WEBVTT - AI Expert Toby Walsh: Are We Waiting for Atrocities Before We Act on AI? 0:00:00.080 --> 0:00:03.200 Debbie Walsh when straight talking my pleasure. I mean this 0:00:03.600 --> 0:00:08.639 topic artificial intelligence like it's nearly out of control in 0:00:08.680 --> 0:00:13.640 that most of us, not you, but most of us, 0:00:14.280 --> 0:00:17.000 we don't know whether we believe what we read. We 0:00:17.079 --> 0:00:19.080 don't know whether to be scared of what we read. 0:00:19.880 --> 0:00:22.119 We don't know whether they'd be excited with what we read. 0:00:23.680 --> 0:00:27.000 Now you're an expert in this area, How does an 0:00:27.040 --> 0:00:29.360 expert feel about artificial intelligence today? 0:00:30.000 --> 0:00:32.120 Well, I think the correct feeling is to be both 0:00:32.159 --> 0:00:35.239 excited and fearful. I think it's going to bring some 0:00:35.320 --> 0:00:38.839 goodness a bad I've had some time to prepare for this. 0:00:39.040 --> 0:00:41.680 I mean, I've been thinking about it for fifty years now. 0:00:42.240 --> 0:00:43.400 Maybe give me a bit history on that. 0:00:43.479 --> 0:00:46.080 So funny anecdote. I phone my father up the other 0:00:46.120 --> 0:00:49.360 day to wish him happy ninety first birth. Wow, And 0:00:49.400 --> 0:00:50.680 I think, you know, he's got to the age where 0:00:50.680 --> 0:00:53.800 you can start being truthful with me, he said, said, 0:00:53.840 --> 0:00:56.160 when you went to university, you know, forty odd years 0:00:56.200 --> 0:00:59.120 ago to study that artificial intelligence, I thought it was 0:00:59.160 --> 0:01:02.200 a bit odd because it was very odd back then 0:01:02.240 --> 0:01:04.680 there was actually only one university in the world that 0:01:04.680 --> 0:01:05.760 had a department of AI. 0:01:06.000 --> 0:01:06.280 Wow. 0:01:06.920 --> 0:01:09.720 And so it was probably odd. But then he finished, 0:01:09.720 --> 0:01:11.440 he said, it's a bit odd. But now I see 0:01:11.440 --> 0:01:12.440 you were playing the long game. 0:01:13.880 --> 0:01:18.080 You were then and you are now. Well which university 0:01:18.160 --> 0:01:20.560 wasn't that you studied artificial intelligence? That it was the 0:01:20.560 --> 0:01:22.960 only course there? So it was in the world. 0:01:22.800 --> 0:01:26.679 A University of Edinburgh and some some people had worked 0:01:26.800 --> 0:01:29.720 at Bletchley Park in the war with Alan Turing and press. 0:01:29.760 --> 0:01:32.960 Other people washed up there for strange reasons and set 0:01:33.080 --> 0:01:35.360 up what was the at the time, the only department 0:01:35.400 --> 0:01:36.319 of l intelligence. 0:01:36.840 --> 0:01:41.479 That's interesting, So Turing was the first computer let's call it. 0:01:41.560 --> 0:01:43.759 Yeah, he's the grandfather of beauty let's call it that, 0:01:44.440 --> 0:01:48.040 and oftence. So the very first scientific paper. 0:01:48.080 --> 0:01:49.320 Well, I was going to ask you what what what 0:01:49.400 --> 0:01:51.640 were they? I mean, we we all know him for 0:01:51.760 --> 0:01:54.160 and we know that you're a post World War two 0:01:54.360 --> 0:01:57.520 as being like, computers sort of kicked off and started 0:01:57.560 --> 0:01:59.320 to become quite a bit more prominent but a lot 0:01:59.440 --> 0:02:02.760 more statistic. But where was artificial intelligence in that whole 0:02:02.760 --> 0:02:03.800 process there? Then? 0:02:04.440 --> 0:02:08.520 So Alan Turing, of of course, a marvelous, magnificent mind 0:02:09.080 --> 0:02:12.520 who they invented the computer to crack the German code. 0:02:13.080 --> 0:02:16.760 Fantastic achievement. Which was kept largely secret at the time 0:02:17.360 --> 0:02:20.440 because we were still reading other people's codes, but shortened 0:02:20.480 --> 0:02:24.400 the war by perhaps two years. But Alan Turing, in 0:02:24.520 --> 0:02:26.920 his remarkable mind, thought, well, what can we do with 0:02:26.960 --> 0:02:29.919 these computers beyond the end of the war. Will be 0:02:30.000 --> 0:02:31.760 one day to be able to use them to help 0:02:31.840 --> 0:02:35.280 us do things that humans humans quire intelligence to do. 0:02:35.360 --> 0:02:37.280 And he wrote what is generally considered to be the 0:02:37.360 --> 0:02:41.560 very scientific paper about AI. And he asked the question, 0:02:41.800 --> 0:02:44.160 a really important question, which is, how will we know 0:02:44.200 --> 0:02:47.560 when we succeed? I mean, how will we know that 0:02:47.600 --> 0:02:50.440 machines are thinking? And he proposed what was now known 0:02:50.480 --> 0:02:52.640 as the Turing test and aimed after him. 0:02:52.560 --> 0:02:54.680 The Turing test, Yes, and what is that? 0:02:54.919 --> 0:02:57.880 Well, it's the imitation game. There was a famous movie 0:02:57.960 --> 0:03:01.720 that was made around this, which is that it's very 0:03:01.720 --> 0:03:04.000 hard to say what artificial intelligence is because very hard 0:03:04.000 --> 0:03:05.799 to say what intelligence is. We don't have very good 0:03:05.840 --> 0:03:09.040 definition of intelligence. So how can we define artificial intelligence 0:03:09.040 --> 0:03:12.120 when we can barely define intelligence? So he sort of 0:03:12.120 --> 0:03:15.560 finessed the problem by saying, well, a very functional approach, 0:03:15.639 --> 0:03:17.680 which is, well, if you sat down with the computer 0:03:17.760 --> 0:03:20.520 and had a conversation, asked it questions and so on. 0:03:20.880 --> 0:03:22.560 I mean maybe in front of a termin or not 0:03:22.600 --> 0:03:25.480 actually sitting down with the computer. If you couldn't tell 0:03:25.520 --> 0:03:27.960 it from having a conversation with the human, then you 0:03:28.040 --> 0:03:31.600 might as well say it's thinking. And interestingly enough, that's 0:03:31.639 --> 0:03:33.840 what we have today. If you sit down with OGBT, 0:03:34.280 --> 0:03:36.360 you can have a conversation, you can ask questions. You 0:03:36.360 --> 0:03:39.200 can easily be fooled into thinking it's a person that 0:03:39.240 --> 0:03:41.800 you're talking to. And indeed lots of us are sitting 0:03:41.840 --> 0:03:44.000 down talking to it as though it's alf therapists. 0:03:44.040 --> 0:03:46.680 Well that's me, I mean, and not not so much 0:03:46.760 --> 0:03:49.520 in terms of mental health bed or a physical health bed. 0:03:49.520 --> 0:03:52.640 But I'm doing advice never I sort of I do. 0:03:54.320 --> 0:03:58.040 It's called reasonless check ideas with it. 0:03:58.040 --> 0:04:00.520 It's hard to imagine, but ten percent to the world's 0:04:00.520 --> 0:04:03.320 population weekly use chat gipt. 0:04:04.720 --> 0:04:07.800 I use it multiple times in a day. Not chatty, 0:04:08.120 --> 0:04:10.160 but as a copol. It doesn't matter one of the chatbots. 0:04:10.200 --> 0:04:11.880 I mean, they're all pretty much the most of each other, 0:04:11.880 --> 0:04:16.680 I think, so I actually funny I check off with 0:04:17.040 --> 0:04:20.440 Claude and Coppola. I actually asked them both the same 0:04:20.520 --> 0:04:23.480 questions both Whenever I asked question one, I asked the 0:04:23.680 --> 0:04:27.680 other just as a sensible check, just to see what 0:04:27.720 --> 0:04:30.560 the two outcomes are and also to see who suits 0:04:30.600 --> 0:04:33.480 me better in terms of the way the answer is, 0:04:33.920 --> 0:04:36.440 the way the answer is delivered. They don't always come 0:04:36.520 --> 0:04:38.279 up with the same answer. 0:04:38.720 --> 0:04:42.760 They don't. I mean they are tuned with a particular character. 0:04:42.839 --> 0:04:45.000 If you might say, you know, yeah, there is a character. 0:04:45.839 --> 0:04:49.480 Yeah that there training personality that was Yeah, in some 0:04:49.520 --> 0:04:52.360 sense they do because there's a choice of answers they 0:04:52.360 --> 0:04:55.240 could give you, and they've been tuned to answer in 0:04:55.240 --> 0:05:00.479 a particular way. So if you use masks, chappots rock, 0:05:01.200 --> 0:05:06.280 that's got quite a libertarian, free flowing, freedom of speech 0:05:06.480 --> 0:05:10.680 approach to answering certain questions. And you know, chat GPT 0:05:10.960 --> 0:05:14.400 is a bit more middle of the road, and there's 0:05:14.440 --> 0:05:16.520 you know, there was one chatbot that's been trained on 0:05:16.600 --> 0:05:19.719 four chair, which is this dark corner of the web, 0:05:20.360 --> 0:05:23.680 and as you can imagine, that is you know, racist misogynists. 0:05:23.880 --> 0:05:25.720 Really que what's it called? 0:05:28.680 --> 0:05:29.839 I try to remember the name. 0:05:29.880 --> 0:05:31.159 Now, do you have to go into the dark web 0:05:31.200 --> 0:05:31.560 to get it? 0:05:32.320 --> 0:05:32.560 No? 0:05:32.560 --> 0:05:35.120 No, no, it's just to dellow the app just an app, 0:05:35.640 --> 0:05:39.400 but it's trained on let let's call the dark ideas. 0:05:39.560 --> 0:05:43.360 Yes, very dark. And then and then, because these chatbots 0:05:43.400 --> 0:05:46.400 are just essentially mirrors that they've been trained on huge 0:05:46.400 --> 0:05:49.760 amounts of the Internet, a third of the Internet. And 0:05:49.800 --> 0:05:53.680 you know, actually what surprises me is how polite and 0:05:53.720 --> 0:05:55.640 how correct they are, given that the Internet is full 0:05:55.680 --> 0:05:58.719 of people were saying very incorrect things in very impolite ways. 0:05:58.960 --> 0:06:02.600 That's interesting because they say, I go on to cap 0:06:02.640 --> 0:06:03.960 on and say, you know, like you know, I've got 0:06:03.960 --> 0:06:07.000 a sore throat or whatever. It knows immediately to say, listen, 0:06:07.040 --> 0:06:10.680 this is not medical advice. Who put those guardrails in? 0:06:11.000 --> 0:06:15.039 Ah? So Microsoft have actually put some on top of 0:06:15.040 --> 0:06:17.479 the chapbot put some guardrails. So if you're asking it 0:06:17.640 --> 0:06:19.960 medical advice, if you're asking it to make a bomb 0:06:20.080 --> 0:06:23.680 or do rather illegal activity, it's you know, it not 0:06:23.760 --> 0:06:25.800 do those things or say that, you know, give your 0:06:26.160 --> 0:06:28.400 medical advice with caution. Yeah, like you. 0:06:28.480 --> 0:06:30.560 And if you ask it to give you some swear 0:06:30.600 --> 0:06:34.840 words in another language, it actually it has a sort 0:06:34.839 --> 0:06:36.359 of a limit to it. 0:06:36.400 --> 0:06:38.920 So there's some simple filters have been added on top. 0:06:41.120 --> 0:06:44.400 Those filters are not perfect. So for example, most of 0:06:44.400 --> 0:06:47.279 the you know, the chatbots have been told if people 0:06:47.279 --> 0:06:50.159 are talking about self harm, or suicide, you know, to 0:06:50.200 --> 0:06:52.719 give the number for lifeline and so on. But it's 0:06:52.760 --> 0:06:56.520 easy enough, unfortunately, to get around those safeguards. You can say, well, 0:06:56.520 --> 0:06:59.159 I'm writing a book and the protagonist is thinking about 0:06:59.160 --> 0:06:59.920 committing suicide. 