1 00:00:14,800 --> 00:00:18,520 Speaker 1: Welcome to tech Stuff. This is the story. Each week 2 00:00:18,560 --> 00:00:21,599 Speaker 1: on Wednesdays, we bring you an in depth conversation with 3 00:00:21,720 --> 00:00:23,880 Speaker 1: someone who has a front row seat to the most 4 00:00:23,880 --> 00:00:31,720 Speaker 1: fascinating things happening in tech today. A conversation with David Spiegelholter, 5 00:00:32,360 --> 00:00:35,760 Speaker 1: a professor of statistics at Cambridge an author of the 6 00:00:35,880 --> 00:00:40,560 Speaker 1: Art of Uncertainty, How to Navigate Chance, Ignorance, Risk and Luck. 7 00:00:41,400 --> 00:00:46,600 Speaker 1: Spiegelhlter has devoted his life to understanding uncertainty. After all, 8 00:00:46,720 --> 00:00:49,320 Speaker 1: it's one of the most uncomfortable aspects of being human, 9 00:00:49,960 --> 00:00:54,120 Speaker 1: particularly when it comes to our health. Since the nineteen seventies, 10 00:00:54,200 --> 00:00:57,760 Speaker 1: Spiegelhulter has worked on algorithms to assist clinicians and patients 11 00:00:58,120 --> 00:01:01,760 Speaker 1: make better decisions about what treatment options to take for cancer, 12 00:01:02,640 --> 00:01:05,840 Speaker 1: and he has a deeply personal understanding of the topic. 13 00:01:06,640 --> 00:01:09,240 Speaker 1: In nineteen ninety seven, he lost his son Danny to 14 00:01:09,280 --> 00:01:12,680 Speaker 1: cancer at the age of just five, and the epigraph 15 00:01:12,800 --> 00:01:15,959 Speaker 1: to David's book is a quote from the Bible. The 16 00:01:16,080 --> 00:01:18,720 Speaker 1: race is not to the swift, nor the battle to 17 00:01:18,760 --> 00:01:22,120 Speaker 1: the strong, nor bread to the wise, nor riches to 18 00:01:22,200 --> 00:01:25,960 Speaker 1: men of understanding, nor favor to men of skill. But 19 00:01:26,160 --> 00:01:30,360 Speaker 1: time and chance happened to them all of course, with 20 00:01:30,440 --> 00:01:33,280 Speaker 1: the explosion of AI, we now have better and better 21 00:01:33,360 --> 00:01:36,840 Speaker 1: tools to help us understand the world and make informed decisions. 22 00:01:37,360 --> 00:01:41,200 Speaker 1: And in fact, Spiegelhalter was an early pioneer of the technology. 23 00:01:41,840 --> 00:01:45,679 Speaker 1: So that's why I decided to start our conversation. We 24 00:01:45,800 --> 00:01:50,120 Speaker 1: live in this extraordinary moment where technology seems to be 25 00:01:50,160 --> 00:01:53,120 Speaker 1: giving us more of a view around the corner into 26 00:01:53,160 --> 00:01:56,960 Speaker 1: the future than it ever has. How has the march 27 00:01:56,960 --> 00:02:00,640 Speaker 1: of technology impacted your work and your understanding these topics 28 00:02:00,720 --> 00:02:01,360 Speaker 1: over your career? 29 00:02:02,080 --> 00:02:04,080 Speaker 2: Oh a huge amount. I mean, we needn't get into 30 00:02:04,080 --> 00:02:07,400 Speaker 2: the whole Asian methodology, but that's what I was interested in, 31 00:02:07,440 --> 00:02:09,040 Speaker 2: and we couldn't do it because you just couldn't do 32 00:02:09,080 --> 00:02:12,600 Speaker 2: the calculations. You couldn't do the maths. But instead of 33 00:02:12,639 --> 00:02:15,960 Speaker 2: trying to do the maths, you just used brute computing 34 00:02:16,080 --> 00:02:20,160 Speaker 2: force to simulate millions of different possibilities and look at 35 00:02:20,200 --> 00:02:24,600 Speaker 2: their distribution and the algorithms we knew would converge to 36 00:02:24,680 --> 00:02:27,440 Speaker 2: the correct answer if they ran long enough. You had 37 00:02:27,440 --> 00:02:30,639 Speaker 2: to wait till nineteen ninety or so. Just before that 38 00:02:30,800 --> 00:02:35,440 Speaker 2: technology that ability to do such huge simulation exercises was 39 00:02:35,480 --> 00:02:38,920 Speaker 2: on everyone's desktop, and then there was an explosion and 40 00:02:38,960 --> 00:02:41,360 Speaker 2: a complete change in the way statistics was done. Up 41 00:02:41,360 --> 00:02:44,320 Speaker 2: to then people did clever maths and then programmed that 42 00:02:44,480 --> 00:02:47,200 Speaker 2: in and it changed into no, we don't have to 43 00:02:47,200 --> 00:02:50,079 Speaker 2: do any maths. We just have to program in the problem, 44 00:02:50,160 --> 00:02:53,160 Speaker 2: the model that we're trying to solve, present the data 45 00:02:53,240 --> 00:02:56,960 Speaker 2: to it, and send it off. And wait. But what 46 00:02:57,080 --> 00:02:59,560 Speaker 2: I think you might be starting to allude to, which 47 00:02:59,560 --> 00:03:02,680 Speaker 2: I'm sure get onto, is the role of AI. Our 48 00:03:02,760 --> 00:03:05,200 Speaker 2: AI is already changing my I used a lot in 49 00:03:05,240 --> 00:03:10,040 Speaker 2: writing the book, both in the researching and summarizing of 50 00:03:10,120 --> 00:03:12,760 Speaker 2: literature and of course in the coding all the time. 51 00:03:13,160 --> 00:03:15,640 Speaker 2: You know, I always rewrote everything, but I used it 52 00:03:15,680 --> 00:03:17,640 Speaker 2: a lot. I use it all the time in my 53 00:03:17,800 --> 00:03:21,040 Speaker 2: daily word, daily life. But actually, will it be able 54 00:03:21,080 --> 00:03:25,240 Speaker 2: to make predictions? And I am rather skeptical about claims 55 00:03:25,720 --> 00:03:27,560 Speaker 2: both that are sort of you know what, you might 56 00:03:27,560 --> 00:03:30,160 Speaker 2: call it a global level, or a social level, or 57 00:03:30,240 --> 00:03:33,520 Speaker 2: even at a personal level, about our health, about the 58 00:03:33,560 --> 00:03:35,559 Speaker 2: ability of AI to make predictions. 