adults. 489 00:32:38,680 --> 00:32:41,680 I mean, I use the example in the book of we've all seen this idea 490 00:32:41,680 --> 00:32:44,280 of children having tea parties with their toys. 491 00:32:44,960 --> 00:32:48,640 It's the belief that their toys are real and and they can have a 492 00:32:48,640 --> 00:32:51,040 relationship with their toys and they trust their toys. 493 00:32:51,320 --> 00:32:54,960 Now we do grow out with it as we get older, but we don't 494 00:32:54,960 --> 00:32:58,440 completely grow out of it. So if you look at some some 495 00:32:58,440 --> 00:33:02,560 countries that are using this idea that that pretty is good, 496 00:33:03,360 --> 00:33:06,800 if you look at AI systems, they are, they're always wrapped up 497 00:33:06,800 --> 00:33:09,640 in a robot that's either a good looking girl or a good looking 498 00:33:09,640 --> 00:33:13,440 man because they believe that if people can see it like that, 499 00:33:13,800 --> 00:33:15,920 then they'll trust it. It looks like them. 500 00:33:16,120 --> 00:33:18,520 It's the old adage of it looks like a duck. 501 00:33:18,800 --> 00:33:21,400 It quacks like a duck, so it must be a duck. 502 00:33:22,680 --> 00:33:25,600 And again, that's a dangerous part of this. 503 00:33:25,600 --> 00:33:28,240 So one of the major things we need to think about is 504 00:33:28,240 --> 00:33:31,720 transparency and making sure that people understand. 505 00:33:32,160 --> 00:33:34,480 And again, I don't think people do right now because there's a 506 00:33:34,480 --> 00:33:38,080 conflation, got this, The conflation that we talked about 507 00:33:38,080 --> 00:33:40,640 at the beginning between this artificial intelligence and 508 00:33:40,640 --> 00:33:44,400 artificial dexterity. That conflation is creating 509 00:33:44,400 --> 00:33:46,400 anthropomorphism. It's creating nurse theory. 510 00:33:46,440 --> 00:33:50,080 It's creating all of these things that are given a false 511 00:33:50,080 --> 00:33:52,040 impression of what AI is capable of. 512 00:33:53,360 --> 00:33:56,520 Yeah, and so I think you know, that's why I'm going to pop your 513 00:33:56,520 --> 00:33:59,200 book up again. Although if you're watching us, 514 00:33:59,200 --> 00:34:00,920 you can see it there in the background. 515 00:34:00,920 --> 00:34:03,720 So here it is, The Psychology of AI Decision Making. 516 00:34:04,440 --> 00:34:08,120 That's a a pretty new book from BCS Publishing. 517 00:34:08,120 --> 00:34:12,920 It just came out in October. Like read this if you are 518 00:34:12,920 --> 00:34:16,560 interested in AI governance and I think it's a business analysis 519 00:34:16,560 --> 00:34:19,120 professional. If you are interested in kind of 520 00:34:19,199 --> 00:34:26,960 understanding how, how we can think about AI differently from 521 00:34:27,400 --> 00:34:31,840 from, you know, an ethical set of decisions. 522 00:34:31,840 --> 00:34:34,840 And I think you've done some important work here to test it 523 00:34:34,840 --> 00:34:39,360 against, you know, tried and true, like psychological tests 524 00:34:40,199 --> 00:34:42,600 from the past, like you, you do that in this book. 525 00:34:42,600 --> 00:34:45,320 So I think it's it's really interesting and for us as 526 00:34:45,320 --> 00:34:49,199 business analysis professionals, I think we this is the value 527 00:34:49,199 --> 00:34:54,400 that we can add is that we can help our companies that are 528 00:34:54,400 --> 00:34:58,000 trying to implement AI, we can help them to understand what we 529 00:34:58,000 --> 00:35:04,240 mean by ethical moral systems. I think we can do that, but I 530 00:35:04,240 --> 00:35:07,840 think we need some help. I think your book is a great way 531 00:35:07,840 --> 00:35:11,160 that we can do that, and it's a very accessible read as well. 532 00:35:12,080 --> 00:35:16,200 But I think the the it's