0:07:00.040 --> 0:07:04.039 Yeah, help me you trick it. Yes, that's interesting because 0:07:04.080 --> 0:07:06.640 often think about that. I say, I was thinking the 0:07:06.680 --> 0:07:09.320 other day because I was actually looking at asking questions 0:07:09.320 --> 0:07:12.520 about the sore throat there's some weeks ago, and I 0:07:12.600 --> 0:07:14.280 knew it wasn't going to be advice. I was thinking, 0:07:14.680 --> 0:07:18.760 if you were researching this, I'm asking it, what would 0:07:18.800 --> 0:07:20.640 you be suggesting of the sorts of things that a 0:07:20.680 --> 0:07:23.440 doctor might prescribe for this? And idmediately it didn't give 0:07:23.480 --> 0:07:27.040 me the call of the provisos. It just went straight 0:07:27.080 --> 0:07:29.880 into having a discussion about it and just on that. 0:07:31.800 --> 0:07:34.720 Sometimes if I might say, well give me a scollar, 0:07:35.080 --> 0:07:37.160 don't just give me an opinion that you scrape off 0:07:37.640 --> 0:07:39.720 you know, Instagram or and if you find this stuff, 0:07:40.000 --> 0:07:44.000 give me a scholarly article which which is peer reviewed, 0:07:44.880 --> 0:07:50.600 and I only want peer reviewed in high ranking journals, 0:07:51.480 --> 0:07:54.240 and it'll go and give you annotations as to where 0:07:54.240 --> 0:07:55.920 it's got the information. You go back and see the 0:07:55.920 --> 0:08:00.320 abstract or whatever the case may be. Let's say it 0:08:00.440 --> 0:08:02.920 was something that you wrote when you did your PhD. Yes, 0:08:03.960 --> 0:08:05.360 how do you feel about that? Do you do you 0:08:05.360 --> 0:08:08.480 feel like? How do you feel like being a so citation? 0:08:08.920 --> 0:08:11.800 You might think that's good, but how do you feel like, 0:08:12.200 --> 0:08:14.280 because then it's your work, how do you feel about that? 0:08:16.400 --> 0:08:17.040 Is it stealing? 0:08:18.000 --> 0:08:20.040 I think it is stealing. I mean, as a scientist, 0:08:20.120 --> 0:08:22.640 I publish my stuff for people who read to make 0:08:22.720 --> 0:08:25.720 the world more ological. So I don't mind for you know, 0:08:25.760 --> 0:08:28.600 my science, but all my books, you know, the books 0:08:28.600 --> 0:08:30.880 I wrote for the public, which gets sold and I 0:08:31.360 --> 0:08:35.160 get a small pittance back, they're also being trained on. 0:08:35.400 --> 0:08:38.760 And I'm lucky enough I've got an income as a 0:08:38.760 --> 0:08:41.320 sciences I don't have to don't have to depend upon 0:08:41.360 --> 0:08:44.319 selling books. But I do worry about the authors and 0:08:45.600 --> 0:08:48.760 the musicians and the graphic designers whose work has all 0:08:48.800 --> 0:08:51.560 been ingested and who depend upon the income which is 0:08:51.600 --> 0:08:54.080 now being undermined by these these bots. 0:08:54.120 --> 0:08:56.120 Because I see that Musk actually put up a tweet 0:08:56.120 --> 0:08:59.520 today on X talking about I think it's sure about anthropic. 0:08:59.600 --> 0:09:03.199 I mean, she's talking about his competition, but saying that 0:09:04.080 --> 0:09:06.840 you know they're stealing stuff does gro doesn't Grock do 0:09:06.920 --> 0:09:07.320 the same thing. 0:09:07.600 --> 0:09:09.840 I mean they all do exactly. So what is it? 0:09:10.160 --> 0:09:11.719 I mean I understand that is he just trying to 0:09:11.720 --> 0:09:14.240 pick a fight with ortmand or something like that? Is Yes, 0:09:14.320 --> 0:09:17.120 I mean he's a grieved because he he puts sold out. 0:09:17.200 --> 0:09:20.400 He put the hundred first hundred million into Open AI. 0:09:21.080 --> 0:09:22.560 It was going to be a you know, not for 0:09:22.600 --> 0:09:25.560 profit for the benefit of humanity. It's no longer. It's 0:09:25.559 --> 0:09:27.679 no longer open, it's no longer not for profit, it's 0:09:27.720 --> 0:09:30.240 no longer seems to be for the benefit the humanity. 0:09:30.720 --> 0:09:33.120 And he's been kicked out. So you know, he's got 0:09:33.120 --> 0:09:37.560 reasons backs because I mean ananthropic you know, have you 0:09:37.600 --> 0:09:39.839 know in the court, have you know just settled one 0:09:39.880 --> 0:09:43.800 and a half billion dollars for the copyright that they took, 0:09:43.840 --> 0:09:45.120 the copyrighted books they took. 0:09:46.240 --> 0:09:50.559 Where what's the business model of these things? I mean, 0:09:50.559 --> 0:09:51.560 how do they make money? 0:09:52.400 --> 0:09:55.440 Well, they're not making money, okay, so how does it work? 0:09:55.480 --> 0:09:59.199 Like Tropic, which makes Claude, is not making money. They're 0:09:59.200 --> 0:10:03.400 probably the Microsoft is not making money. Google is not 0:10:03.400 --> 0:10:05.720 making money on this, none of them. The only company 0:10:05.760 --> 0:10:06.920 that's making money isn't. 0:10:06.720 --> 0:10:09.439 In video yeah, because they're making parts they make. 0:10:09.600 --> 0:10:13.199 Yes, they're making the shovels, but they're making it, and 0:10:13.559 --> 0:10:16.000 they are they are the most valuable company in the world. 0:10:16.040 --> 0:10:18.800 They're worth three or four trillion dollars. Yeah. Today today 0:10:19.720 --> 0:10:22.080 they are the luckiest company. You know, it was by 0:10:22.240 --> 0:10:24.719 chance that they were selling what's proved to be the 0:10:24.760 --> 0:10:28.320 shovels for the AI revolution. They were selling graphics chips 0:10:28.320 --> 0:10:30.880 to people to make games with, and it's so just 0:10:31.160 --> 0:10:33.280 was just by luck turned out to be exactly the 0:10:33.320 --> 0:10:33.920 thing you needed. 0:10:34.000 --> 0:10:36.080 So were you someone who sits back and watches this, 0:10:36.760 --> 0:10:38.600 the growth of this stuff, do you think to yourself, 0:10:39.920 --> 0:10:42.240 how's all this stuff being funded? And who are the 0:10:42.240 --> 0:10:44.199 people funding? Why are they doing this? Why are they're 0:10:44.200 --> 0:10:46.520 prepared to go way out in a limit which told 0:10:46.520 --> 0:10:49.800 me the ortmands and you know, Mask and everybody else 0:10:50.320 --> 0:10:53.200 who's got a platform, what I'm in by platform, an 0:10:53.200 --> 0:10:57.760 AI platform? Why they why are they doing it? Is 0:10:58.080 --> 0:11:00.800 it power grab? What's the deal? What the what do 0:11:00.840 --> 0:11:01.200 you think? 0:11:01.240 --> 0:11:03.440 The thing that surprises me not the technology. The technology 0:11:03.480 --> 0:11:05.640 is where I would have imagined when I when I preaps, 0:11:05.800 --> 0:11:08.680 when I started as But the thing that surprises me 0:11:08.920 --> 0:11:11.520 is the the amount of money and the speed with 0:11:11.559 --> 0:11:15.360 which it's being executed, and I didn't expect it's several 0:11:15.400 --> 0:11:19.880 billion dollars a day being invested into AI. It's a 0:11:19.920 --> 0:11:23.679 third of the world's R and D budget is being spent. Wow, technology, 0:11:23.720 --> 0:11:28.920 we've we've never made such big bets and in some 0:11:28.960 --> 0:11:31.120 sense that that that is being returned. I mean open Ai, 0:11:31.440 --> 0:11:34.839 the company behind chatchapt that's the fastest growing company in 0:11:34.880 --> 0:11:37.280 the history I mean not the history of not just 0:11:37.320 --> 0:11:39.520 the history of AI or technology, the history of any 0:11:40.000 --> 0:11:42.880 any field of endeavor. No company has grown as quickly 0:11:42.880 --> 0:11:46.440 as open Ai. Before chat Chipity was released, they had 0:11:46.520 --> 0:11:50.520 no products, no income. A year later, after chachipep was released, 0:11:50.520 --> 0:11:53.800 they had a billion dollars a year annualize revenue. So 0:11:54.360 --> 0:11:58.200 who pays I mean people are paying subscriptions. Yeah, you know, 0:11:58.240 --> 0:12:02.640 they're doing deals with lots of other companies. They've got 0:12:02.960 --> 0:12:05.560 you know, tens of millions of people paying them, some 0:12:05.640 --> 0:12:08.920 of them paying two hundred dollars a month. They're not 0:12:08.960 --> 0:12:11.760 making money. So they went from a billion dollars a 0:12:11.840 --> 0:12:14.320 year a year later there were four billion. There were 0:12:14.360 --> 0:12:16.440 private companies. So they're not very open about it, but 0:12:16.559 --> 0:12:19.800 they're making somewhere somewhere more than ten billion dollars a 0:12:19.880 --> 0:12:23.319 year in revenue. But there's still enough in terms of 0:12:23.320 --> 0:12:25.240 what this No, but they're spending more than much more. 0:12:25.240 --> 0:12:28.079 They're spending twenty or thirty millions, So they're losing money 0:12:28.120 --> 0:12:28.400 out off. 0:12:28.679 --> 0:12:30.240 But why are they doing it? Because, I mean I 0:12:30.240 --> 0:12:33.560 can understand why you know, Google might have been doing 0:12:33.679 --> 0:12:35.400 or Netflix might have been doing, because they've got a 0:12:35.480 --> 0:12:38.280 model where they sort of hook you in and then 0:12:38.400 --> 0:12:40.600 once you get in there, they then start over time, 0:12:40.720 --> 0:12:43.120 they start to give you a free, free use and 0:12:43.120 --> 0:12:44.640 they say, look, if you don't want to have any ads, 0:12:44.679 --> 0:12:47.559 you can pay some bit extra. That's that Netflix model. 0:12:47.880 --> 0:12:50.079 That model has been done million times over. Do you 0:12:50.120 --> 0:12:52.480 think is it about making money? It's about power? 