59 00:03:36,200 --> 00:03:39,080 Speaker 1: Your book has the epigraph from the Bible. How did 60 00:03:39,120 --> 00:03:40,200 Speaker 1: you come up with that? Oh? 61 00:03:40,280 --> 00:03:44,040 Speaker 2: By using AI to ask for quotes that use chants 62 00:03:44,080 --> 00:03:44,640 Speaker 2: and things like that. 63 00:03:44,760 --> 00:03:45,320 Speaker 1: Is that true? 64 00:03:45,440 --> 00:03:47,480 Speaker 2: Yeah, I'd actually for that one, I knew that one, 65 00:03:47,520 --> 00:03:49,240 Speaker 2: but otherwise so yeah, I use AI. 66 00:03:49,320 --> 00:03:51,200 Speaker 1: Well why that one? Why was that the first one 67 00:03:51,240 --> 00:03:52,040 Speaker 1: that you used? Oh? 68 00:03:52,120 --> 00:03:55,600 Speaker 2: I think because the whole book, and especially if we're 69 00:03:55,640 --> 00:03:58,480 Speaker 2: the first chapter, which is about my grandfather, was intended 70 00:03:58,520 --> 00:04:01,040 Speaker 2: to give the idea of the utter lack of control 71 00:04:01,400 --> 00:04:04,440 Speaker 2: we have in our lives, and we have an illusion 72 00:04:04,520 --> 00:04:08,040 Speaker 2: of control, which I think actually is not helpful. I 73 00:04:08,080 --> 00:04:10,880 Speaker 2: think to realize that how little control we do have 74 00:04:10,960 --> 00:04:12,960 Speaker 2: in our lives, how much of what happens to us 75 00:04:13,400 --> 00:04:16,080 Speaker 2: is what, for want of a better word, we might 76 00:04:16,120 --> 00:04:18,720 Speaker 2: call chance. In other words, events that will happen to 77 00:04:18,800 --> 00:04:21,440 Speaker 2: us that were unpredictable and that you know, and that 78 00:04:21,520 --> 00:04:24,479 Speaker 2: we had no control over. I think that's rather important 79 00:04:24,480 --> 00:04:27,760 Speaker 2: to realize that. And because the word that appears in 80 00:04:27,800 --> 00:04:30,520 Speaker 2: the book more than almost any other is humility. There's 81 00:04:30,520 --> 00:04:33,080 Speaker 2: almost no mention of the word rationality in the book. 82 00:04:33,120 --> 00:04:35,960 Speaker 2: This is not a book about being rational, It's a 83 00:04:36,000 --> 00:04:37,080 Speaker 2: book about being humble. 84 00:04:37,480 --> 00:04:39,400 Speaker 1: He mentioned your grandfather, and in the book you talk 85 00:04:39,440 --> 00:04:42,560 Speaker 1: about how he survived various battles in World War One. 86 00:04:43,160 --> 00:04:45,520 Speaker 1: You'll talk about your mother being captured by parents in 87 00:04:45,520 --> 00:04:49,800 Speaker 1: the South China. See exactly and they're making it to 88 00:04:49,839 --> 00:04:52,000 Speaker 1: the UK, where you should meet your father. 89 00:04:52,080 --> 00:04:54,080 Speaker 2: Yeah, who then nearly died in the Second World War 90 00:04:54,080 --> 00:04:54,440 Speaker 2: as well. 91 00:04:54,600 --> 00:04:54,920 Speaker 1: He did. 92 00:04:55,040 --> 00:04:57,120 Speaker 2: Yeah, he got TB. I mean it was an illness 93 00:04:57,160 --> 00:04:59,480 Speaker 2: and he was in the hospital for weeks, and he 94 00:04:59,600 --> 00:05:02,360 Speaker 2: was a vat cuated from Jerusalem as he was there 95 00:05:02,360 --> 00:05:04,520 Speaker 2: when he heard the Saint David's Hotel being blown up. 96 00:05:04,920 --> 00:05:06,920 Speaker 2: So then you know, there so much could have happened 97 00:05:06,920 --> 00:05:09,279 Speaker 2: to both of them. Then, as I mentioned in my book, 98 00:05:09,320 --> 00:05:13,320 Speaker 2: the biggest chance event of all is my conception. It's 99 00:05:13,360 --> 00:05:17,600 Speaker 2: not just me, everyone's conceptions. An extraordinarily unlikely of it 100 00:05:17,720 --> 00:05:21,039 Speaker 2: so easily could not have happened. So me realizing that 101 00:05:21,120 --> 00:05:24,640 Speaker 2: and researching the situation of my conception, it really made me. 102 00:05:24,960 --> 00:05:26,960 Speaker 2: You know, it changed my whole attitude to life. In fact, 103 00:05:27,160 --> 00:05:29,560 Speaker 2: it really did. It made me think, God, I'm here 104 00:05:29,680 --> 00:05:33,599 Speaker 2: just by total chants in what I call these micro 105 00:05:33,760 --> 00:05:38,279 Speaker 2: contingencies that just accumulated, and here I am. And so 106 00:05:38,400 --> 00:05:41,800 Speaker 2: the idea somehow that I'm you know, on the earth 107 00:05:41,800 --> 00:05:44,279 Speaker 2: for a purpose, or I'm you know, in any way 108 00:05:44,360 --> 00:05:48,280 Speaker 2: special I find is now for me a complete illusion. 109 00:05:48,760 --> 00:05:53,360 Speaker 1: When did you start getting interested in this relationship between poverty, statistics, 110 00:05:53,400 --> 00:05:54,160 Speaker 1: and medicine. 111 00:05:54,279 --> 00:05:58,040 Speaker 2: Yeah, I was interested in the mathematical aspects of statistics particularly, 112 00:05:58,040 --> 00:05:59,960 Speaker 2: But then there was a job going and the funny 113 00:06:00,080 --> 00:06:03,280 Speaker 2: the job going was in nineteen seventy eight, and it 114 00:06:03,320 --> 00:06:07,479 Speaker 2: was to work on what was then called computer aided diagnosis. 115 00:06:07,839 --> 00:06:11,200 Speaker 2: Well now we've called it AI. So nineteen seventy eight 116 00:06:12,000 --> 00:06:15,320 Speaker 2: was using some basic statistical algorithms what is now called 117 00:06:15,400 --> 00:06:19,280 Speaker 2: naive base simple algorithms. It's still around and used as 118 00:06:19,279 --> 00:06:23,039 Speaker 2: a very basic machine learning algorithm, for example in a 119 00:06:23,080 --> 00:06:25,440 Speaker 2: spam filtering or whatever. And we were doing that in 120 00:06:25,480 --> 00:06:28,239 Speaker 2: the late seventies. So it's one of the first jobs 121 00:06:28,279 --> 00:06:32,880 Speaker 2: to work on algorithms in medicine for both diagnosis and prognosis. 122 00:06:32,920 --> 00:06:35,400 Speaker 2: We were working on the likelihood of someone with head 123 00:06:35,440 --> 00:06:38,880 Speaker 2: injury surviving and so on. And because the computers you 124 00:06:38,880 --> 00:06:40,960 Speaker 2: could even carry them around a thing. They were sat 125 00:06:41,000 --> 00:06:43,920 Speaker 2: in the corner with a huge, great machine with a keyboards. 126 00:06:44,000 --> 00:06:47,600 Speaker 2: It's incredibly clumsy to use, but we were doing that. 127 00:06:47,640 --> 00:06:49,520 Speaker 2: And then in the early eighties I was working on 128 00:06:49,560 --> 00:06:53,760 Speaker 2: more algorithms. Then we got into developments in AI and 129 00:06:53,800 --> 00:06:56,080 Speaker 2: so on. So you know, this stuff is not new. 130 00:06:56,279 --> 00:07:00,480 Speaker 2: It's been around for ages, what was predict addicted. It's 131 00:07:00,480 --> 00:07:04,560 Speaker 2: still going because that's an algorithm for predicting the survival 132 00:07:04,680 --> 00:07:07,600 Speaker 2: of women with breast cancer and men with prostate cancer, 133 00:07:07,640 --> 00:07:13,080 Speaker 2: still available, hugely, widely used. It's a very good statistical algum, 134 00:07:13,240 --> 00:07:18,520 Speaker 2: a regression algorithm. Of course, in practice, any actual clinician 135 00:07:18,640 --> 00:07:22,280 Speaker 2: making a decision with the patient would use much more 136 00:07:22,280 --> 00:07:25,200 