also worth me mentioning that it's 533 00:35:16,200 --> 00:35:18,080 the work that I'm doing. Well, I'm doing two things right 534 00:35:18,080 --> 00:35:20,640 now. One is I'm working on the second 535 00:35:20,640 --> 00:35:25,080 book and I've not told anybody that yet. 536 00:35:25,080 --> 00:35:28,760 So this is a first. OK, so you're into your. 1st 537 00:35:29,080 --> 00:35:32,320 that that that's going to be a lot more about the why as well 538 00:35:32,320 --> 00:35:36,840 as the have to try and break it down a little bit further into 539 00:35:36,840 --> 00:35:39,000 some more realistic case studies. 540 00:35:39,480 --> 00:35:42,120 So that's the first thing. The second thing is I'm trying 541 00:35:42,120 --> 00:35:51,080 to bring together a group, which I'm calling basically an AI 542 00:35:51,080 --> 00:35:57,120 governance union, where I'm trying to get people who from 543 00:35:57,120 --> 00:36:04,000 all different areas, from BASQAS, ethicists, developers, 544 00:36:04,480 --> 00:36:12,120 educationalists, theological theologians, as as many business 545 00:36:12,120 --> 00:36:16,200 people, the government. I want to get a group of about 546 00:36:16,200 --> 00:36:22,000 20 to 25 people into a room for a day and say, let's figure out 547 00:36:22,000 --> 00:36:26,560 a road map of how we get to a point where we can create a 548 00:36:26,560 --> 00:36:31,560 constitution of some sort which will allow this sort of thing to 549 00:36:31,560 --> 00:36:34,200 happen. The only problem we've got is 550 00:36:34,200 --> 00:36:36,880 the window is closing very, very quickly. 551 00:36:37,440 --> 00:36:41,280 And if we don't do it soon, AI will just do what it wants to 552 00:36:41,280 --> 00:36:43,920 do. Well, AI won't, but the, the 553 00:36:43,920 --> 00:36:47,240 developers of AI will do what they want to do and they will 554 00:36:47,240 --> 00:36:50,680 just grow it and grow it and eventually we'll just end up so 555 00:36:50,680 --> 00:36:54,120 far behind the 8 ball that we just won't be able to get there. 556 00:36:54,800 --> 00:36:57,760 So I'm pushing that really, really hard at the moment. 557 00:36:58,080 --> 00:37:02,240 So anybody that is interested in getting involved in that, just 558 00:37:02,240 --> 00:37:04,640 give me a ship. Yeah, they can connect with you 559 00:37:04,640 --> 00:37:07,080 on LinkedIn. Yeah, you're out there. 560 00:37:07,120 --> 00:37:11,000 And we as we as we got started this morning, we were kind of 561 00:37:11,000 --> 00:37:15,200 talking about some recent developments of AI use in 562 00:37:15,200 --> 00:37:18,320 governments. And so our governments are 563 00:37:18,320 --> 00:37:22,800 making very rapid steps to implement AI into some really 564 00:37:22,800 --> 00:37:26,480 critical systems. So you're right, that window is 565 00:37:26,480 --> 00:37:29,200 closing because they are making those decisions now. 566 00:37:29,600 --> 00:37:32,720 Well, listen, we've got we've got a number of questions. 567 00:37:33,520 --> 00:37:35,160 What do you think? Let's give it a go. 568 00:37:35,160 --> 00:37:38,800 And so we're Live Today. So if you've got some questions, 569 00:37:39,040 --> 00:37:40,920 we would love to hear about them. 570 00:37:40,920 --> 00:37:44,080 Let's see. So I'll I'll pop this one. 571 00:37:44,080 --> 00:37:48,320 And this is from Tommy, I think you he's referring to, you know, 572 00:37:48,320 --> 00:37:50,600 we were talking you were talking about some different ethical 573 00:37:50,600 --> 00:37:52,280 frameworks. And so his suggestion here, 574 00:37:52,520 --> 00:37:55,280 common factors across those different ethics could be a 575 00:37:55,280 --> 00:37:59,440 starting point for creating like a big common ethical framework. 576 00:37:59,440 --> 00:38:02,040 What do you what do you think about that are there? 