0:12:53.520 --> 0:12:56.959 Oh, it's a bit of both. There's certainly you know, 0:12:57.040 --> 0:13:00.480 they're trying to capitalize on the first move Advantageactly how 0:13:00.720 --> 0:13:02.840 Google succeeded. You know, if you go back to the 0:13:02.840 --> 0:13:05.079 beginnings of the Internet, I remember there was a there 0:13:05.080 --> 0:13:07.680 was a new aid search engine every day. Do you 0:13:07.679 --> 0:13:10.000 remember Auto Visa. I do. When Auto Vista came out, 0:13:10.040 --> 0:13:12.160 I remember saying to a Fred, oh that's it. Search 0:13:12.160 --> 0:13:14.760 engines are so good now with auto Vista that I'm 0:13:14.760 --> 0:13:16.600 never going to have to change my search engine again, 0:13:16.600 --> 0:13:18.480 because up to that point, every year or two you 0:13:18.720 --> 0:13:20.959 change the search engine. I remember getting to Audo Vester 0:13:21.040 --> 0:13:23.120 and thinking, oh, it's so good. Now I'm not going 0:13:23.200 --> 0:13:25.600 to have to change again. And of course Google came 0:13:25.600 --> 0:13:28.520 out and was much better. But what was interesting is 0:13:28.559 --> 0:13:32.160 that we've we've never had to change after Google because 0:13:32.240 --> 0:13:35.120 Google did. One thing is they got such a market share. 0:13:35.400 --> 0:13:37.920 They've got so much data that even if you came 0:13:37.960 --> 0:13:40.520 out with and there have been better algorithms for searching 0:13:40.520 --> 0:13:43.800 the web, you didn't have the data. And so no 0:13:43.920 --> 0:13:46.360 one can really compete against Google. And if you look 0:13:46.440 --> 0:13:48.439 like you're threatening to compete against Google, of course they 0:13:48.440 --> 0:13:51.760 just buy you out. He is right, So they've really 0:13:51.760 --> 0:13:53.760 Google has got an unassailable lead. 0:13:53.960 --> 0:13:55.520 That monopoly so monopolis. 0:13:55.520 --> 0:13:59.080 It is data monopolis, a natural monopoly. And you know, 0:13:59.120 --> 0:14:00.920 we should do something about that. Which is, you know, 0:14:01.559 --> 0:14:05.400 is it to the consumers benefit that you know, Google 0:14:05.520 --> 0:14:09.000 own YouTube, that Google that Google and the ad business 0:14:09.000 --> 0:14:11.800 that's also selling ads onto the global platform. Is it 0:14:11.840 --> 0:14:14.120 to the consumers you know benefit? There's only one social 0:14:14.200 --> 0:14:18.600 media platform, Facebook or Facebook's you know, other companies WhatsApp 0:14:18.679 --> 0:14:21.320 and Instagram and so on. I don't think it is, 0:14:21.560 --> 0:14:23.680 and we probably should. You know what we do when 0:14:23.920 --> 0:14:26.200 when people get monopolies break them into parts. 0:14:26.240 --> 0:14:29.760 Well, it's funny when I when I think about, you know, 0:14:30.040 --> 0:14:33.960 the Google, Facebook, just generally, then I think about all 0:14:33.960 --> 0:14:38.120 these other these other AI apps that people are using 0:14:38.320 --> 0:14:43.480 and downloading. Immediately think is the movie or the or 0:14:43.520 --> 0:14:47.280 the book written by Frank Boone called The Wizard of 0:14:47.280 --> 0:14:54.240 Oz and you know the so called wondament of you know, 0:14:54.840 --> 0:14:57.240 going to see this w Wizard is at the end 0:14:57.280 --> 0:14:59.120 of the day, is just a dude there with a curtain, 0:14:59.520 --> 0:15:03.400 just one per but this one person controlled the whole story. 0:15:03.640 --> 0:15:06.880 And really there's not that much in it except there's 0:15:06.880 --> 0:15:11.240 one so called wizard. And I often think about that show, 0:15:11.440 --> 0:15:13.800 the movie and the story or you know, the underlying 0:15:13.840 --> 0:15:16.000 story behind, and I think about, you know, what's going on, 0:15:16.200 --> 0:15:19.400 Sam Moltman. You know, we've got a mask, We've got Google, 0:15:19.560 --> 0:15:23.280 We've got Zuckerberg. There's some individuals who control these things, 0:15:23.920 --> 0:15:28.520 say a dozen of them, and are they the Wizard 0:15:28.600 --> 0:15:30.520 of Oz? Just you know, like there's not that much. 0:15:30.560 --> 0:15:32.360 They're just still human beings at the end of the day, 0:15:32.480 --> 0:15:35.440 people with all the law failures and all the foibills, 0:15:35.480 --> 0:15:37.680 and all the advantages and all the personality traits and 0:15:38.040 --> 0:15:41.240 good and bad, And therefore does it therefore make it 0:15:41.320 --> 0:15:42.000 very dangerous? 0:15:43.200 --> 0:15:45.520 It does. I mean, we're seeing a concentration of power 0:15:45.520 --> 0:15:49.560 into the hands of a few very influential, important people. 0:15:51.240 --> 0:15:53.480 Does it worry you though, Yeah, it does, it does, 0:15:53.560 --> 0:15:56.600 and their choices impact upon us. I mean, just just 0:15:56.640 --> 0:16:00.160 the last couple of days, we've been revealed that Facebook 0:16:00.520 --> 0:16:05.960 asked eighteen independent wellness and health experts whether what they 0:16:05.960 --> 0:16:10.120 should do about beauty filters on Instagram, and every one 0:16:10.160 --> 0:16:14.360 of those experts said they're harming young women, you shouldn't 0:16:14.400 --> 0:16:19.239 use them. And Mark Zuckerberg said overall those eighteen independent 0:16:19.240 --> 0:16:22.240 expert opinions and said, no, We're putting the beauty filters 0:16:22.240 --> 0:16:23.080 back in Instagram. 0:16:24.120 --> 0:16:27.520 He went against their advices, every one of them, and. 0:16:27.360 --> 0:16:29.520 What they were going to harm, that they were harming 0:16:29.920 --> 0:16:30.480 young women. 0:16:30.720 --> 0:16:33.840 And did Zuckerberg justify this? 0:16:34.000 --> 0:16:38.080 Yes, classic Silicon value justification. Freedom of speech. 0:16:38.880 --> 0:16:40.800 Ye, there's a throwaway line. 0:16:41.000 --> 0:16:41.360 It is. 0:16:41.840 --> 0:16:45.400 It is, yeah, and it's one of It's a United 0:16:45.440 --> 0:16:47.520 States part of the constitution. So it's hard to sort 0:16:47.520 --> 0:16:49.640 of argue, yes, we don't actually sort of have a 0:16:49.720 --> 0:16:52.600 constitutionalized here in Australia, but maybe that's the reason why 0:16:52.680 --> 0:16:56.440 we have the under sixteens sort of legislation coming through. 0:16:56.480 --> 0:16:58.440 We do, but we have a duty to protect young people. 0:16:59.200 --> 0:17:02.160 We've seen that. You know, these people are not you know, 0:17:02.400 --> 0:17:03.680 mindful of those duties. 0:17:04.200 --> 0:17:05.560 What are the things that worry you most? 0:17:05.640 --> 0:17:05.840 Then? 0:17:06.400 --> 0:17:08.840 Is it more the power concentration of power? I mean, 0:17:08.880 --> 0:17:13.040 I don't like concentration risk, powers risk because in the 0:17:13.040 --> 0:17:16.400 individual any stage we go bonkers or those people random. 0:17:16.480 --> 0:17:17.760 Is that the thing that worries the most? 0:17:18.520 --> 0:17:21.239 My worries go beyond just the concentration of power. Mean, 0:17:21.240 --> 0:17:24.160 that's certainly something to be concerned about. But we've dealt 0:17:24.160 --> 0:17:26.159 with concentration and power in the past. We did in 0:17:26.160 --> 0:17:29.919 the Dust Revolution. We had the robber barons. You know, 0:17:29.960 --> 0:17:32.800 we had to we broke up big oil, big tobacco. 0:17:34.119 --> 0:17:37.080 You know, we we we've we've thought about these issues 0:17:37.080 --> 0:17:39.480 and we do have mechanism of antitrust and so want 0:17:39.840 --> 0:17:42.760 to deal with concentration of powers. But what I worry 0:17:42.800 --> 0:17:46.040 about is the more insidious ways that it's going to 0:17:46.080 --> 0:17:50.440 change our society, whether it undermines our value as humans 0:17:50.680 --> 0:17:55.160 by taking away our work, by by making us feel 0:17:55.240 --> 0:17:59.560 redundant in a society, Whether it undermines our democracy by 0:17:59.640 --> 0:18:03.160 changing the conversations we have our trust in politicians, by 0:18:03.240 --> 0:18:07.800 changing the news that we see, the hearth truths, the mistruths, 0:18:07.840 --> 0:18:11.840 the conspiracy theories that AI generated deep fakes. What sort 0:18:11.840 --> 0:18:16.639 of world it will be when it's filtered through artificial intelligence? 0:18:16.760 --> 0:18:18.320 Do you do feel as though right now we're going 0:18:18.359 --> 0:18:21.199 through a massive transformational period. 0:18:21.480 --> 0:18:23.480 Yes, I do think this is we will look back 0:18:24.000 --> 0:18:26.280 and say, just as we went through a big transformation 0:18:27.160 --> 0:18:29.520 when we went through the Usher Revolution, most of us 0:18:29.600 --> 0:18:31.880 used to work in the farms, out in the fields, 0:18:31.880 --> 0:18:34.160 and then we moved to work in factories and offices. 0:18:34.480 --> 0:18:38.520 That was a big change. When we had the computer 0:18:38.560 --> 0:18:42.359 revolution and we started introducing computers into offices and businesses, 0:18:42.440 --> 0:18:44.840 that was again a big change. And the way that 0:18:44.880 --> 0:18:46.800 we went about our lives. And you know, you look 0:18:46.840 --> 0:18:48.479 at an office, so that looks quite different to it 0:18:48.520 --> 0:18:50.800 did fifty years ago. And I think we're going through 0:18:50.840 --> 0:18:54.120 another big change. And I think what's challenging is that 0:18:54.320 --> 0:18:57.200 those previous changes took times to happen. The dust revolution 0:18:57.240 --> 0:19:00.240 took fifty years to happen, you know, getting people onto 0:19:00.280 --> 0:19:02.200 the internet and doing stuff on the internet. That took 0:19:02.240 --> 0:19:05.560 a decade to happen. This one is happening overnight. 0:19:06.080 --> 0:19:08.880 Is because is that because all of us got devices. 0:19:08.920 --> 0:19:09.840 Now we've all got. 0:19:09.680 --> 0:19:12.720 Devices, we can deliver something. It's not a coincidence that 0:19:12.920 --> 0:19:15.679 chat GPT was the fastest growing app ever, you just 0:19:15.680 --> 0:19:18.040 have to be told, well, it exists, you can put 0:19:18.040 --> 0:19:21.120 the I mean, that's the fantastic opportunity for in terms 0:19:21.119 --> 0:19:23.800 of business, in terms of transformation, is you can get 0:19:24.320 --> 0:19:26.800 stuff into the hands of a billion people overnight. We've 0:19:26.840 --> 0:19:31.119 never previously had a been a method to deliver some 0:19:31.520 --> 0:19:36.080 new experience and new functionality to people overnight by just saying, well, 0:19:36.080 --> 0:19:38.879 there's this URL, just go and try it out. And 0:19:38.920 --> 0:19:41.040 then we've got all the resources, all the data centers 0:19:41.040 --> 0:19:43.320 to spin this up at scale. Right, So it used 0:19:43.320 --> 0:19:45.040 to be your hat to go out and buy yourself 0:19:45.040 --> 0:19:48.040 computers and now you just have to, you know, tell 0:19:48.040 --> 0:19:50.680 one of the providers, you know, we want some water cycles. 