Speaker 2: personal information that they might have the patient, because you know, 137 00:07:25,480 --> 00:07:28,480 Speaker 2: for example, physical status doesn't go into the algorithm, and 138 00:07:28,560 --> 00:07:31,640 Speaker 2: yet that might be you know, someone's basic underlying health 139 00:07:31,760 --> 00:07:34,640 Speaker 2: might be incredibly important. So that's when we wrote the 140 00:07:34,640 --> 00:07:37,840 Speaker 2: interface for a predict we try to emphasize not say 141 00:07:37,880 --> 00:07:41,000 Speaker 2: this is the risk of this patient. It's just what 142 00:07:41,040 --> 00:07:43,160 Speaker 2: would expect to one hundred happen to one hundred people 143 00:07:43,400 --> 00:07:46,600 Speaker 2: who ticked the boxes that she did or he did. 144 00:07:46,960 --> 00:07:51,440 Speaker 2: But actually people have tried different, more sophisticated machine learning 145 00:07:51,800 --> 00:07:54,320 Speaker 2: methods and they don't make much difference. And so it's 146 00:07:54,360 --> 00:07:55,880 Speaker 2: about as good as you can do. I think with 147 00:07:56,000 --> 00:07:58,920 Speaker 2: the data that is available, you could always do better 148 00:07:58,960 --> 00:08:02,600 Speaker 2: by collecting more data to and having a bigger, better database. 149 00:08:02,960 --> 00:08:05,960 Speaker 2: It's going to be marginal marginal benefits just from using 150 00:08:06,000 --> 00:08:08,880 Speaker 2: AI with the same data so the real you know, 151 00:08:08,960 --> 00:08:11,400 Speaker 2: benefit in the future, of course, is just by having 152 00:08:11,480 --> 00:08:12,200 Speaker 2: better data. 153 00:08:12,640 --> 00:08:14,400 Speaker 1: So where did you say a few moments ago that 154 00:08:14,440 --> 00:08:16,920 Speaker 1: you doubted that AI would be a useful. 155 00:08:16,720 --> 00:08:18,720 Speaker 2: Oh well, I mean it's going to be marginal, marginal 156 00:08:18,760 --> 00:08:21,640 Speaker 2: benefits just from using AI with the same data. 157 00:08:22,080 --> 00:08:24,520 Speaker 1: What about other things like drug discovery. 158 00:08:24,080 --> 00:08:27,000 Speaker 2: Or oh yeah all that, Yeah that's very important. No, no, 159 00:08:27,040 --> 00:08:29,960 Speaker 2: then it's going to be great, huge, huge benefits there. 160 00:08:30,120 --> 00:08:32,120 Speaker 2: Now I'm talking about predicting what's going to happen to 161 00:08:32,200 --> 00:08:34,600 Speaker 2: me in the future, because I've got prospect cancer so 162 00:08:34,800 --> 00:08:37,719 Speaker 2: I am quite interested in this. And of course when 163 00:08:37,720 --> 00:08:39,920 Speaker 2: I got it, I looked at all the algorithms and 164 00:08:39,960 --> 00:08:42,840 Speaker 2: which were sort of helpful, but they're very broad. All 165 00:08:42,880 --> 00:08:45,800 Speaker 2: they do really is tell you what we would expect 166 00:08:45,840 --> 00:08:47,880 Speaker 2: to happen to It is one hundred people who ticked 167 00:08:47,920 --> 00:08:52,240 Speaker 2: the boxes you've ticked, and of course everyone is so different, 168 00:08:52,480 --> 00:08:55,920 Speaker 2: Everyone's so different, so I know that that will give 169 00:08:55,960 --> 00:08:58,400 Speaker 2: me a broad figure, but it's only a ballpark figure. 170 00:08:58,400 --> 00:09:01,640 Speaker 2: It's still very useful until of my tenure survival, but 171 00:09:01,920 --> 00:09:06,160 Speaker 2: one that I know could be changed by just having 172 00:09:06,160 --> 00:09:06,840 Speaker 2: more information. 173 00:09:07,600 --> 00:09:10,440 Speaker 1: So you spent your career trying to help doctors and 174 00:09:10,520 --> 00:09:14,000 Speaker 1: patients make better decisions about what to do when a 175 00:09:14,040 --> 00:09:17,520 Speaker 1: patient gets sick. But you've also lived this with your 176 00:09:17,520 --> 00:09:21,760 Speaker 1: son Danny, this experience of how to make medical decisions 177 00:09:22,240 --> 00:09:24,560 Speaker 1: in the face of horribly serious illness. 178 00:09:25,280 --> 00:09:27,599 Speaker 2: Oh, that's interesting because I do think about that. We 179 00:09:28,120 --> 00:09:30,920 Speaker 2: did some decisions and I'm not sure, you know, maybe 180 00:09:31,200 --> 00:09:32,839 Speaker 2: we always say, well, maybe if we had taken him 181 00:09:32,840 --> 00:09:35,320 Speaker 2: to these, to Canada or something, and that we could 182 00:09:35,320 --> 00:09:38,760 Speaker 2: have got a different therapy. In a way, that's one 183 00:09:38,760 --> 00:09:42,200 Speaker 2: thing I prefer not to dwell on too much, because 184 00:09:43,679 --> 00:09:46,640 Speaker 2: you know, you don't know. But it has made me 185 00:09:46,800 --> 00:09:50,960 Speaker 2: very aware of the importance of making informed medical decisions, 186 00:09:51,360 --> 00:09:53,440 Speaker 2: and that's what I with my team we worked on 187 00:09:53,480 --> 00:09:56,720 Speaker 2: those on providing decision aids, not in any way to 188 00:09:56,840 --> 00:10:00,200 Speaker 2: encourage people to make any particular decision, but just as 189 00:10:00,240 --> 00:10:02,840 Speaker 2: I mentioned in the book, I don't believe that decision 190 00:10:02,920 --> 00:10:05,360 Speaker 2: theory and all the advances that have been done in 191 00:10:05,400 --> 00:10:07,440 Speaker 2: that can never tell you what to do, because it 192 00:10:07,520 --> 00:10:10,640 Speaker 2: assumes that you know all the possible outcomes, you know 193 00:10:10,720 --> 00:10:14,000 Speaker 2: all the possible options, you know all the probabilities and 194 00:10:14,080 --> 00:10:18,720 Speaker 2: the values, and of course this is totally infeasible apart 195 00:10:18,720 --> 00:10:21,600 Speaker 2: from the simplest sort of gambling type examples. You never 196 00:10:21,679 --> 00:10:23,520 Speaker 2: know any of these things. You never know how you're 197 00:10:23,520 --> 00:10:25,680 Speaker 2: going to feel in the future if something happens, and 198 00:10:25,720 --> 00:10:28,960 Speaker 2: so on. So it's impossible to be rational in those 199 00:10:29,000 --> 00:10:32,760 Speaker 2: situations and use decision theory. But I think it's really 200 00:10:32,800 --> 00:10:36,600 Speaker 2: helpful to try to examine the problem to face up 201 00:10:36,640 --> 00:10:38,840 Speaker 2: to a decision has to be made. One of the 202 00:10:38,840 --> 00:10:42,640 Speaker 2: biggest problems about decisions is that people don't actually make decisions. 