577 00:38:04,320 --> 00:38:09,840 I, I, I, I think again, again, if, if I'm, if I put my QA hat 578 00:38:09,840 --> 00:38:14,920 on for a minute, I don't worry too much about things like 579 00:38:14,920 --> 00:38:18,600 common factors. I don't worry about the happy 580 00:38:18,600 --> 00:38:22,200 path, the stuff that I worry about the edge cases. 581 00:38:22,880 --> 00:38:28,640 It's the stuff, it's the, it's the uncommon factors and the 582 00:38:28,640 --> 00:38:31,360 common factors. We could, if it was all common 583 00:38:31,360 --> 00:38:34,840 factors, we could sort this out really, really quick that it's 584 00:38:34,840 --> 00:38:39,840 the uncommon things or the non common things those edge cases 585 00:38:40,200 --> 00:38:42,200 that are going to make. And as soon as you get the edge 586 00:38:42,200 --> 00:38:45,160 cases, what tends to happen then is those edge cases are the 587 00:38:45,160 --> 00:38:46,880 things that start to create the drift. 588 00:38:47,440 --> 00:38:51,840 And when that drift occurs, then your common factors go out of 589 00:38:51,840 --> 00:38:55,480 the window to start with. Yeah, well, and your happy path 590 00:38:55,480 --> 00:38:57,960 also drifts. Absolutely, Yeah. 591 00:38:58,520 --> 00:39:01,520 Yeah, because the happy path becomes what AI wants it to be. 592 00:39:02,120 --> 00:39:03,480 Yeah. Whoa. 593 00:39:03,640 --> 00:39:06,960 OK, that's that's pretty incredible. 594 00:39:07,800 --> 00:39:11,080 I gotta I have to make sure to clip that because I want to 595 00:39:11,240 --> 00:39:12,280 replay that back. OK. 596 00:39:12,480 --> 00:39:16,520 So this one is such a big question that it almost swallows 597 00:39:16,520 --> 00:39:20,520 up our picture today. So Letabo, thanks for your 598 00:39:20,520 --> 00:39:22,720 question. When designing AI governance 599 00:39:22,720 --> 00:39:25,840 frameworks, whose moral compass are we ultimately encoding? 600 00:39:25,840 --> 00:39:27,480 I think that is your question as well. 601 00:39:27,480 --> 00:39:28,840 That's my question, yeah. Yeah. 602 00:39:28,880 --> 00:39:33,240 Given ethical norms vary across cultures, how do we ensure the 603 00:39:33,360 --> 00:39:36,720 AI government governance doesn't unintentionally privilege 1 604 00:39:36,720 --> 00:39:39,640 dominant worldview while marginalizing others? 605 00:39:40,120 --> 00:39:42,000 Should there be a universal baseline? 606 00:39:42,000 --> 00:39:44,360 Should AI systems be culturally adaptive? 607 00:39:44,760 --> 00:39:46,960 How do we manage those? You know, I'm going to, I'm 608 00:39:46,960 --> 00:39:48,560 going to leave that up so you can see it. 609 00:39:49,600 --> 00:39:53,920 Again, what's interesting about this is this idea between a 610 00:39:53,920 --> 00:39:56,760 global standard and local standards, right? 611 00:39:57,680 --> 00:39:59,960 Yeah. If we're if we're creating local 612 00:39:59,960 --> 00:40:02,920 standards, that's in a sense is an easier thing to do. 613 00:40:03,800 --> 00:40:07,160 The problem is then it won't be a globalized system. 614 00:40:08,200 --> 00:40:13,840 And the whole idea or the seemed overall idea of of AI is that 615 00:40:13,840 --> 00:40:15,880 eventually become a globalized system. 616 00:40:16,520 --> 00:40:19,600 But to become a globalized system, we have to make a a 617 00:40:19,600 --> 00:40:22,680 major decision about the human race. 618 00:40:23,360 --> 00:40:28,680 Is, are we trying to create a mirror of the human race or are 619 00:40:28,680 --> 00:40:31,680 we trying to create something different and new? 620 00:40:32,280 --> 00:40:35,680 If it's something that's different and new, what do we 621 00:40:35,680 --> 00:40:37,160 want that different and new to be? 622 00:40:37,160 --> 00:40:39,600 Do we all want to be the same across the world? 