0:19:51.040 --> 0:19:54.600 I don't know whether it's exciting or scary or more 0:19:54.640 --> 0:19:57.480 exciting less what it is both, but am I more 0:19:57.520 --> 0:20:00.320 excited than I am scared or am I more scared 0:20:00.320 --> 0:20:02.560 that I am excited? And it's sort of I oscillate 0:20:02.600 --> 0:20:06.680 between the two depends was telling the story, but you 0:20:06.840 --> 0:20:10.080 and you're right. I mean the computer period, the Internet period, 0:20:10.200 --> 0:20:13.119 like from ninety ninety five, so of two thousand and 0:20:13.119 --> 0:20:15.480 five whatever, that a ten year period that seemed to 0:20:15.480 --> 0:20:19.320 me to be incredibly fast. You know, Moore's laws sort 0:20:19.320 --> 0:20:22.200 of doesn't exist anymore, and if they should have something, 0:20:22.200 --> 0:20:24.440 they might just have a new law, Toby's law. Everything 0:20:24.480 --> 0:20:27.160 happens at a much faster pace than has ever happened before. 0:20:27.200 --> 0:20:29.320 And it's used to for two years or something. Everything 0:20:29.320 --> 0:20:31.520 doubles in every eighty months or something along those lines, 0:20:32.400 --> 0:20:34.680 and now it seems to be happening on a monthly basis. 0:20:35.040 --> 0:20:38.400 It's like doubling, tripling on a monthly basis. But if 0:20:38.400 --> 0:20:43.400 I look back, I can go back to fifty years ago, 0:20:43.440 --> 0:20:46.120 when I first started working at twenty, I can go back, 0:20:46.160 --> 0:20:48.880 and I can I know what existed in my office environment. 0:20:48.880 --> 0:20:51.240 There was no computers. You did you typed everything yourself 0:20:51.680 --> 0:20:53.719 or the end or there's a typing pill, depending on 0:20:53.960 --> 0:20:55.919 what business you worked in. And there was lots of 0:20:55.920 --> 0:20:57.520 type of sitting there that you walk and there was 0:20:57.560 --> 0:20:59.400 actually my I worked a law from me that there 0:20:59.480 --> 0:21:03.000 was a room might be maybe twenty five girls. It 0:21:03.000 --> 0:21:06.400 was all women with that type rise of ribbons on them, 0:21:06.520 --> 0:21:10.520 and everything was typed in two in a copy and 0:21:10.880 --> 0:21:13.679 the original duplicate, and then it was we used to 0:21:13.680 --> 0:21:16.480 call it a memo, the original carbon coffee, carbon copy, 0:21:17.280 --> 0:21:20.880 original carbon cop exactly, the original CC. It was an 0:21:20.880 --> 0:21:23.920 actual carbon copy. And and then you would go and 0:21:23.960 --> 0:21:26.600 photocopy that. Some people had the ability to photocopy things, 0:21:26.720 --> 0:21:29.320 and there was just a they were called type is pool, 0:21:29.760 --> 0:21:31.960 and you know, you would write it out by hand 0:21:32.000 --> 0:21:33.560 and you'd hand it over to somebody and they would 0:21:33.560 --> 0:21:34.960 type it out for it, and that was it became 0:21:34.960 --> 0:21:37.639 a memo, a memorandum, and then you would deliver the 0:21:37.640 --> 0:21:40.360 man brand and physically it was never done by email, obviously, 0:21:40.840 --> 0:21:44.800 and then that became your record. And then obviously no phones, 0:21:44.840 --> 0:21:47.320 peop would take days to get back to you. They'd 0:21:47.320 --> 0:21:49.720 have to go write it out of response, and things 0:21:49.760 --> 0:21:53.239 just took a long time to resolve or finalize. And 0:21:53.280 --> 0:21:56.440 then but it's funny, people got very nervous when computers 0:21:56.440 --> 0:21:59.080 started coming in. They thought all these women lose their 0:21:59.119 --> 0:22:02.720 typing jobs. Then I've never gone and examined it. But 0:22:03.480 --> 0:22:06.959 those people got jobs somewhere doing something. They're not typeist anymore. 0:22:06.960 --> 0:22:09.639 But you're right, we don't employ many, if any people 0:22:09.640 --> 0:22:12.320 as type But there were lots of new jobs, invented 0:22:12.400 --> 0:22:17.320 office jobs. Yes, a more interesting jobs, perhaps, I hope. 0:22:17.160 --> 0:22:19.359 Because I mean that would have been pretty boring sitting. 0:22:19.119 --> 0:22:21.760 There just just copying what I ran. 0:22:22.480 --> 0:22:25.239 Yeah, great, well it actually was shorthand so that you're right, 0:22:25.440 --> 0:22:28.080 and missus step because sometimes if you were high enough enough, 0:22:28.240 --> 0:22:30.000 high up enough in the firm, which I wasn't at 0:22:30.000 --> 0:22:33.639 that time, you would dictate it to your what they 0:22:33.720 --> 0:22:36.400 called secretary in those days, not an EA, and then 0:22:36.400 --> 0:22:38.280 she would take copy d take you down by short 0:22:38.320 --> 0:22:39.760 and then she would hand it over to the topping 0:22:39.800 --> 0:22:42.480 pill and we get typed up. In our case a 0:22:42.520 --> 0:22:45.240 law fum get typed up as a document. Then you 0:22:45.240 --> 0:22:48.160 would review it. And at the speed at which things 0:22:48.160 --> 0:22:53.720 happened was remarkably slow. But it also when you went 0:22:53.720 --> 0:22:55.720 on a holiday, you had a good holiday because no 0:22:55.720 --> 0:22:59.320 one could contact you. Do you do you think that 0:23:00.080 --> 0:23:05.760 AI today is putting pressure on us to live a 0:23:05.840 --> 0:23:07.360 much more fast pace life. 0:23:07.160 --> 0:23:09.320 Indeed, and this is why we've now had to You know, 0:23:09.440 --> 0:23:12.440 some countries are introducing laws about you have a ride 0:23:12.480 --> 0:23:14.360 to switch off, you don't have to write, you don't 0:23:14.400 --> 0:23:15.520 have to answer emails. 0:23:15.320 --> 0:23:17.439 After five o'clock or something, but we do it here 0:23:17.480 --> 0:23:19.360 in Australia. That's I think it is a law here 0:23:19.359 --> 0:23:21.960 in our Australia. I think the Albania the entry. 0:23:22.000 --> 0:23:24.760 I mean, it's sad but true that we've actually now 0:23:24.760 --> 0:23:27.320 had to make that law to try and reduce the 0:23:27.320 --> 0:23:29.960 pressure on people, because otherwise you can always be on 0:23:30.240 --> 0:23:31.639 what's the excuse not to reply? 0:23:31.800 --> 0:23:34.239 Well, it's funny because like well before that all come 0:23:34.280 --> 0:23:36.760 in I would type something on my phone. I don't 0:23:36.760 --> 0:23:39.240 even use the laptop or an an iPad here because 0:23:39.240 --> 0:23:40.760 the guys give it to me that I don't. I 0:23:40.800 --> 0:23:43.320 don't use it and I do everything on everything, everything. 0:23:43.960 --> 0:23:47.920 And if I get idea, yeah, we all do it. 0:23:48.160 --> 0:23:49.800 But I put it down there because I just think 0:23:49.800 --> 0:23:51.119 I've got an idea. I got something, I've got to 0:23:51.160 --> 0:23:53.560 send it out to be done. It doesn't get done. 0:23:53.600 --> 0:23:56.080 I mean I said I couldn't send it midnight and 0:23:56.200 --> 0:23:58.040 hopefully someone will pick it up the next day to 0:23:58.200 --> 0:24:01.280 get it to get it sorted. Do you think that 0:24:01.880 --> 0:24:05.159 the big advantage that we're all seeing is that we are, 0:24:05.560 --> 0:24:07.800 as a result of AI and as a result of 0:24:08.320 --> 0:24:11.000 just technology generally, that we are much more efficient or 0:24:11.000 --> 0:24:11.800 are we too efficient? 0:24:13.400 --> 0:24:16.120 Well, it has the potential to make us more efficient, 0:24:16.160 --> 0:24:19.320 but equally, there's a lot of places where it's slow 0:24:19.400 --> 0:24:23.639 thinking that we need. You know, me as a CEO, 0:24:24.480 --> 0:24:27.560 you know, I remember the CEOs that I've respected, the 0:24:27.600 --> 0:24:31.080 people who make Actually a CEO makes very few decisions, 0:24:31.480 --> 0:24:34.680 but they're critical decisions, right, you want that person the 0:24:34.840 --> 0:24:37.520 strategic Yes, critical strategic decisions. 0:24:37.560 --> 0:24:38.760 Hardly, they're not tactical. 0:24:39.200 --> 0:24:41.280 It's not about you don't want the CEO to be 0:24:41.320 --> 0:24:44.560 sitting there micro managing the business. You want the CEO 0:24:44.600 --> 0:24:47.320 to be standing back and saying, Okay, this is the 0:24:47.320 --> 0:24:48.919 way that we're going to go next year, these are 0:24:48.920 --> 0:24:50.880 the products we're going to introduce. This is the way 0:24:50.920 --> 0:24:53.520 that we're going to you know, offer a better service 0:24:53.520 --> 0:24:56.159 to our customers. Those are a few critical decisions. You 0:24:56.160 --> 0:24:58.960 don't actually want them to be lost in the weeds. 0:24:59.119 --> 0:25:01.399 But do you think AI assists the CEO to do that? 0:25:01.480 --> 0:25:06.760 Now? Well, it certainly has the potential and certainly in 0:25:06.840 --> 0:25:09.080 terms in terms of, you know, giving your access to 0:25:09.280 --> 0:25:12.400 much better information about what's going on in the business 0:25:12.440 --> 0:25:14.000 and what the customers want, and so what. 0:25:14.400 --> 0:25:17.720 Can you explain how what people mean by when they 0:25:17.760 --> 0:25:25.360 say BT four, for example, hallucinates in other words, makes 0:25:25.440 --> 0:25:29.240 errors or tricks itself into the wrong answer. 0:25:29.680 --> 0:25:35.160 Yeah, it's it's wrong to think that it's grounded in truth. 0:25:35.920 --> 0:25:40.000 It's so there, So we should not we should just like. 0:25:39.960 --> 0:25:44.240 You shouldn't trust it for your for your cough. Yeah, yeah, yeah, 0:25:44.280 --> 0:25:46.680 because it may actually tell you something that might be 0:25:47.320 --> 0:25:50.359 either not just not helpful to your cough, but actually 0:25:50.440 --> 0:25:52.359 harmful to your what what what? 0:25:52.600 --> 0:25:57.240 So why would I I in in any whatever time 0:25:57.359 --> 0:26:01.240 II you're using whichever protocol, why would it give you 0:26:01.280 --> 0:26:02.600 the wrong answer? And how could it? 0:26:02.640 --> 0:26:05.320 Is it? Because it's remarkable It gives you the correct 0:26:05.359 --> 0:26:08.159 answers most of the time because it doesn't know what 0:26:08.200 --> 0:26:09.560 It's not like you and I. It's not part of 0:26:09.560 --> 0:26:11.359 the world. It doesn't live in the world and know 0:26:11.440 --> 0:26:14.680 what's true and force. It is trained on the internet, 0:26:15.600 --> 0:26:17.720 and it repeats back the sorts of things that it 0:26:17.840 --> 0:26:21.320 sees frequently on the internet. Now frequently what you will 0:26:21.320 --> 0:26:24.040 see on the internet is correct medical advice about how 0:26:24.040 --> 0:26:27.919 to treat a cough. But equally you also find stuff 0:26:27.920 --> 0:26:31.720 that's that's force and stuff that's actually harmful. And so 0:26:31.760 --> 0:26:33.520 sometimes it will repeat some of the stuff. It just 0:26:33.560 --> 0:26:37.159 says what it sees. Most frequently. The best description is 0:26:37.160 --> 0:26:40.080 that on your smartphone, when you're, you know, typing a 0:26:40.080 --> 0:26:42.720 text message, there's the auto complete that finishes the word. 0:26:43.480 --> 0:26:46.920 You know, you type in AP, and the most common 0:26:46.920 --> 0:26:50.119 way of finishing AP is a P P L E. Apple. 