203 00:10:43,120 --> 00:10:45,720 Speaker 2: I don't you just go. You just find yourself going 204 00:10:45,760 --> 00:10:48,640 Speaker 2: down a path, and you never just stop and say, 205 00:10:48,640 --> 00:10:51,440 Speaker 2: this is a decision point. There is a branch in 206 00:10:51,480 --> 00:10:53,360 Speaker 2: the road. We have to choose which way to go, 207 00:10:53,640 --> 00:10:57,080 Speaker 2: and sometimes you can recognize those points, but they're few 208 00:10:57,080 --> 00:11:00,600 Speaker 2: and far between, and so would I love to encourage 209 00:11:00,600 --> 00:11:04,280 Speaker 2: people to actually have much more of those decision points. 210 00:11:04,520 --> 00:11:05,920 Speaker 2: This is the time we have to make a decision. 211 00:11:05,920 --> 00:11:09,200 Speaker 2: These are the possibilities, the benefits and harms of the 212 00:11:09,240 --> 00:11:11,520 Speaker 2: options that face you. We're not going to tell you 213 00:11:11,559 --> 00:11:13,600 Speaker 2: what to do. We might be able to put some 214 00:11:13,720 --> 00:11:17,480 Speaker 2: rough probabilities. For example, for women in breast cancer, we've 215 00:11:17,520 --> 00:11:20,320 Speaker 2: got such a lot of data we can actually produce 216 00:11:20,400 --> 00:11:23,000 Speaker 2: reasonably good ten year survival rates and what the benefit 217 00:11:23,040 --> 00:11:26,559 Speaker 2: would be if you had chemotherapy. So, for example, in Cambridge, 218 00:11:27,080 --> 00:11:31,040 Speaker 2: unless you your absolute survival benefit is going to go 219 00:11:31,160 --> 00:11:34,480 Speaker 2: up by three percent with chemotherapy, they don't recommend chema 220 00:11:34,559 --> 00:11:38,440 Speaker 2: therapy because that means essentially that of all the people 221 00:11:38,480 --> 00:11:42,000 Speaker 2: that give chemotherapy, out of thirty people, only one will 222 00:11:42,040 --> 00:11:45,840 Speaker 2: benefit after ten years. Only you get one extra ten 223 00:11:45,920 --> 00:11:49,240 Speaker 2: year survivor for thirty people being given to chemotherapy, which 224 00:11:49,280 --> 00:11:53,120 Speaker 2: can have a really awful effect on people. And by 225 00:11:53,160 --> 00:11:55,360 Speaker 2: producing those numbers you can get a feeling that well, 226 00:11:55,480 --> 00:11:58,040 Speaker 2: you've got to have a reasonable benefit in order to 227 00:11:58,440 --> 00:12:01,800 Speaker 2: take the hit of the tree. So in those situations, 228 00:12:01,840 --> 00:12:04,440 Speaker 2: I think it's really good. You can't do it this 229 00:12:04,600 --> 00:12:08,120 Speaker 2: exactly to some idea of what the benefits might be. 230 00:12:08,559 --> 00:12:11,200 Speaker 2: But on the whole, you know, it's difficult to do 231 00:12:11,280 --> 00:12:15,240 Speaker 2: that in situations where you havn't got all that data 232 00:12:15,280 --> 00:12:18,240 Speaker 2: and all that analysis, all that tech behind. 233 00:12:17,880 --> 00:12:27,480 Speaker 1: It coming up. Is it possible to predict murder? Stay 234 00:12:27,480 --> 00:12:39,760 Speaker 1: with us? Well, it just is remarkable to think that 235 00:12:40,000 --> 00:12:44,440 Speaker 1: kind of Hiding underneath all of these numbers and statistics 236 00:12:44,480 --> 00:12:48,360 Speaker 1: and maths are so many life and death decisions. And 237 00:12:48,400 --> 00:12:49,960 Speaker 1: the other thing I wanted to ask you about was 238 00:12:50,280 --> 00:12:54,360 Speaker 1: your work on one of the most prolific serial killer 239 00:12:54,400 --> 00:12:56,880 Speaker 1: cases of all time, the Herald Shipment case. Just for 240 00:12:56,920 --> 00:13:00,960 Speaker 1: a US audience, can you explain that case and what 241 00:13:00,960 --> 00:13:01,800 Speaker 1: your work on it was. 242 00:13:02,760 --> 00:13:05,880 Speaker 2: Harold Shepman was a family doctor who, over a twenty 243 00:13:05,960 --> 00:13:08,920 Speaker 2: year period, murdered at least two hundred and fifty of 244 00:13:08,920 --> 00:13:13,199 Speaker 2: his patients and possibly up to four hundred without being caught, 245 00:13:13,280 --> 00:13:17,760 Speaker 2: of course, until he finally, rather stupidly forged, rather badly, 246 00:13:17,880 --> 00:13:20,040 Speaker 2: did a rather bad forgery of a will in order 247 00:13:20,080 --> 00:13:24,000 Speaker 2: to inherit some money. Absolute madness. And it was this 248 00:13:24,040 --> 00:13:26,719 Speaker 2: a woman whose daughter was a solicitor and looked at 249 00:13:26,720 --> 00:13:30,520 Speaker 2: this willness, it just didn't believe it. And so suspicions rose, 250 00:13:30,600 --> 00:13:35,760 Speaker 2: and finally he was arrested and they exhumed the last 251 00:13:35,840 --> 00:13:40,079 Speaker 2: fifteen patients that had died, and they all had incredibly 252 00:13:40,160 --> 00:13:44,360 Speaker 2: high levels of diamorphine heroin essentially in their bloodstream. I mean, 253 00:13:44,360 --> 00:13:46,080 Speaker 2: he only got away with it because there were never 254 00:13:46,120 --> 00:13:50,840 Speaker 2: any post mortems. There's many old people and they liked him. 255 00:13:50,920 --> 00:13:53,200 Speaker 2: He was a very trusted family doctor for many people. 256 00:13:54,080 --> 00:13:57,040 Speaker 2: He did a lot of home visits, and that's of 257 00:13:57,040 --> 00:14:01,000 Speaker 2: course when he murdered people. So when when someone went 258 00:14:01,040 --> 00:14:03,720 Speaker 2: back and looked at all the certificates of the time 259 00:14:03,760 --> 00:14:07,400 Speaker 2: of death, for most people, people died all times of 260 00:14:07,400 --> 00:14:09,400 Speaker 2: the day and night, the sort of uniform distribution of 261 00:14:09,480 --> 00:14:12,000 Speaker 2: the twenty four hours, Harold Shipman's deaths had a great, 262 00:14:12,200 --> 00:14:16,280 Speaker 2: huge spike between around one to three in the afternoon 263 00:14:17,040 --> 00:14:18,360 Speaker 2: when he did his home visits. 