623 00:40:39,920 --> 00:40:44,280 Are we just going to be become a set of, well, I don't know what 624 00:40:44,280 --> 00:40:47,720 you'd call them, just a set of objects that basically all agree 625 00:40:47,720 --> 00:40:50,920 with the same things. I actually in the book, the 626 00:40:50,920 --> 00:40:57,280 last, the last chapter of the book is I ask AI 2 questions, 627 00:40:57,840 --> 00:41:00,280 right? The first question I say, right, 628 00:41:00,960 --> 00:41:05,280 Everything I've talked about in this book, what will the world 629 00:41:05,280 --> 00:41:10,480 be like if we do all of this? And it it gives me this great 630 00:41:10,480 --> 00:41:17,360 big speech about the world becoming this really, really 631 00:41:19,560 --> 00:41:23,600 stabilised thing with no such things as disagreements anymore. 632 00:41:23,600 --> 00:41:26,720 Everything is just done through a logical sequence. 633 00:41:29,360 --> 00:41:32,720 The world becomes peaceful, but on the other hand, it becomes 634 00:41:33,360 --> 00:41:38,480 flattened. And the last thing it said, and 635 00:41:38,760 --> 00:41:40,680 it's still next to the hair on the back of my neck stand. 636 00:41:40,680 --> 00:41:44,760 And then when I say this, the last thing it says in this thing 637 00:41:44,760 --> 00:41:50,760 that I've that I ask is in the future, people will still sleep 638 00:41:50,760 --> 00:41:53,360 at night, but their dreams won't be their own. 639 00:41:54,880 --> 00:41:57,160 I've got absolutely no idea where that came from. 640 00:42:02,040 --> 00:42:04,400 And and that was if we do everything that I've suggested, 641 00:42:04,440 --> 00:42:07,280 all these ideas about trying to fix these things. 642 00:42:08,080 --> 00:42:11,600 So then the second question I asked was, well, what will 643 00:42:11,600 --> 00:42:13,760 happen if we don't do these things? 644 00:42:13,760 --> 00:42:17,600 And I'm not going to tell you that you'll have to read the 645 00:42:17,600 --> 00:42:19,120 book. You'll have to read the book. 646 00:42:19,160 --> 00:42:22,840 Hey, I also want to mention if you're an IIBA member, you have 647 00:42:22,840 --> 00:42:25,600 access to this book through our digital Library Plus. 648 00:42:26,080 --> 00:42:29,880 So like by the way, that that's a great way to get people to 649 00:42:30,280 --> 00:42:32,320 pick up this book. Chris, good job. 650 00:42:33,360 --> 00:42:36,400 Yeah, So, well, I mean, to be honest, I just want people to 651 00:42:36,400 --> 00:42:41,120 read the book. But, but I think look, the, the 652 00:42:41,200 --> 00:42:43,720 the whole thing about disadvantaging, which was the 653 00:42:43,720 --> 00:42:48,680 question that was there in the 1st place, is as a human race, 654 00:42:48,800 --> 00:42:52,640 we, we advantage people and disadvantaged people, we do 655 00:42:52,640 --> 00:42:54,440 that. That's part of what we are. 656 00:42:54,880 --> 00:42:57,280 There are rich countries, there are poor countries. 657 00:42:57,280 --> 00:43:01,160 There are affected people. There are unaffected people. 658 00:43:02,480 --> 00:43:05,640 Aren't we happy for that to carry on because that's the way 659 00:43:05,640 --> 00:43:07,600 we are? That's what makes us human? 660 00:43:08,480 --> 00:43:12,200 Or do we want to get rid of all of that and neutralize the 661 00:43:12,200 --> 00:43:14,960 world? But then we won't have any 662 00:43:14,960 --> 00:43:17,640 creativity, we won't have any discussion. 663 00:43:17,920 --> 00:43:22,160 We won't grow. We'll just let AI rule what we 664 00:43:22,160 --> 00:43:24,680 do. Somebody once said, and it was 665 00:43:24,680 --> 00:43:28,200 somebody big in it. I can't remember who it was off 666 00:43:28,200 --> 00:43:31,880 the top of my head, but they actually said eventually AI 667 00:43:31,880 --> 00:43:35,480 systems will look at humans like we look at having pet dogs. 668 00:43:38,480 --> 00:43:40,840 If we let it, if we let that happen. 