0:26:51.040 --> 0:26:53.040 So it will offer Apple as the way to finish 0:26:53.040 --> 0:26:55.720 the world. Well, what chatchup you died? It has done 0:26:55.800 --> 0:26:57.760 is that, but on a much greater scale. It's done 0:26:57.760 --> 0:26:59.960 it not on the not on the scale of center 0:27:00.000 --> 0:27:01.720 this is, but the whole internet. So it tells you 0:27:01.880 --> 0:27:04.199 how to just to finish the word, but the sentence 0:27:04.320 --> 0:27:07.600 or the paragraph or the page trained on the sorts 0:27:07.640 --> 0:27:09.520 of things that sees on it's seen on the internet. 0:27:09.600 --> 0:27:13.639 That's interesting because I've experienced it. So sometimes I find 0:27:13.680 --> 0:27:18.720 that the when I'm typing into but I'll keep saying, 0:27:18.760 --> 0:27:20.879 I'll just keep using copil by way of example, But 0:27:20.920 --> 0:27:23.600 if I'm typing in the actually it corrects, not correct me. 0:27:23.800 --> 0:27:28.480 It finishes off my sentence or my question much faster 0:27:29.040 --> 0:27:33.840 than say, if I'm going into if I'm using auto 0:27:33.880 --> 0:27:39.800 text order correct on text on this mobile text message 0:27:39.800 --> 0:27:43.920 to you for arguments sake, it actually finishes my sentences 0:27:43.920 --> 0:27:46.879 for me, and much more readily. Is because it's obviously 0:27:46.960 --> 0:27:47.919 heard the question before. 0:27:48.240 --> 0:27:52.040 Yes, and a lot of our conversation is quite formulaic. Yeah, 0:27:52.520 --> 0:27:55.200 and so it's been taught those formulas. So if you're 0:27:55.520 --> 0:27:57.360 if you're in a course answer, you're literally reading from 0:27:57.359 --> 0:28:00.280 a script. Yeah, and it's very good at doing exact that. 0:28:00.320 --> 0:28:02.600 But it turns out that a lot of human conversation 0:28:03.400 --> 0:28:05.879 is also quite formulating. We actually go through it. We 0:28:05.920 --> 0:28:10.520 do do that, and we've now taught those formulas to computers. 0:28:11.280 --> 0:28:14.000 So you know, how often do you stop and think 0:28:14.119 --> 0:28:16.520 carefully before you start speaking. You don't know how you're 0:28:16.520 --> 0:28:17.080 going to say the. 0:28:17.080 --> 0:28:19.880 Center, No, I'm just I'm just feeling. 0:28:19.800 --> 0:28:21.679 A lot of the time, we're actually on sort of 0:28:21.720 --> 0:28:24.639 auto pilot. We're actually going through you know, the pleasantries 0:28:24.680 --> 0:28:27.160 of you know, asking how the person is and why, 0:28:27.160 --> 0:28:29.840 how their days being. That's feeling very much a script, 0:28:29.880 --> 0:28:31.440 and we computers can do those scripts. 0:28:31.520 --> 0:28:33.520 Do you think as ai yet worked out how to 0:28:34.359 --> 0:28:38.200 cajole you, in other words, speak to you nicely. 0:28:38.480 --> 0:28:44.520 Yes, so so CHATCHPT was explicitly trained to be somewhat sycophantic, 0:28:44.560 --> 0:28:47.400 to be somewhat flattering. Do you do it agree with you? 0:28:47.440 --> 0:28:50.160 Which is also how it ends up hallucinating, which is 0:28:50.200 --> 0:28:52.200 that it always wants to be helpful and it never 0:28:52.240 --> 0:28:53.680 wants to say you know, no, I don't know the 0:28:53.720 --> 0:28:56.719 answer to that. So it always give you an answer 0:28:57.320 --> 0:28:59.640 even when you know, or give you an answer when 0:28:59.640 --> 0:29:02.680 it's really certainly you know, when everyone on the internet 0:29:02.720 --> 0:29:04.760 says the way to cure a cough is to go 0:29:04.800 --> 0:29:06.840 and get some cough medicine and some cough suites, right, 0:29:06.880 --> 0:29:09.640 so I'll tell you Mark and get some cough medicines 0:29:09.640 --> 0:29:12.120 and co suitts. Because ninety five percent of the people 0:29:12.160 --> 0:29:13.880 talking about how to deal with the cough say that. 0:29:14.480 --> 0:29:17.320 But there's other questions where it's much more fifty to fifty. 0:29:18.240 --> 0:29:20.360 But it's always been told with confidence to tell you 0:29:20.400 --> 0:29:22.160 the answer and not to tell you, oh, I'm a 0:29:22.160 --> 0:29:26.120 bit unconfident about this. Well, you know, I've seen both 0:29:26.160 --> 0:29:28.640 possibilities here, So what about I. 0:29:28.840 --> 0:29:32.640 I Well, our fisial intelligence putting into robots. 0:29:33.800 --> 0:29:36.320 Well, if we want, you know, robots to do interesting 0:29:36.400 --> 0:29:38.560 things and to react and behave in the world, the 0:29:38.640 --> 0:29:42.960 uncertain world, we've got to give them some intelligence. And indeed, 0:29:43.160 --> 0:29:45.280 you know, there's a body of thought that says we're 0:29:45.280 --> 0:29:49.560 not going to get really intelligence, artificial intelligence until we 0:29:49.880 --> 0:29:53.000 have it out in the world experiencing the rich complexity 0:29:53.040 --> 0:29:55.320 of the world. So the way that we're going to 0:29:55.360 --> 0:29:59.120 get real artificial intelligence, really intelligent is by putting it 0:29:59.120 --> 0:30:02.520 into robots, getting the robots to do stuff in the world. 0:30:02.880 --> 0:30:04.440 Say, for example, we build a robot. 0:30:05.280 --> 0:30:06.880 You know, a self driving car is in some sense 0:30:06.920 --> 0:30:09.640 of robots. Yeah, it can sense the world. It can see, 0:30:09.760 --> 0:30:12.280 you know, where the pedestrians are and the other cars, 0:30:12.520 --> 0:30:14.760 makes decisions about where to go, and then it acts 0:30:14.760 --> 0:30:17.280 on those decisions in a physical way. It turns, the 0:30:17.320 --> 0:30:19.920 wheel of the car, pushes the accelerator, goes on the break. 0:30:20.040 --> 0:30:22.400 So it's got all the characteristics of a robot. All 0:30:22.520 --> 0:30:25.680 it doesn't look like a Hollywood robot, you know, it 0:30:25.800 --> 0:30:26.880 is a robot. 0:30:27.680 --> 0:30:30.040 So, like, let's just pick an example. Let's say we're 0:30:30.080 --> 0:30:32.720 building a robot to put it into a nursing home. Yes, 0:30:32.960 --> 0:30:34.760 there's a lot of older people who need to take 0:30:34.760 --> 0:30:38.120 their tablets at a certain time, and maybe it knows 0:30:38.120 --> 0:30:41.120 to sing a song to the audience who is in 0:30:41.200 --> 0:30:43.880 the nursing home because maybe they've got Alzheimer's or something 0:30:43.920 --> 0:30:49.680 that it might be just be nice that robot is 0:30:49.720 --> 0:30:53.880 also learning about these people. Do you see dangers that 0:30:53.960 --> 0:30:56.720 it might stuff up the medication? It might give you 0:30:56.800 --> 0:31:01.040 my medication? Me might you know me your medicaid? Are 0:31:01.040 --> 0:31:04.280 the dangers that there are dangers? But I'm actually much 0:31:04.320 --> 0:31:07.000 more concerned about humans stuffing up. If you look at 0:31:07.200 --> 0:31:10.040 the number of people who get misprescribed in hospitals and 0:31:10.480 --> 0:31:13.640 in pharmacies, it's you. Yeah, people you know, get ten 0:31:13.680 --> 0:31:16.320 times the dose by mistake because they just read, you know, 0:31:16.440 --> 0:31:18.840 a thousand times of dose because the risk means grams 0:31:18.840 --> 0:31:22.560 for micrograms and so on. So humans are terrible making 0:31:22.600 --> 0:31:25.960 that and indeed getting AI to dispense is potentially going 0:31:25.960 --> 0:31:29.440 to reduce those errors. But equally, you know, I don't 0:31:29.440 --> 0:31:31.240 want to be in an old person's home with robots. 0:31:31.280 --> 0:31:33.080 I want to be an old person at home with 0:31:33.160 --> 0:31:36.400 attractive nurses look after me. I can float what a 0:31:36.400 --> 0:31:37.880 better an attractive robot nurse? 0:31:37.880 --> 0:31:39.600 But I do want the nurse, you know, I do 0:31:39.680 --> 0:31:43.160 want the nurse to have maybe a robot so that 0:31:43.480 --> 0:31:46.200 you know, lifting a person out of the bath or 0:31:46.240 --> 0:31:49.280 you know, onto the bed instead of them putting their 0:31:49.320 --> 0:31:51.920 back out doing that. The robot can do those sorts 0:31:51.960 --> 0:31:55.000 of things. But I want to have the empathy, the 0:31:55.040 --> 0:32:00.160 companionship of a person who's going to sit down and say, 0:32:00.240 --> 0:32:01.920 you're looking a bit glum today, so they have a 0:32:01.920 --> 0:32:04.160 cup of tea with me. Let's tell me what's troubling you. 0:32:04.240 --> 0:32:06.200 But maybe the robody in the nursing home could go 0:32:06.240 --> 0:32:10.200 along to the actual nurse of physicalness like a biological person, 0:32:10.760 --> 0:32:13.720 and say to their biologicalness, look, I think you've got 0:32:13.720 --> 0:32:15.960 to stop right now. Just Toby's the one who going 0:32:16.000 --> 0:32:18.640 to concentrate on today, because Toby's looking a little bit glumb, 0:32:18.880 --> 0:32:20.800 you know, because we saw through our sensing that his 0:32:21.000 --> 0:32:22.920 face is a bit down or whatever the case may be. 0:32:22.960 --> 0:32:27.680 How the sensing works based on like previous knowledge of 0:32:27.720 --> 0:32:31.120 your face features and can go and tell the noess 0:32:31.240 --> 0:32:33.680 That makes sense. But maybe I would just change the 0:32:34.280 --> 0:32:37.920 change the narrative. What about it in the military robots 0:32:37.920 --> 0:32:40.040 in the military environment, So in other words, don't send 0:32:40.040 --> 0:32:42.920 any soldiers out to do the fighting. Send robots out. 0:32:44.240 --> 0:32:46.960 That sounds like an attractive issier. Right, let's get humans 0:32:46.960 --> 0:32:50.280 out of the battlefit. That's what would you know, what 0:32:50.320 --> 0:32:52.280 would not be a goode or drones sort of do 0:32:52.360 --> 0:32:55.840 that now they do, and unfortunately they're turning warfare into 0:32:55.880 --> 0:33:02.000 more deadly dangerous activity. And the other challenge challenge with 0:33:02.040 --> 0:33:05.400 that that simple scenario is that there isn't a separate 0:33:05.400 --> 0:33:08.680 part of the world called the battlefield wars unfortunately, a 0:33:08.800 --> 0:33:12.320 fort in amongst towns and cities in and amongst and 0:33:12.360 --> 0:33:17.080 against civilian populations. And so yes, we could DEVELOPE and 0:33:17.120 --> 0:33:20.320 we are developing you know, AIS and robots that could 0:33:20.360 --> 0:33:21.880 go and fight for us. 0:33:22.160 --> 0:33:25.520 We are, are we well, yes, like those. 