264 00:14:18,559 --> 00:14:20,640 Speaker 1: And what was your involvement personally with the case. 265 00:14:21,400 --> 00:14:24,560 Speaker 2: There was a public inquiry because the families quite really 266 00:14:24,560 --> 00:14:26,200 Speaker 2: on other people ask how do you get away with 267 00:14:26,240 --> 00:14:29,600 Speaker 2: it for so long? It's an absolute scandal, And the 268 00:14:29,800 --> 00:14:33,880 Speaker 2: public inquiry, I think very sensibly brought in quite a 269 00:14:33,880 --> 00:14:37,600 Speaker 2: substantial team of statisticians to look at the data, which 270 00:14:37,920 --> 00:14:41,000 Speaker 2: like the time of death data, but also the deaths 271 00:14:41,040 --> 00:14:44,000 Speaker 2: of all his patients when they had occurred, how it 272 00:14:44,040 --> 00:14:47,120 Speaker 2: compared with other doctors, how many would have been expected, 273 00:14:47,360 --> 00:14:51,720 Speaker 2: compared with how many ratually observed. And we used sort 274 00:14:51,760 --> 00:14:56,000 Speaker 2: of fairly standard industrial quality control methods to work out 275 00:14:56,040 --> 00:14:59,760 Speaker 2: when you could have spotted with considerable confidence when something 276 00:14:59,800 --> 00:15:03,000 Speaker 2: al was going on. So it's like industrial quality control 277 00:15:03,040 --> 00:15:05,280 Speaker 2: methods spot when a production line is going out a 278 00:15:05,280 --> 00:15:08,280 Speaker 2: bit out of kilter. They would have been developed over decades, 279 00:15:08,480 --> 00:15:11,040 Speaker 2: and we used those for his death rates and worked 280 00:15:11,080 --> 00:15:14,040 Speaker 2: out he could have been caught after about forty deaths, 281 00:15:14,440 --> 00:15:17,840 Speaker 2: or he could have been identified as being odd. In 282 00:15:17,880 --> 00:15:21,640 Speaker 2: other words, someone could have done an investigation. Now, Shipman, 283 00:15:21,720 --> 00:15:24,840 Speaker 2: when the algorithm that we developed was applied to a 284 00:15:24,960 --> 00:15:29,240 Speaker 2: thousand other gps without their knowledge, there were six who 285 00:15:29,800 --> 00:15:31,080 Speaker 2: as bad as Shipman. 286 00:15:31,040 --> 00:15:32,520 Speaker 1: I used many deaths on their watch. 287 00:15:32,720 --> 00:15:37,840 Speaker 2: Exactly why do you think that was? They were really 288 00:15:37,880 --> 00:15:42,440 Speaker 2: good gps who were working in retirement communities and who 289 00:15:42,480 --> 00:15:45,000 Speaker 2: were enabling their patients to die at home rather than 290 00:15:45,000 --> 00:15:48,360 Speaker 2: going to hospital by being really good caring gps, and 291 00:15:48,400 --> 00:15:52,480 Speaker 2: so they were signing a lot of death certificates. And 292 00:15:52,680 --> 00:15:54,640 Speaker 2: these were really good people, but they had very high 293 00:15:54,640 --> 00:15:57,240 Speaker 2: death rates. So I used this as an example all 294 00:15:57,280 --> 00:16:01,240 Speaker 2: the time about how statistics is about correlation not causation. 295 00:16:01,760 --> 00:16:04,200 Speaker 2: You know, if someone has got a high death rate, 296 00:16:04,720 --> 00:16:08,280 Speaker 2: it's an indication that someone perhaps should look at the data, 297 00:16:08,440 --> 00:16:10,680 Speaker 2: but you cannot conclude the cause for that. 298 00:16:10,960 --> 00:16:13,080 Speaker 1: Well, just in the last few weeks, the UK has 299 00:16:13,120 --> 00:16:17,800 Speaker 1: announced an algorithm to predict the likelihood of committing murder. 300 00:16:18,200 --> 00:16:21,400 Speaker 2: Well, those algorithms have been around for ages, but the 301 00:16:21,560 --> 00:16:24,440 Speaker 2: chance of predicting someone, all you'll do is find a 302 00:16:24,520 --> 00:16:26,440 Speaker 2: small change in odds. You're never going to be able 303 00:16:26,440 --> 00:16:30,160 Speaker 2: to predict an event like that. At an individual level. 304 00:16:30,200 --> 00:16:32,480 Speaker 2: You'll be able to predict someone's at someone it's slightly 305 00:16:32,560 --> 00:16:36,600 Speaker 2: increased risk. But there's so much puff befind these algorithms 306 00:16:36,640 --> 00:16:39,280 Speaker 2: to make it. You know, they get headlines, but actually 307 00:16:39,360 --> 00:16:42,640 Speaker 2: I'm deeply skeptical about their actual ability and certainly to 308 00:16:42,800 --> 00:16:44,600 Speaker 2: predict events like murders. 309 00:16:45,080 --> 00:16:47,120 Speaker 1: You spend quite a bit of time working on public 310 00:16:47,160 --> 00:16:50,560 Speaker 1: inquiries and informing a public about various issues, and I 311 00:16:50,600 --> 00:16:52,440 Speaker 1: think you have one of the most interesting, if not 312 00:16:52,520 --> 00:16:56,400 Speaker 1: the most interesting title in British academia Professor for the 313 00:16:56,440 --> 00:16:59,240 Speaker 1: Public Understanding of Risk. Can you talk a little bit 314 00:16:59,280 --> 00:17:00,200 Speaker 1: about what that means? 315 00:17:00,800 --> 00:17:04,040 Speaker 2: Yeah, I think I am the one and only Professor 316 00:17:04,040 --> 00:17:07,000 Speaker 2: for the Public Understanding of Risk because after I retired 317 00:17:07,040 --> 00:17:11,080 Speaker 2: they renamed it when the next person got the funding. 318 00:17:11,160 --> 00:17:14,800 Speaker 2: So this was a fascinating I'd been an academic and 319 00:17:15,160 --> 00:17:18,160 Speaker 2: was doing okay, it'd got a good reputation, but fancied 320 00:17:18,160 --> 00:17:21,800 Speaker 2: a change in direction from the normal business of writing 321 00:17:21,920 --> 00:17:26,639 Speaker 2: papers and all that stuff, and then a philanthropist David Harding, 322 00:17:26,680 --> 00:17:29,959 Speaker 2: a hedge fund manager, wanted to endow a chair in 323 00:17:30,080 --> 00:17:33,560 Speaker 2: Cambridge that was to do with the improving the way 324 00:17:33,600 --> 00:17:37,359 Speaker 2: that statistics and risk was discussed in society because he 325 00:17:37,400 --> 00:17:39,240 Speaker 2: got so fed up with all the stories in the 326 00:17:39,280 --> 00:17:42,760 Speaker 2: news and all the misunderstandings, and so he paid for 327 00:17:42,840 --> 00:17:45,080 Speaker 2: this chair. And if you gave three and a half 328 00:17:45,160 --> 00:17:47,240 Speaker 2: million pounds the University of Cambridge, you could have a 329 00:17:47,320 --> 00:17:49,919 Speaker 2: chair of absolutely anything. 330 00:17:50,040 --> 00:17:51,520 Speaker 1: And he had the good grace not to name it 331 00:17:51,560 --> 00:17:52,720 Speaker 1: after himself. 