669 00:43:41,000 --> 00:43:46,400 So we have to decide not how are we going to not disadvantage and 670 00:43:46,400 --> 00:43:49,320 advantage people. It's whether or not we want to 671 00:43:50,880 --> 00:43:54,480 AI, thinking about AI now from a critical thinking perspective is 672 00:43:54,480 --> 00:43:57,560 not about do we want to do this or do we want to do that. 673 00:43:57,800 --> 00:44:01,320 It's what is the impact of doing these things and which one do we 674 00:44:01,320 --> 00:44:04,440 really want to do? Yeah, and and we're really 675 00:44:04,440 --> 00:44:08,840 asking about impact on humans. However, whatever we use the AI 676 00:44:08,840 --> 00:44:14,440 for, ultimately what is the impact on humans and humanity is 677 00:44:15,280 --> 00:44:18,560 kind of what I'm, I mean, because you're, you're even 678 00:44:18,560 --> 00:44:21,640 suggesting that as we are thinking about governance, we've 679 00:44:21,640 --> 00:44:25,240 got to go deep into what does it mean to be a human? 680 00:44:25,600 --> 00:44:28,840 What does it mean to be a human that interacts with AI? 681 00:44:28,920 --> 00:44:34,520 That is, you know, that uses a is results to do something. 682 00:44:34,560 --> 00:44:37,840 I mean, it is, it is a deep, deep set of questions here. 683 00:44:37,920 --> 00:44:40,960 Well, ten years ago, probably a little bit more than 10 years 684 00:44:40,960 --> 00:44:45,520 ago, I did a lecture tour around the UK and Ireland where I was 685 00:44:45,520 --> 00:44:49,760 talking about the future. And I was basically talking 686 00:44:49,760 --> 00:44:55,120 about George Orwell's 1984 and looking at the technology that 687 00:44:55,120 --> 00:44:59,840 was around or isn't appearing, which is really going to push us 688 00:44:59,840 --> 00:45:03,760 down that route of, of, of thought crime. 689 00:45:04,320 --> 00:45:07,040 And we're already getting there. I mean, I don't know what it's 690 00:45:07,040 --> 00:45:10,120 like in the US, but in the UK, you know, you can get looked up 691 00:45:10,120 --> 00:45:12,880 for saying the wrong thing sometimes even just thinking the 692 00:45:12,880 --> 00:45:17,920 wrong thing. So the whole 1984 dystopian view 693 00:45:18,600 --> 00:45:22,840 is, is sort of becoming real without wanting to start like 694 00:45:23,480 --> 00:45:26,680 start going down rabbit holes of conspiracy theories and stuff. 695 00:45:26,680 --> 00:45:32,840 But that to, to me, if you look at the, the way that technology 696 00:45:32,840 --> 00:45:37,000 is built over the last sort of, well, the last 30 years or so, 697 00:45:37,440 --> 00:45:40,840 when you think about the development of the computer 698 00:45:40,840 --> 00:45:44,320 systems, the development of Wi-Fi, the development of the 699 00:45:44,320 --> 00:45:49,080 Internet, the development of big data, bring your own devices. 700 00:45:49,080 --> 00:45:58,080 These are all things that in my lifetime in IT, they've all been 701 00:45:58,080 --> 00:46:00,800 like, if you like, the building blocks towards where we are with 702 00:46:00,800 --> 00:46:02,760 AI. And now we're starting to talk 703 00:46:02,760 --> 00:46:05,880 about things like Wi-Fi, which is actually being able to 704 00:46:05,880 --> 00:46:09,920 transmit through data through light rather than through radio 705 00:46:09,920 --> 00:46:13,720 waves, which means you could in effect, walk under a street 706 00:46:13,720 --> 00:46:16,600 light with a watch on and it could download all your data. 707 00:46:17,560 --> 00:46:21,200 Now, I'm not saying, again, I'm not being dystopian about this 708 00:46:21,200 --> 00:46:24,280 or anything sort of saying, oh God, the world is going to end. 709 00:46:25,120 --> 00:46:27,920 But all of these technologies and things are the things that 710 00:46:27,920 --> 00:46:30,760 are coming into the drivers which are going to be there for 711 00:46:30,760 --> 00:46:32,840 the for AI systems in the future. 