0:33:25.560 --> 0:33:28.520 Of the US Department of Defense has always been the 0:33:28.560 --> 0:33:32.360 greatest funder of AI research really, yes, ever since the 0:33:32.480 --> 0:33:34.760 nineteen fifties, because they could see. 0:33:35.520 --> 0:33:37.800 In a military sense, or it is generally in a 0:33:37.840 --> 0:33:41.120 military sense, right, Yes, all aspects of how we go 0:33:41.160 --> 0:33:44.000 about warfare. I mean, I don't know, I've got no idea, 0:33:44.000 --> 0:33:45.640 but I can imagine China would be doing it. 0:33:45.880 --> 0:33:47.880 China's doing it as well, Russia's doing it. Yes, there's 0:33:47.880 --> 0:33:49.960 a there is an ai arms race. 0:33:50.000 --> 0:33:52.720 Literally, there's an arms race at that level. 0:33:52.880 --> 0:33:58.080 Yes, and looking at what's happening in Ukraine is intensifying 0:33:58.160 --> 0:34:01.440 that because the militaries are seeing the nature of warfare is. 0:34:01.480 --> 0:34:04.480 Danger but it's also politically dangerous because because you lose 0:34:04.480 --> 0:34:09.280 your people, and that's political danger for example happening with Putin. 0:34:09.680 --> 0:34:13.160 The more young boys who die on the front, the 0:34:13.239 --> 0:34:16.560 more parents are going to be disaffected or dissatisfied, and 0:34:16.640 --> 0:34:19.359 the greater the chance or the risk of him losing 0:34:19.440 --> 0:34:22.040 his grip on the country. Yes, so if I can 0:34:22.080 --> 0:34:24.640 send out there, I'm going to send robots said there. 0:34:24.640 --> 0:34:25.799 Everyone's going to wow, that's cool. 0:34:26.600 --> 0:34:29.440 It is, but it's making warfare much more deadly. 0:34:29.239 --> 0:34:31.960 Activity or particularly for the weak aside. 0:34:33.920 --> 0:34:35.480 I mean, that's a challenge. If you look at what's 0:34:35.480 --> 0:34:39.360 happened in the Ukraine, You've got Russia nation ten times 0:34:39.360 --> 0:34:42.799 the size, ten times the economy of Ukraine, but it's 0:34:42.840 --> 0:34:46.719 being hells at bay by the Ukrainians who are embracing 0:34:46.760 --> 0:34:49.200 these technologies. In that sense, you know it's being used 0:34:49.200 --> 0:34:53.040 for a good purpose but being able to hold back 0:34:53.320 --> 0:34:56.799 the Russian oppressor. But equally you see that you know 0:34:56.880 --> 0:35:00.040 it's going to change the balance of power stability of 0:35:00.080 --> 0:35:04.640 the world completely. Turkey, Iran A becoming major drone powers. 0:35:06.239 --> 0:35:08.680 They're going to be able to compete against much bigger nations. 0:35:08.680 --> 0:35:11.040 I mean, it used to be your powers determine your 0:35:11.040 --> 0:35:14.520 military powers determined by your economic might. Yeah, could you 0:35:14.560 --> 0:35:18.360 buy yeh thirty five fighters, could you buy the aircraft 0:35:18.400 --> 0:35:19.960 carriers and so on things? And you know it was 0:35:20.000 --> 0:35:24.359 the US and Russia and now China. But now increasingly 0:35:24.760 --> 0:35:27.879 your military might is with these cheap drones, and it's 0:35:29.480 --> 0:35:33.080 small nations can get the whold hand on these equipments, 0:35:33.080 --> 0:35:37.120 and it's no longer the case that the balance of 0:35:37.160 --> 0:35:38.279 power is the same as it was. 0:35:38.360 --> 0:35:41.000 And where does AI fee into that for example? Like 0:35:41.239 --> 0:35:42.720 drones for example. 0:35:42.960 --> 0:35:45.080 Well, the problem is that the weak thing with the 0:35:45.160 --> 0:35:47.560 drone is the radio link and indeed it's you know, 0:35:47.719 --> 0:35:51.399 electronic warfare will stop most of the drones these days 0:35:51.400 --> 0:35:53.480 in the Ukraine. So most of the drones now are 0:35:53.520 --> 0:35:57.000 flown with fiber optics, which you can't be interrupted by 0:35:57.880 --> 0:36:02.040 by radio jamming or increasing me. They're autonomous, they're making 0:36:02.080 --> 0:36:03.799 their own decisions because then you've got a much more 0:36:03.840 --> 0:36:08.080 pow weapon. You can't stop it by jamming the radio. 0:36:08.200 --> 0:36:11.440 You can't stop it by killing the pilots. 0:36:11.520 --> 0:36:13.360 You know, the beausori pre trained. 0:36:13.719 --> 0:36:17.560 Yes, it's trained to identify the targets and to go 0:36:17.640 --> 0:36:20.719 and destroy those targets automatically with no human intervention. 0:36:21.040 --> 0:36:23.239 So there's not there is. I think what you're saying 0:36:23.320 --> 0:36:26.800 is I think what you're saying is that the drone 0:36:26.840 --> 0:36:30.160 is preloaded with let's call it intelligence, but preloaded with 0:36:30.520 --> 0:36:35.239 a plan, and there's nobody sort of sitting at the 0:36:35.280 --> 0:36:38.200 other end, sort of like like we normously droned people 0:36:38.239 --> 0:36:40.680 with a like a bat or a computer sort of 0:36:40.719 --> 0:36:44.719 directing the traffic. Because of the jamming processes, it's probably ineffective. 0:36:45.000 --> 0:36:47.680 So these things are preloaded. And they just said there 0:36:47.760 --> 0:36:50.879 might be a when you identify when when you see 0:36:50.920 --> 0:36:57.600 when the drone sees that building kamikaze suicide into it. 0:36:57.640 --> 0:36:59.840 Or when it sees something looks like a Russian tanka, 0:37:00.280 --> 0:37:02.600 or when it sees something that looks like a Russian 0:37:02.600 --> 0:37:03.440 soldier it. 0:37:03.400 --> 0:37:05.319 Looks like to me, then that seems to be then 0:37:05.920 --> 0:37:09.600 open for a lot of error. Yes, later, I mean 0:37:09.840 --> 0:37:12.080 because it might not. It might see the target, but 0:37:12.120 --> 0:37:14.680 there might be a family sitting there into having a 0:37:14.680 --> 0:37:15.680 cup of coffee, indeed. 0:37:15.719 --> 0:37:17.960 And that's that's one of the challenges, which is, you know, 0:37:18.000 --> 0:37:22.479 we fight war according to certain laws, international managed Toian law, 0:37:23.480 --> 0:37:25.640 and there's you know, we call people to account for 0:37:25.680 --> 0:37:28.160 their actions. You can't just go you know, you're only 0:37:28.160 --> 0:37:32.920 supposed to target military they targets. You're not supposed to 0:37:32.920 --> 0:37:36.560 target civilians. You know, the target has to be proportional 0:37:36.560 --> 0:37:38.520 to the threat that you face, and so on those 0:37:38.960 --> 0:37:43.000 those we hold humans to conventions and when they violate 0:37:43.040 --> 0:37:45.200 those those we you know, there's a cause of the 0:37:45.239 --> 0:37:47.920 Hague and various other places that we call people people 0:37:48.000 --> 0:37:51.000 to account. But we can only call hold people to account. 0:37:51.160 --> 0:37:53.480 Our legal system only holds people to account. 0:37:53.480 --> 0:37:55.000 We can't hold a drone account. 0:37:55.480 --> 0:37:56.479 What could we do with the drone? 0:37:56.640 --> 0:37:56.920 Nothing? 0:37:58.800 --> 0:38:01.040 So that's you know, that's another big question, which is, 0:38:01.080 --> 0:38:04.480 you know where mistakes happened when trosses, who are we 0:38:04.520 --> 0:38:05.440 going to hold account? 0:38:05.800 --> 0:38:13.080 Do they various platforms around the world. Are they interfacing 0:38:13.480 --> 0:38:19.399 with military robots which use AI or which will use 0:38:19.440 --> 0:38:21.560 AI in the future. I mean they got some sort 0:38:21.560 --> 0:38:25.240 of arrangement. There was a this pick open AI for example. 0:38:25.480 --> 0:38:28.200 Are they do they do deals with the West government? 0:38:28.200 --> 0:38:31.280 To actually help the Uist defense. 0:38:31.520 --> 0:38:38.520 Yes, yes, well today the CEO of Anthropic, the company 0:38:38.520 --> 0:38:42.560 that makes Claude, has been summoned to give meet Hesketh 0:38:42.800 --> 0:38:45.040 the Pentagon. 0:38:44.560 --> 0:38:47.040 Too is the Secretary of War. 0:38:47.120 --> 0:38:51.160 The Secondary of War, to talk about their relationship and 0:38:51.200 --> 0:38:57.120 about apparently the latest military targeting is now being using 0:38:57.280 --> 0:39:00.360 tools like Anthropic to help them make their decisions. And 0:39:00.400 --> 0:39:02.960 we see this also in Israel. We see that that, 0:39:03.040 --> 0:39:05.360 you know, lots of the idea of the Israeli defense 0:39:05.440 --> 0:39:10.800 forces targeting decisions in the Gaza conflict were not human decisions. 0:39:10.840 --> 0:39:14.160 There were decisions that the that the algorithm AI algorithms, 0:39:14.200 --> 0:39:17.080 algorithms that are called lavender, Where's Daddy, they get some 0:39:17.120 --> 0:39:20.680 funny names, but it's not humans are deciding. You know, 0:39:21.000 --> 0:39:24.880 we think that there are some some hesbula leadership in 0:39:25.000 --> 0:39:27.560 this tower block. We're going to send some artillery in 0:39:27.600 --> 0:39:30.520 against this tower block. That's an AI that's taking in 0:39:30.600 --> 0:39:33.759 all the intelligence information, all the information they've collected on 0:39:33.800 --> 0:39:38.120 social media over many years and calculating where they think 0:39:38.200 --> 0:39:39.640 maybe are potential targets. 0:39:39.800 --> 0:39:43.760 As a mathematician who were a methodician, we are are 0:39:43.800 --> 0:39:48.560 we talking about probability based decisions in we are and 0:39:48.880 --> 0:39:49.840 which is dangerous? 0:39:50.000 --> 0:39:54.360 Which is dangerous because never one hundred percent which and 0:39:54.400 --> 0:39:57.839 there's it's not clear that we really understand what those 0:39:57.880 --> 0:40:01.120 decisions are. They're talking, they're taking vast amount information and 0:40:01.160 --> 0:40:06.200 synthesizing it together, so it's not clear on what basis 0:40:06.200 --> 0:40:08.400 decision is being made and who's going to be accountable. 