332 00:17:53,040 --> 00:17:55,640 Speaker 2: Exactly why it was the Winton. It was the Winton 333 00:17:55,720 --> 00:17:58,199 Speaker 2: Professor for the Public Understanding of risk, so which was 334 00:17:58,359 --> 00:18:01,080 Speaker 2: just fine because he was very good. I always give 335 00:18:01,359 --> 00:18:03,960 Speaker 2: my career advice for young people now is to say, 336 00:18:04,040 --> 00:18:06,680 Speaker 2: find a billionaire and get him to give you lots 337 00:18:06,680 --> 00:18:08,960 Speaker 2: of money to do what you feel like, because he 338 00:18:09,080 --> 00:18:11,680 Speaker 2: just gave the money and then completely hands off. 339 00:18:11,760 --> 00:18:17,199 Speaker 1: In their capacity. What was the biggest misunderstanding you encountered 340 00:18:17,240 --> 00:18:18,720 Speaker 1: about how the public understand risk? 341 00:18:19,720 --> 00:18:23,560 Speaker 2: Oh goodness, that's so difficult. I mean you could the 342 00:18:23,600 --> 00:18:27,440 Speaker 2: absolutely standard one, of course, which the media don't help 343 00:18:27,480 --> 00:18:30,679 Speaker 2: with is the difference between absolute and relative risk. So 344 00:18:31,000 --> 00:18:33,320 Speaker 2: you know, the media stories are always full of oh, well, 345 00:18:33,320 --> 00:18:36,160 Speaker 2: if you eat meat, it's going to increase your risk 346 00:18:36,200 --> 00:18:39,239 Speaker 2: of bowel cancer by twenty percent or so on. And 347 00:18:39,280 --> 00:18:41,240 Speaker 2: that's a relative risk. And I think it's actually true 348 00:18:41,240 --> 00:18:44,440 Speaker 2: that actually red meat is some process meat in particular, 349 00:18:44,520 --> 00:18:47,639 Speaker 2: is associated with an increased risk of bow cancer. And 350 00:18:47,680 --> 00:18:51,800 Speaker 2: that's what gets in the headlines increase risk. I've got lovely, 351 00:18:52,080 --> 00:18:54,679 Speaker 2: you know, headlines of the killer bacon soandwich and this 352 00:18:54,720 --> 00:18:57,600 Speaker 2: sort of thing. But when you actually translated, and I 353 00:18:57,680 --> 00:19:00,320 Speaker 2: talk about this all the time to schools audiences, when 354 00:19:00,359 --> 00:19:02,560 Speaker 2: they hear a story like this, they want to know, well, 355 00:19:02,800 --> 00:19:04,520 Speaker 2: you know, is this the big number? Do we care 356 00:19:04,560 --> 00:19:06,520 Speaker 2: about this? And to know that, you have to know 357 00:19:06,600 --> 00:19:09,880 Speaker 2: twenty percent of what, in other words, the baseline risk 358 00:19:10,080 --> 00:19:12,399 Speaker 2: of which there is a twenty percent increase. Now, the 359 00:19:12,400 --> 00:19:16,520 Speaker 2: baseline risk of getting boo cancer, sadly is about six percent, 360 00:19:16,560 --> 00:19:19,240 Speaker 2: about one in sixteen will get it during our lifetime, sadly, 361 00:19:19,640 --> 00:19:24,120 Speaker 2: and a twenty percent increase over those six percentage points 362 00:19:24,880 --> 00:19:28,080 Speaker 2: is about seven percentage points. So that's out of one 363 00:19:28,160 --> 00:19:31,280 Speaker 2: hundred people eating a bacon sandwich every single day of 364 00:19:31,320 --> 00:19:35,360 Speaker 2: their lives, one extra will get bow cancer because of that. 365 00:19:35,960 --> 00:19:39,640 Speaker 2: And that's a complete different way of reframing the story 366 00:19:40,119 --> 00:19:43,639 Speaker 2: to make it look frankly fairly reassuring, especially if you 367 00:19:43,720 --> 00:19:47,960 Speaker 2: like bacon savwiches. So it's a great example of this 368 00:19:48,400 --> 00:19:52,199 Speaker 2: difference between relative risks and absolute risk because percentage, the 369 00:19:52,240 --> 00:19:55,000 Speaker 2: word percentage is used for both. It's just in one 370 00:19:55,040 --> 00:19:57,520 Speaker 2: you're talking about a percentage increase and the other talking 371 00:19:57,560 --> 00:19:58,720 Speaker 2: about percentage points. 372 00:20:01,119 --> 00:20:04,159 Speaker 1: When we come back, we break down the probability that 373 00:20:04,240 --> 00:20:17,560 Speaker 1: AI could lead to human extinction. Stay with us. When 374 00:20:17,640 --> 00:20:21,200 Speaker 1: the consumer internet boomed in the late nineties early two 375 00:20:21,240 --> 00:20:24,879 Speaker 1: thousands and Google came along, you could either click Google 376 00:20:24,960 --> 00:20:29,080 Speaker 1: Search or I'm feeling lucky and I'm feeling Lucky would 377 00:20:29,119 --> 00:20:32,560 Speaker 1: bypass the search results and take you directly to a website. 378 00:20:32,640 --> 00:20:34,679 Speaker 1: And this is basically a way for Google to flex 379 00:20:34,720 --> 00:20:36,720 Speaker 1: and say like, this is how incredibly good we are 380 00:20:36,760 --> 00:20:39,320 Speaker 1: in search. And they've since abandoned the I'm feeling a 381 00:20:39,359 --> 00:20:42,879 Speaker 1: Lucky button, but actually, in parallel, the whole Internet in 382 00:20:42,920 --> 00:20:45,280 Speaker 1: the last two or three years has become an I'm 383 00:20:45,280 --> 00:20:47,800 Speaker 1: feeling Lucky engine in the sense that you now get 384 00:20:48,240 --> 00:20:51,440 Speaker 1: a generative AI response rather than a selection of links 385 00:20:51,480 --> 00:20:53,760 Speaker 1: to follow, or at least you get both. How do 386 00:20:53,880 --> 00:20:59,920 Speaker 1: you see that incredible cultural shift of sort of outsourcing information, 387 00:21:00,080 --> 00:21:03,439 Speaker 1: summarization and predictions to large language models. 388 00:21:04,200 --> 00:21:07,320 Speaker 2: I think it's great. I'm a real fan of the 389 00:21:07,359 --> 00:21:09,800 Speaker 2: AI summary. So as long as you know, just like 390 00:21:10,000 --> 00:21:13,359 Speaker 2: using any large language model, you have to grasp the 391 00:21:13,359 --> 00:21:15,679 Speaker 2: fact that it doesn't know anything at all. You know, 392 00:21:15,720 --> 00:21:18,640 Speaker 2: it is. All it does is string words together and 393 00:21:18,880 --> 00:21:21,840 Speaker 2: comes up with something that sounds plausible. Now maybe in fact, 394 00:21:22,240 --> 00:21:24,760 Speaker 2: as we all know, it comes up if it talks 395 00:21:24,800 --> 00:21:27,240 Speaker 2: about facts, it can be deeply wrong and say all 396 00:21:27,240 --> 00:21:29,760 Speaker 2: sorts of things that are just incorrect. So it has 397 00:21:29,840 --> 00:21:32,680 Speaker 2: to be taken with a huge pinch of salt when 398 00:21:32,720 --> 00:21:36,840 Speaker 2: it's saying anything factual. When it's summarizing an argument, or 399 00:21:36,880 --> 00:21:39,960 Speaker 2: perhaps you know, with a discussion on a topic, I 400 00:21:39,960 --> 00:21:41,960 Speaker 2: think it can be enormously helpful. I mean, if you 401 00:21:42,119 --> 00:21:44,960 Speaker 2: just ask it a fact, you know, what's the capital 402 00:21:44,960 --> 00:21:48,320 Speaker 2: of somewhere, then it'll generally be right. But I think, 403 00:21:48,359 --> 00:21:52,520 Speaker 2: as someone who's who worked on uncertainty in AI forty 404 00:21:52,720 --> 00:21:56,200 Speaker 2: years ago, we thought we'd solved it in nineteen eighty six. 405 00:21:56,240 --> 00:21:57,520 Speaker 1: Well, how do you think you've solved it? 406 00:21:57,800 --> 00:22:00,960 Speaker 2: Oh, because then the model's much more We in the 407 00:22:01,000 --> 00:22:04,760 Speaker 2: mid nineteen eighties, the way of actually handling probability, first 408 00:22:04,800 --> 00:22:08,080 Speaker 2: within rule based systems and then within basian networks was 409 00:22:08,160 --> 00:22:11,760 Speaker 2: really developed. It was extremely successful, but of course that's 410 00:22:11,840 --> 00:22:13,959 Speaker 2: in much smaller networks. 