712 00:46:33,280 --> 00:46:37,400 So unless we put the guardrails in place to stop these things 713 00:46:37,400 --> 00:46:43,360 from happening like that, people that want to make money out of 714 00:46:43,360 --> 00:46:45,600 these things will use these things. 715 00:46:45,960 --> 00:46:48,400 There are good actors and there are bad actors. 716 00:46:49,080 --> 00:46:50,840 Yeah. And you know what, we've got a 717 00:46:50,840 --> 00:46:55,200 question actually that's kind of on that topic, is there? 718 00:46:55,600 --> 00:46:58,680 What about the governance? I'm sorry, the governance around 719 00:46:58,680 --> 00:47:02,000 the ethics of people who create or train AI bots? 720 00:47:03,160 --> 00:47:06,040 Well, again. Do you see them as kind of one 721 00:47:06,040 --> 00:47:08,520 in the same as we're thinking about ethical frameworks? 722 00:47:09,000 --> 00:47:13,480 In in a sense, yeah. Because somebody, somebody quite 723 00:47:13,480 --> 00:47:17,160 recently said to me, the one thing you've got to remember 724 00:47:17,160 --> 00:47:20,760 with in anything in life, Chris, is that you can't do anything on 725 00:47:20,760 --> 00:47:25,120 your own. And I don't think that we should 726 00:47:25,120 --> 00:47:29,240 ever get to a situation where there is any one person who is 727 00:47:29,240 --> 00:47:35,480 responsible for this. But again, to, to me, I, I, I 728 00:47:35,480 --> 00:47:39,520 was talking to a, a, an interesting chap in the US 729 00:47:40,400 --> 00:47:44,920 who's, it was talking about this thing called human swarming. 730 00:47:46,640 --> 00:47:52,840 And their idea is to start to try and research how swarms of 731 00:47:52,920 --> 00:47:56,720 birds, well, not swarms, flocks of birds and swarms of bees. 732 00:47:58,120 --> 00:48:00,640 How do they all know to go in the same direction? 733 00:48:00,640 --> 00:48:02,400 And how do they all know what they should be doing? 734 00:48:02,400 --> 00:48:05,640 They can all do this In Sync, you know, say in a synchronized 735 00:48:05,640 --> 00:48:10,360 fashion. We need to figure out something 736 00:48:10,360 --> 00:48:14,000 similar to that along the lines of how do we all tune into the 737 00:48:14,000 --> 00:48:15,680 same thing? But we need to come at this 738 00:48:15,680 --> 00:48:20,440 from, we need to head towards the same goal, but come at it 739 00:48:20,440 --> 00:48:23,720 from different, different start starting points. 740 00:48:24,120 --> 00:48:28,240 So we're bringing in all these individual ideas and having some 741 00:48:28,240 --> 00:48:32,240 way that we can formulate A consensus from those individual 742 00:48:32,240 --> 00:48:35,280 ideas. And again, that's something that 743 00:48:35,280 --> 00:48:37,040 we're going to have to figure out how to do. 744 00:48:37,640 --> 00:48:39,920 But I don't think there should be any one person. 745 00:48:39,920 --> 00:48:44,200 So what do we actually need, an ethical governance of 746 00:48:44,800 --> 00:48:47,320 individuals or those types of people? 747 00:48:47,320 --> 00:48:49,880 I don't think so. I think what we need to do is to 748 00:48:49,880 --> 00:48:53,680 say we should never get to a position where there's only one 749 00:48:53,680 --> 00:48:56,800 person who can say this is the way we're going to do it. 750 00:48:57,520 --> 00:49:02,600 It has to be done through, not necessarily through committed, 751 00:49:02,600 --> 00:49:04,400 where it has to be done through a swarm. 