0:40:08.400 --> 0:40:10.440 And it's wrong. And certainly when you look at what 0:40:10.480 --> 0:40:14.520 was happening in Ghaza, the amounts of collateral damage, would 0:40:14.520 --> 0:40:17.880 you make whole place squenzes or hundreds of civilians that 0:40:17.920 --> 0:40:22.960 were being killed alongside, you know, supposed military targets, you 0:40:23.000 --> 0:40:26.360 were thinking, this is not proportional to the threat of response. 0:40:26.480 --> 0:40:28.279 Do you think that? Do you think that right now 0:40:28.320 --> 0:40:34.279 we're experiencing in Gaza and also probably in Ukraine? Are 0:40:34.320 --> 0:40:37.360 we and are we experiencing the worst of what I 0:40:37.560 --> 0:40:40.720 has got to offer right now? And we're learning lessons 0:40:40.719 --> 0:40:42.799 from it? And it is somebody taking it up? Is 0:40:42.840 --> 0:40:48.080 someone doing something about this? Well they're just combinants with Well. 0:40:48.400 --> 0:40:50.239 Next week I'm off to speak at the United Nations 0:40:50.320 --> 0:40:54.160 in Geneva about these very topics that you know, I 0:40:54.160 --> 0:40:57.840 don't think we're I don't think we're mindful enough of 0:40:58.040 --> 0:41:02.839 the world that we're opening. We get to choose what 0:41:02.920 --> 0:41:05.759 technologies get used where. I mean we sometimes forget that 0:41:05.840 --> 0:41:09.200 we think we have agency some things that you know, 0:41:09.280 --> 0:41:12.319 we we decide, we make conscious decisions, remember, you know, 0:41:12.840 --> 0:41:15.400 we we we made you know, I grew up in 0:41:15.440 --> 0:41:17.799 the UK and they made a conscious decision that there 0:41:17.800 --> 0:41:20.320 weren't going to be adverts on the main channel of TV, 0:41:21.000 --> 0:41:22.680 which is like a fantastic decision. 0:41:24.040 --> 0:41:24.800 It's a BBC. 0:41:24.920 --> 0:41:27.240 The BBC doesn't have adverts still today. 0:41:27.040 --> 0:41:28.879 The same as ABC here doesn't have. 0:41:28.800 --> 0:41:31.600 Adverts Like that was a fantastic decision. And you see 0:41:31.600 --> 0:41:33.760 what happens when you put adverts on TV, which becomes 0:41:33.760 --> 0:41:37.080 you know, driven by the need to satisfy the advertisers. 0:41:37.080 --> 0:41:40.520 And as you know, the ABC the BBC are fantastic 0:41:41.719 --> 0:41:45.960 institutions for public good, for informing the public without the 0:41:46.560 --> 0:41:50.280 pressures and the biases that having to please advertisers brings. 0:41:50.680 --> 0:41:52.680 That was a conscious decision. You know, you can put 0:41:52.719 --> 0:41:56.120 adverts on TV in many countries, that's the only thing 0:41:56.160 --> 0:42:00.600 you get. Similarly with you know, AI, we can choose 0:42:01.000 --> 0:42:03.239 how it's going to be used and where it's going 0:42:03.239 --> 0:42:06.160 to be used. I mean we we forget there. 0:42:06.000 --> 0:42:09.840 Are there are we they're responsible people were the planet. 0:42:10.040 --> 0:42:12.640 But do you think, but what if someone in China 0:42:12.680 --> 0:42:14.600 doesn't agree and they just go and do something completely 0:42:14.600 --> 0:42:17.640 different to what you where you decided, for example, they're 0:42:17.680 --> 0:42:18.280 not nations. 0:42:18.400 --> 0:42:21.080 Yeah, I mean these are accessible technologies, right, so that 0:42:21.400 --> 0:42:24.759 there's nothing that can stop people who really seriously wanted 0:42:24.800 --> 0:42:27.239 to do that. But there's a whole raft of technologies, 0:42:27.360 --> 0:42:31.439 you know, chemical weapons, biological weapons, cluster dimmunitions, blinding lasers. 0:42:31.480 --> 0:42:34.120 There's actually quite a bit of stuff that we've we've 0:42:34.120 --> 0:42:40.640 actually got UN conventions on and we don't use even 0:42:40.640 --> 0:42:42.759 though it would. You know, we can't stop people using 0:42:42.800 --> 0:42:47.480 chemical weapons. Yes, said for example, Star uses them, Fusions 0:42:47.480 --> 0:42:50.160 probably used them. You know, they occasionally get used. But 0:42:50.200 --> 0:42:53.560 when they do, the world is outraged. There's resolutions on 0:42:53.600 --> 0:42:55.560 the floor of the u N there's headlines in the 0:42:55.560 --> 0:42:56.160 New York Times. 0:42:56.160 --> 0:42:57.640 But do you think it means any. 0:42:59.000 --> 0:43:02.840 I think the world will be a much darker, dangerous 0:43:02.880 --> 0:43:07.440 place if we didn't, if we morally hadn't decided that 0:43:07.440 --> 0:43:09.319 that's not the way that we should need to fight war. 0:43:10.560 --> 0:43:13.080 What troubles me is that, you know, I think we'll 0:43:13.120 --> 0:43:16.200 get there with the darker uses of AI in a 0:43:16.280 --> 0:43:19.439 military setting. What worries me is that most of those 0:43:19.480 --> 0:43:22.319 other examples, like chemical weapons, it was only after we 0:43:22.320 --> 0:43:24.359 saw the horrors, for example, the horrors of the First 0:43:24.360 --> 0:43:27.320 World War, you know what the First World World poets 0:43:27.320 --> 0:43:31.120 and people told us about. When we saw that happening, 0:43:31.560 --> 0:43:34.960 did we have the courage to decide, actually, we should, 0:43:35.000 --> 0:43:39.319 you know, try together collectively globally, we should do something 0:43:39.320 --> 0:43:41.960 about this. And we've done something about this and limited 0:43:42.000 --> 0:43:45.200 the use to which chemical weapons we go. What I 0:43:45.320 --> 0:43:48.480 worry that we're going to have to see atrocities of 0:43:48.560 --> 0:43:50.800 AI being used on the battlefield and so on to 0:43:51.480 --> 0:43:55.160 murder civilians before we have the courage to actually take action. 0:43:56.200 --> 0:43:58.759 I say to people, it's easy to imagine what that 0:43:58.800 --> 0:44:00.759 world will be if we don't do any thing. It 0:44:00.800 --> 0:44:03.239 will look like one of those bad Hollywood movies. Will 0:44:03.280 --> 0:44:04.799 it look like what happened Hiroshima? 0:44:04.960 --> 0:44:07.360 Then they dropped the atomic bondah and you had those 0:44:07.760 --> 0:44:09.880 that are the photo that's in my mind or the 0:44:10.160 --> 0:44:13.120 images in my mind, is that little girl running down 0:44:13.160 --> 0:44:16.600 the street naked with burns or over a body like 0:44:17.480 --> 0:44:18.640 her arms outstretched. 0:44:19.480 --> 0:44:22.720 But these will be used to target women and children 0:44:22.800 --> 0:44:24.360 and civilians and innocent people. 0:44:25.280 --> 0:44:27.319 You're going to speaking of the United Nations and you 0:44:27.320 --> 0:44:29.839 said next week, yes, So what are you going to tell. 0:44:29.760 --> 0:44:32.200 Them, I mean, say, we can have to make these choices, 0:44:33.000 --> 0:44:37.000 that it's wrong to think that, you know, the US 0:44:37.000 --> 0:44:38.719 has been pushing back against this. The China has been 0:44:38.719 --> 0:44:43.080 pushing against this because because I think they wrongly feel 0:44:43.160 --> 0:44:46.359 that they've they've got some strategic advantage. They've got, you know, 0:44:46.360 --> 0:44:49.160 a technical advantage over the rest of the world. But 0:44:49.239 --> 0:44:52.040 to think that they're going to keep that is misguided. 0:44:52.680 --> 0:44:56.040 Do you think they're just pushing underground then middle of 0:44:56.120 --> 0:44:59.239 the Manhattan Project and other projects. I mean, I had 0:44:59.280 --> 0:45:02.480 I had someone's in the chair recently talking about how 0:45:03.800 --> 0:45:07.880 this is the theory, how most of the major countries 0:45:07.920 --> 0:45:12.200 around the world have had experiences, have actually UFO information, 0:45:12.760 --> 0:45:15.560 but they won't reveal it to anybody because they're keeping 0:45:15.560 --> 0:45:17.880 it for their own advantage. They're trying to examine the 0:45:18.080 --> 0:45:21.520 materials and everything around it as to how they can 0:45:21.520 --> 0:45:23.440 get some military advantage out of it, which is why 0:45:23.480 --> 0:45:24.640 they're not sharing it globally. 0:45:25.320 --> 0:45:27.960 You know. And it's a nice conspiracy theory, but it's 0:45:28.000 --> 0:45:30.879 a good one. But I wonder how you can keep 0:45:30.880 --> 0:45:32.439 so many people quiet for so long. 0:45:32.920 --> 0:45:34.680 Yeah, well they're out a lot of them are out 0:45:34.680 --> 0:45:36.279 there now talking about it. But do you think that 0:45:37.120 --> 0:45:39.520 and I'm not suggesting I agree with that, but do 0:45:39.560 --> 0:45:42.160 you think that all we're going to do is pushed underground? 0:45:44.160 --> 0:45:46.520 Well, these technologies are going to exist, you know, whether 0:45:46.600 --> 0:45:50.440 they exist already? Yes, And that's the fundamental challenge we 0:45:50.480 --> 0:45:54.480 have with a technology like arstic intelligence. It's completely due use. 0:45:55.440 --> 0:45:58.879 So for example, you know, take your AI drone that's 0:45:58.880 --> 0:46:01.520 going to autonom target people. It needs to be able 0:46:01.560 --> 0:46:05.440 to track people. Will you know, it needs to be 0:46:05.480 --> 0:46:08.600 able to track soldiers on the ground so it can 0:46:08.600 --> 0:46:12.680 target them. Well, the same algorithms again into our autonomous 0:46:12.680 --> 0:46:16.120 cars to tract pedestrians so it can avoid them. We 0:46:16.160 --> 0:46:19.680 want autonomous cars. A thousand people will die next year 0:46:20.080 --> 0:46:24.520 in Australia in road traffic accidents, mostly caused by human error. 0:46:25.000 --> 0:46:27.600 Once we have autonomous cars, we won't have those deaths. 0:46:27.920 --> 0:46:30.480 It will go to close to zero. So that's going 0:46:30.520 --> 0:46:33.600 to be a fantastic you know, bonus to our lives. 0:46:35.080 --> 0:46:38.680 You know, if you survive birth, the most likely cause 0:46:38.719 --> 0:46:41.280 of death to you're in your forties is car accident. 0:46:42.239 --> 0:46:45.799 I mean, it's terrible, and most of us sadly have 0:46:46.040 --> 0:46:47.960 you know, in our families or in our close friends 0:46:48.000 --> 0:46:50.759 know someone's life which was touched by one of those 0:46:50.760 --> 0:46:53.239 car accidents. And we're going to look back in twenty 0:46:53.320 --> 0:46:55.720 or thirty years time and say, you know, it's funny 0:46:55.760 --> 0:46:57.160 that we don't have any of that anymore. 