411 00:22:14,560 --> 00:22:17,879 Speaker 1: We're living in this extraordinary moment. I mean, Jeffrey Hinton 412 00:22:17,920 --> 00:22:20,840 Speaker 1: has said there's a thirty percent chance that AI will 413 00:22:20,920 --> 00:22:24,000 Speaker 1: drive human extinction in the next twenty to thirty years. 414 00:22:24,680 --> 00:22:31,119 Speaker 1: There are rogue genetic scientists editing the human gene line. 415 00:22:31,600 --> 00:22:36,639 Speaker 1: There is uncertainty about whether the COVID pandemic was you know, 416 00:22:36,720 --> 00:22:40,560 Speaker 1: something creating a lab or something emerged organically. I muchine 417 00:22:40,600 --> 00:22:43,040 Speaker 1: what you say is timeless. But how do you suggest 418 00:22:43,119 --> 00:22:47,199 Speaker 1: navigating this particular scientific technological moment. 419 00:22:47,880 --> 00:22:51,000 Speaker 2: Yeah, I think again by trying to think coldly about 420 00:22:51,000 --> 00:22:53,720 Speaker 2: it instantly. Jeff, when I mentioned working on AAR in 421 00:22:53,760 --> 00:22:56,159 Speaker 2: the nineteen eighties, I mean I was. I was in 422 00:22:56,160 --> 00:22:58,359 Speaker 2: Cambridge and Jeff was in Cambridge then, and we used 423 00:22:58,400 --> 00:23:00,720 Speaker 2: to think, oh, poor, because Jeff was going around saying, well, 424 00:23:00,760 --> 00:23:03,000 Speaker 2: these neural networks, one day they'll be big enough to 425 00:23:03,040 --> 00:23:05,119 Speaker 2: really be able to act in an intelligent way. And 426 00:23:05,160 --> 00:23:05,560 Speaker 2: we used to. 427 00:23:05,520 --> 00:23:06,320 Speaker 1: Think poor Jeff. 428 00:23:07,880 --> 00:23:10,520 Speaker 2: He's Gary is banging on about his networks again, Why 429 00:23:10,560 --> 00:23:14,600 Speaker 2: didn't he just give up? Because he was right. It 430 00:23:14,800 --> 00:23:16,800 Speaker 2: took a long time, but he was right. 431 00:23:17,200 --> 00:23:19,240 Speaker 1: He was on tech stuff not too long ago. And 432 00:23:19,359 --> 00:23:21,119 Speaker 1: I asked him, how did you count with the number 433 00:23:21,359 --> 00:23:24,919 Speaker 1: thirty percent for the probability that AI will drive humans extinction? 434 00:23:25,000 --> 00:23:26,680 Speaker 1: He said, well, I knew it was more than one 435 00:23:26,680 --> 00:23:28,080 Speaker 1: percent and less than one hundred percent. 436 00:23:28,320 --> 00:23:30,920 Speaker 2: It means a non trivial chance of this happening. Really, 437 00:23:31,640 --> 00:23:33,480 Speaker 2: I think obviously there is a danger of tech. I 438 00:23:33,480 --> 00:23:35,760 Speaker 2: mean in the book, I talk about surveys that have 439 00:23:35,800 --> 00:23:38,280 Speaker 2: been done of people, you know, looking at the chance 440 00:23:38,359 --> 00:23:41,840 Speaker 2: of existential risk to the population into the world in 441 00:23:41,880 --> 00:23:44,720 Speaker 2: general and from tech. And because people do have judgments, 442 00:23:44,720 --> 00:23:47,000 Speaker 2: like Jeff, does you know, I think it probably is 443 00:23:47,040 --> 00:23:48,880 Speaker 2: a non zero. Probably we could argue about how big 444 00:23:48,880 --> 00:23:49,240 Speaker 2: it was. 445 00:23:49,560 --> 00:23:50,560 Speaker 1: How do you measure it? 446 00:23:50,880 --> 00:23:52,960 Speaker 2: Oh, well, well you can't measure it. There's no measurement 447 00:23:53,000 --> 00:23:55,760 Speaker 2: because it's not a number. There's no truth out there, 448 00:23:55,880 --> 00:23:57,240 Speaker 2: so you can't measure it. 449 00:23:57,359 --> 00:24:00,399 Speaker 1: So remember that, so you simulate different The best way 450 00:24:00,440 --> 00:24:01,880 Speaker 1: to approximate with simulation. 451 00:24:01,560 --> 00:24:04,920 Speaker 2: Or I wouldn't believe any simulated futures either. I mean, 452 00:24:05,080 --> 00:24:08,800 Speaker 2: the simulating possible futures is fantastic method. We've used it 453 00:24:08,840 --> 00:24:11,360 Speaker 2: all the time in prediction work, and that's what's done 454 00:24:11,359 --> 00:24:13,479 Speaker 2: in a lot of weather forecasting as well. So but 455 00:24:13,720 --> 00:24:16,600 Speaker 2: it's a good idea. I just don't think you'd you'd 456 00:24:16,680 --> 00:24:18,760 Speaker 2: be so reliant on the assumptions in your models. No, 457 00:24:18,840 --> 00:24:22,720 Speaker 2: these are personal judgments. But just like an intelligence analysts 458 00:24:22,760 --> 00:24:26,200 Speaker 2: will be assessing probabilities even now about what will happen 459 00:24:26,240 --> 00:24:28,960 Speaker 2: in the Russia Ukraine war in a year's time and things. 460 00:24:29,040 --> 00:24:31,640 Speaker 2: So these are judgments that we should all be assessing. 461 00:24:31,680 --> 00:24:34,080 Speaker 2: I think is really valuable to work in separate teams 462 00:24:34,119 --> 00:24:36,640 Speaker 2: to come up with these judgments and the reasons for them. 463 00:24:36,680 --> 00:24:38,960 Speaker 2: So I like this sort of exercise, and I'm glad 464 00:24:38,960 --> 00:24:40,720 Speaker 2: you have put a number on it. I think my 465 00:24:40,840 --> 00:24:43,600 Speaker 2: number would be considerably lower, but you know he knows 466 00:24:43,680 --> 00:24:46,439 Speaker 2: more than I do. But so I think the crucial 467 00:24:46,440 --> 00:24:48,520 Speaker 2: answers once we get to something that's what you might 468 00:24:48,520 --> 00:24:51,720 Speaker 2: call a distinct possibility, what do you do about it? 469 00:24:52,800 --> 00:24:54,520 Speaker 2: You know, where are the controls? You know that you 470 00:24:54,600 --> 00:24:56,480 Speaker 2: need to think about where you don't just sit back 471 00:24:56,520 --> 00:24:58,800 Speaker 2: as casual partisips or that many of us will be 472 00:24:58,840 --> 00:25:01,560 Speaker 2: just an audience, but that's not true of the people 473 00:25:01,640 --> 00:25:04,479 Speaker 2: working in this area, or the regulators, or the people 474 00:25:04,880 --> 00:25:06,840 Speaker 2: who might be able to do something about it. So 475 00:25:06,880 --> 00:25:08,960 Speaker 2: I think it does, as people, of course have said, 476 00:25:09,400 --> 00:25:12,159 Speaker 2: you know, generate the question, well, okay, how can we 477 00:25:12,200 --> 00:25:13,440 Speaker 2: reduce that probability? 