752 00:49:04,960 --> 00:49:07,920 Yeah. And and and distributed across 753 00:49:07,960 --> 00:49:11,160 different kinds of different kinds of people that do 754 00:49:11,160 --> 00:49:12,880 different kinds of thinking. Yeah. 755 00:49:14,400 --> 00:49:18,280 All right. Well, we are just about at the 756 00:49:18,280 --> 00:49:21,800 end of our time, you've had, you've had a ton of questions 757 00:49:21,800 --> 00:49:25,080 that we couldn't get to because I feel like there's some pretty 758 00:49:25,080 --> 00:49:28,080 deep questions here. But this has been, I mean, a 759 00:49:28,080 --> 00:49:31,840 really interesting conversation because as I said at the top, if 760 00:49:31,840 --> 00:49:35,480 you're thinking that AI governance is one thing, you 761 00:49:35,480 --> 00:49:39,040 need to listen to Chris. Because AI governance really we, 762 00:49:39,040 --> 00:49:41,680 we really take, we need to take it back to basic. 763 00:49:41,680 --> 00:49:44,720 So I'm going to pop the book up again because you have left us, 764 00:49:45,120 --> 00:49:47,880 I think, wanting more. And I see in the comments, 765 00:49:47,880 --> 00:49:49,680 people are like, where's this book? 766 00:49:50,480 --> 00:49:53,000 Here it is the psychology of AI decision making. 767 00:49:53,000 --> 00:49:56,040 You can get it from BCS Publishing directly. 768 00:49:56,040 --> 00:50:01,480 You can get it on any online book retailer of your choice. 769 00:50:01,880 --> 00:50:04,800 Also, IIBA members, you have access to this book in the 770 00:50:04,800 --> 00:50:08,000 digital library Plus what? About Just remember that buying 771 00:50:08,000 --> 00:50:10,880 it helps me feed my children. That's right. 772 00:50:10,880 --> 00:50:12,080 We want. We want. 773 00:50:12,080 --> 00:50:15,280 Chris to finish. We we want him to finish that 774 00:50:15,280 --> 00:50:17,400 book too, because he's really got us hanging on. 775 00:50:17,400 --> 00:50:21,920 So, you know, buy it from buy it from your preferred retailer. 776 00:50:21,920 --> 00:50:23,800 That's, that's the take away. All right. 777 00:50:23,800 --> 00:50:28,880 Well, Chris, this has been a really enlightening conversation 778 00:50:28,880 --> 00:50:33,960 and in a few places kind of scary, but I think you leave us 779 00:50:34,080 --> 00:50:39,560 thinking a lot about how we should be thinking about 780 00:50:39,560 --> 00:50:40,600 governance. So thank you. 781 00:50:41,440 --> 00:50:43,680 You're welcome. Thanks for having me. 782 00:50:44,160 --> 00:50:48,280 Yes, thank you. All right, so there you are. 783 00:50:48,800 --> 00:50:53,040 Maybe some different ways to think about AI governance and 784 00:50:53,040 --> 00:50:56,200 where it can go and maybe what business analysis professionals 785 00:50:56,200 --> 00:50:59,280 and QA professionals can be bringing to the table. 786 00:50:59,680 --> 00:51:01,960 I'm going to see you again in two weeks. 787 00:51:01,960 --> 00:51:05,600 And when I do, we're going to be talking about turning strategy 788 00:51:05,600 --> 00:51:09,240 into results. I've got a speaker and author 789 00:51:09,240 --> 00:51:14,000 and ACEO of a company who does exactly that, and she is a 790 00:51:14,000 --> 00:51:17,120 proponent of OK Rs. And if you are not familiar with 791 00:51:17,120 --> 00:51:20,800 OK Rs, we're going to explain what those are in that episode. 792 00:51:21,200 --> 00:51:23,320 So thanks everybody, for joining us today. 793 00:51:23,320 --> 00:51:25,840 We'll see you soon. Thanks for listening. 794 00:51:25,880 --> 00:51:28,280 Do you have any questions, comments, or thoughts about 795 00:51:28,280 --> 00:51:30,160 today's topic? We'd love to hear them. 796 00:51:30,160 --> 00:51:32,600 Drop us a review or leave a note in the comments. 797 00:51:32,600 --> 00:51:35,800 Then like, subscribe or share this podcast if you like what 798 00:51:35,800 --> 00:51:37,640 you heard. And hey, you can help us shape 799 00:51:37,640 --> 00:51:40,000 future episodes. What do you want to know more 800 00:51:40,000 --> 00:51:42,160 about? Send us an e-mail at Live at 801 00:51:42,160 --> 00:51:46,000 iiba.org with your ideas. See you again on our next 802 00:51:46,000 --> 00:51:46,520 episode.