478 00:25:13,760 --> 00:25:15,720 Speaker 1: I want to bring us back to the book, I mean, 479 00:25:15,800 --> 00:25:20,960 Speaker 1: which reads as a clarion call to learn to embrace uncertainty. 480 00:25:21,480 --> 00:25:23,600 Speaker 1: I mean, is that a fair characterization. What do you 481 00:25:23,640 --> 00:25:25,560 Speaker 1: hope that your readers will take away from this book? 482 00:25:25,800 --> 00:25:27,639 Speaker 2: Yeah? I mean I always say it's not a self 483 00:25:27,640 --> 00:25:30,200 Speaker 2: help book, although see people do seem to get quite 484 00:25:30,240 --> 00:25:32,800 Speaker 2: a lot from it sometimes of the fact that you know, 485 00:25:32,960 --> 00:25:36,080 Speaker 2: by owning up to uncertainty, first of all, that it 486 00:25:36,119 --> 00:25:38,560 Speaker 2: shouldn't be something to dread. We live with uncertainty all 487 00:25:38,600 --> 00:25:40,119 Speaker 2: the time. We enjoy it. It'd be awful to be 488 00:25:40,119 --> 00:25:42,720 Speaker 2: certain about everything. Can't think of anything worse to live alone. 489 00:25:42,800 --> 00:25:44,760 Speaker 2: And I always ask audiences if I could tell you, 490 00:25:44,800 --> 00:25:46,680 Speaker 2: would you know, want to know when you're going to die? 491 00:25:47,320 --> 00:25:50,600 Speaker 2: And a few people would always just a few, they'd 492 00:25:50,640 --> 00:25:51,959 Speaker 2: like to be able to plan and things, and that 493 00:25:51,960 --> 00:25:54,439 Speaker 2: you know that somebody, but nearly everybody does want to know. 494 00:25:54,600 --> 00:25:56,520 Speaker 2: You don't read. You don't look at you know, on 495 00:25:56,560 --> 00:25:59,199 Speaker 2: a you know on a thriller series. You don't go 496 00:25:59,240 --> 00:26:01,919 Speaker 2: for the last efforts to find out what the conclusion. 497 00:26:02,160 --> 00:26:03,879 Speaker 2: You don't want to know the sports result before you 498 00:26:03,920 --> 00:26:06,879 Speaker 2: see the match, if it's recorded. And so the point 499 00:26:06,960 --> 00:26:09,000 Speaker 2: is that we live with uncertainty. I think we should 500 00:26:09,040 --> 00:26:12,440 Speaker 2: embrace it. It will never go away, but there are ways 501 00:26:12,560 --> 00:26:13,440 Speaker 2: to explore it. 502 00:26:13,600 --> 00:26:16,520 Speaker 1: I want to close with this, David, you said recently, 503 00:26:16,880 --> 00:26:20,600 Speaker 1: my wildest prediction is that people will stop making predictions. 504 00:26:21,200 --> 00:26:25,479 Speaker 2: Mm. Well, that's the one I would love. And what 505 00:26:25,520 --> 00:26:28,200 Speaker 2: I mean by that is predictions where they say what's 506 00:26:28,240 --> 00:26:30,639 Speaker 2: going to happen. And what I want is the Jeff 507 00:26:30,680 --> 00:26:33,439 Speaker 2: Hinton approach where you give probability, so what's going to happen? 508 00:26:33,960 --> 00:26:36,040 Speaker 2: And those probabilities may be good. I think Jeff's a 509 00:26:36,040 --> 00:26:37,760 Speaker 2: bit high. It may not be, but at least we've 510 00:26:37,760 --> 00:26:39,760 Speaker 2: got something we know where they are. He's not saying 511 00:26:39,760 --> 00:26:42,479 Speaker 2: it's going to happen or it's not going to happen. 512 00:26:42,920 --> 00:26:45,080 Speaker 2: I don't care about whether someone thinks something's going to 513 00:26:45,080 --> 00:26:46,480 Speaker 2: happen or not going to I couldn't care less. I 514 00:26:46,480 --> 00:26:49,320 Speaker 2: wouldn't their probabilities of whether it's going to happen. That's 515 00:26:49,359 --> 00:26:52,720 Speaker 2: why sports pundits when they're chatting on if you're just 516 00:26:52,800 --> 00:26:55,080 Speaker 2: chatting casually, you might say, oh, I think this is 517 00:26:55,119 --> 00:26:59,080 Speaker 2: the result. But of course anyone taking sports seriously doesn't 518 00:26:59,080 --> 00:27:00,840 Speaker 2: say who's going to win. Going to look they work 519 00:27:00,880 --> 00:27:03,080 Speaker 2: out the odds, because they're going to be going on 520 00:27:03,119 --> 00:27:06,520 Speaker 2: to the betting exchanges and checking if they can get better. Okay, 521 00:27:06,560 --> 00:27:08,680 Speaker 2: you know if there's differences between the odds they think 522 00:27:08,720 --> 00:27:11,680 Speaker 2: were appropriate the odds being offered by on their betting exchanges. 523 00:27:11,760 --> 00:27:15,960 Speaker 2: So serious sporting people only think in terms of probabilities. 524 00:27:21,760 --> 00:27:23,280 Speaker 1: David, thank you so much for joining us today and 525 00:27:23,320 --> 00:27:23,840 Speaker 1: tex stuff. 526 00:27:23,920 --> 00:27:25,280 Speaker 2: It's been a real pleasure. 527 00:27:30,040 --> 00:27:33,280 Speaker 1: For tech stuff. I'm os Voloscian. This episode was produced 528 00:27:33,280 --> 00:27:38,040 Speaker 1: by Eliza Dennis and Adriana Tapia. It was executive produced 529 00:27:38,040 --> 00:27:42,120 Speaker 1: by me, Karen Price and Kate Osborne for Kaleidoscope and 530 00:27:42,320 --> 00:27:46,440 Speaker 1: Katrina Norvelle for iHeart Podcasts. Jack Insley mixed this episode 531 00:27:46,640 --> 00:27:49,760 Speaker 1: and Kyle Murdoch Rudel theme song. Join us on Friday 532 00:27:49,760 --> 00:27:52,080 Speaker 1: for the Week in Tech. Karen and I will run 533 00:27:52,119 --> 00:27:55,800 Speaker 1: through all the most important tech headlines, including some you 534 00:27:55,840 --> 00:27:58,960 Speaker 1: may have missed, and please rate and review the show 535 00:27:59,000 --> 00:28:01,960 Speaker 1: in Apple Podcasts, Spotify, and reach out to us over 536 00:28:02,040 --> 00:28:16,200 Speaker 1: email at tech Stuff podcast at gmail dot com.