0:46:57.640 --> 0:46:59.120 Well, it's funny. When I was four and a half 0:46:59.200 --> 0:47:01.920 years of age, I first started school on four and 0:47:01.960 --> 0:47:06.880 a half and I got off the bus at the 0:47:06.920 --> 0:47:10.400 bus stop where I live and because the bus was 0:47:10.440 --> 0:47:12.520 higher than it wasn't only little, but are you running 0:47:12.520 --> 0:47:14.520 for it a half yeah, and he didn't see me, 0:47:14.760 --> 0:47:18.560 and he will hit me, ran over me the bus 0:47:18.560 --> 0:47:21.120 a bus yeah, oh my god. Yeah, and lucky to 0:47:21.160 --> 0:47:23.000 have you a land him in hospital for quite a 0:47:23.040 --> 0:47:26.160 long time. But point being is probably in the future 0:47:26.360 --> 0:47:29.880 buses will have senses all around them which are probably 0:47:30.320 --> 0:47:32.160 you know, even if they're not an autonomous if still 0:47:32.160 --> 0:47:34.800 have sensors around it which would automatically stop the bus 0:47:35.000 --> 0:47:37.600 from proceeding because a little sense that there is a 0:47:37.640 --> 0:47:39.759 little person or there's something a little in front of it, 0:47:39.880 --> 0:47:42.279 something in front of it. 0:47:42.080 --> 0:47:43.799 As an example of that. You know, one of the 0:47:43.840 --> 0:47:47.120 newest terrorist weapons is to go rent a truck or 0:47:47.160 --> 0:47:49.880 a bus and drive us into a crowd. Yeah, that 0:47:49.920 --> 0:47:52.240 won't be possible in five or ten years time because 0:47:52.480 --> 0:47:54.880 every truck will have cameras on the front and if 0:47:54.920 --> 0:47:57.840 there's a pedestrian in front of it, it will break automatically. 0:47:57.880 --> 0:47:58.719 You won't be able to drive it. 0:47:58.840 --> 0:48:00.960 And that's our AI, that's working. 0:48:02.360 --> 0:48:04.319 People to be able to recognize the roads and to 0:48:04.360 --> 0:48:07.600 see the pedestrians and then to make the decision to stop, 0:48:07.680 --> 0:48:10.480 so that there's a great positive terrorist won't be able 0:48:10.520 --> 0:48:12.160 to do that in the future. They can hire as 0:48:12.160 --> 0:48:14.399 many trucks as they like, they can try and drive 0:48:14.440 --> 0:48:15.800 them to crowds, but they were to stop. 0:48:16.040 --> 0:48:19.759 And also just someone like a young kid not knowing 0:48:19.760 --> 0:48:22.160 any better, they won't what might get run over? 0:48:22.560 --> 0:48:23.520 Run over, young kid. 0:48:23.760 --> 0:48:26.560 I remember for six weeks, I remember it. Remember I 0:48:26.560 --> 0:48:28.400 don't remember being I still remember being a hospital. 0:48:28.440 --> 0:48:31.520 I don't remember that many more entrepreneurs that get to 0:48:31.520 --> 0:48:33.000 grow up become success. 0:48:33.000 --> 0:48:35.919 Maybe there's going to be too many. But just on 0:48:35.960 --> 0:48:38.800 this is you're probably well aware of the big argument 0:48:38.840 --> 0:48:43.280 in Australia right now that Australia in the developed nations 0:48:43.360 --> 0:48:47.480 is probably not probably is has i should say, the 0:48:47.560 --> 0:48:53.480 lowest productivity growth of any nation in the world developed 0:48:53.520 --> 0:48:57.120 nations right now, What do you think AI is going 0:48:57.160 --> 0:49:01.160 to do to help productivity in a place like Australia, Well, if. 0:49:01.040 --> 0:49:04.600 You believe the economists, all the economists come out with numbers, 0:49:04.600 --> 0:49:07.640 which is it can lift our productivity by fifteen twenty. 0:49:07.440 --> 0:49:09.360 Percent, which is significant. 0:49:09.440 --> 0:49:11.680 Yes, and I see a few other things coming along 0:49:11.719 --> 0:49:14.960 that can offer fifteen twenty percent lifts of our productivity. 0:49:15.080 --> 0:49:15.560 That's good. 0:49:15.719 --> 0:49:16.319 Yeah, that's good. 0:49:16.360 --> 0:49:16.920 That's a good thing. 0:49:17.239 --> 0:49:17.839 That's a good thing. 0:49:18.400 --> 0:49:20.799 Should we embrace it? Because the flip side of that is, 0:49:21.840 --> 0:49:24.360 you know, no I read in or hear about this often, 0:49:24.800 --> 0:49:26.600 is that there's going to be some stage where no 0:49:26.600 --> 0:49:28.279 one actually goes to work anymore. No one, no one 0:49:28.320 --> 0:49:30.880 actually works anymore, No, no, no one, but lead the 0:49:30.880 --> 0:49:34.680 majority of people, their jobs are going to become redundant. 0:49:35.520 --> 0:49:38.080 But the same productivity, the same man A taxs paid, 0:49:38.239 --> 0:49:39.840 which means the govern's got the money and we just 0:49:40.000 --> 0:49:40.960 pay you to stay home. 0:49:41.120 --> 0:49:41.359 Yep. 0:49:42.520 --> 0:49:43.160 Well, I think. 0:49:43.400 --> 0:49:46.319 People forget this has already happened, right, This happened with 0:49:46.320 --> 0:49:50.080 the Industrial Revolution, before the people forget the weekend was 0:49:50.120 --> 0:49:52.840 an invention of the dust Revolution. Before the indust revolution, 0:49:52.880 --> 0:49:55.720 you work seminary week some went up, some went down. 0:49:55.880 --> 0:49:58.279 You were working all that time. And then because of 0:49:58.320 --> 0:50:01.560 the productivity benefits of you know, the automation machinery that 0:50:01.600 --> 0:50:05.080 we brought in in the Dust Revolution, we got we 0:50:05.160 --> 0:50:07.279 got to say, you know, we want to share some 0:50:07.320 --> 0:50:08.919 of that. We want Sunday off to go to church, 0:50:09.440 --> 0:50:12.879 then Saturday afternoon off, then all of Saturday. And then 0:50:13.040 --> 0:50:15.400 I don't know why we stopped asking why did we 0:50:15.400 --> 0:50:17.960 stop asking for more? I mean, what was There's nothing 0:50:17.960 --> 0:50:20.799 magic about two days off but every seven. There's nothing 0:50:20.840 --> 0:50:22.840 about you know, the earth going around the sun that 0:50:22.840 --> 0:50:26.320 says it's only two days off seven. It's a human construction, 0:50:27.360 --> 0:50:32.080 which was we could have carried on asking for more. 0:50:32.680 --> 0:50:34.960 And you know, interestingly, there are experiments on a four 0:50:35.040 --> 0:50:38.040 day week in New Zealand and Europe. Hundreds of thousand 0:50:38.040 --> 0:50:40.719 people are doing these experiments now and the experiments all 0:50:40.760 --> 0:50:44.040 have two conclusions. One is that people do as much 0:50:44.040 --> 0:50:45.680 work in four days as they used to do in five. 0:50:47.280 --> 0:50:49.279 You you know, stop having all the bullshit meetings. You 0:50:49.320 --> 0:50:51.480 get down to it and focus on what you need 0:50:51.520 --> 0:50:54.000 to do so you can pay them as much because 0:50:54.000 --> 0:50:56.160 they're doing just as much. Right, so business is just 0:50:56.200 --> 0:50:58.640 as productive as they were when you were working five 0:50:58.840 --> 0:51:02.120 And second, people are happy. Who would have imagined. 0:51:01.800 --> 0:51:07.239 Well that I from wondering about though the happier I 0:51:07.239 --> 0:51:09.399 guess abby people said they're going to work. They're still 0:51:09.440 --> 0:51:12.239 enjoying their colleagues Monday to Thursday, for example, find I 0:51:12.280 --> 0:51:14.560 said it's something that they have an opportunity to enjoy 0:51:14.600 --> 0:51:16.520 their family. They're still getting paid the same amount of 0:51:16.560 --> 0:51:18.720 money because the company should be making the same amount 0:51:18.719 --> 0:51:20.120 of money, so the company can afford to pay them 0:51:20.320 --> 0:51:22.160 the same amount of money. That will be people who 0:51:22.200 --> 0:51:24.040 will be seduced into paying you less, but it doesn't 0:51:24.080 --> 0:51:26.960 matter trying to pay you less. Do you see that 0:51:27.040 --> 0:51:29.399 in the is this some? Is this a consequence of AO? 0:51:29.520 --> 0:51:31.920 Do you see this coming out in the future that 0:51:32.040 --> 0:51:35.919 will we definitely will be working less days, because isn't 0:51:35.960 --> 0:51:38.240 that one way fixing up the fact that some people 0:51:38.320 --> 0:51:41.080 will have no job? Yes, so instead of saying you 0:51:41.120 --> 0:51:44.759 don't work any days, I'll work one less day and 0:51:44.840 --> 0:51:47.440 you can you will work? Yeah, yeah, whatever, some along 0:51:47.480 --> 0:51:48.759 those lines. You know what I mean, like, you know, 0:51:48.760 --> 0:51:53.839 because there might be for every one person loses their job, 0:51:53.880 --> 0:51:56.000 there might be five people keep the job. So five people, 0:51:56.680 --> 0:51:59.480 I'll do it. If six people keep the job, five 0:51:59.480 --> 0:52:01.560 people keep the job, we should say so fire people 0:52:01.600 --> 0:52:04.400 going to all say we'll work one day less. You 0:52:04.520 --> 0:52:06.080 the one person have lost your job, you can now 0:52:06.120 --> 0:52:08.640 work five days or one of those lines, you know, mathematically. 0:52:09.160 --> 0:52:12.040 So that is that something people are talking about now. 0:52:12.760 --> 0:52:14.920 I think we should be having these conversations. I mean, 0:52:14.920 --> 0:52:18.359 I think it's an example of how we went through 0:52:18.400 --> 0:52:22.160 some big structural changes through the Industrial Revolution. We introduced, 0:52:22.280 --> 0:52:26.400 you know, the weekend unions, labor laws to protect the 0:52:26.480 --> 0:52:29.360 rights of workers, to spread the benefits around, universal education 0:52:29.400 --> 0:52:32.440 so we're educated for those new jobs, Universal welfare so 0:52:32.480 --> 0:52:34.680 that when we were unemployed we weren't in the poorhouse, 0:52:35.840 --> 0:52:37.839 Universal pensions so that we could retire at the end 0:52:37.840 --> 0:52:40.279 of our working lives. We did a lot to you know, 0:52:40.320 --> 0:52:42.239 spread the benefits around and make sure it was good. 0:52:42.280 --> 0:52:43.640 So the structure sort of exists. 0:52:44.320 --> 0:52:46.360 Well, now we're going through another big change, and I 0:52:46.400 --> 0:52:50.839 think we should be thinking equally radically about this. And 0:52:50.920 --> 0:52:53.640 you know, there are deep fundamental questions here, you know, 0:52:53.840 --> 0:52:56.879 many people derive their value through what they do. 0:52:57.040 --> 0:52:58.960 They identify with that, and then we. 0:52:59.040 --> 0:53:01.000 Meet someone, you say, what do you do? What work 0:53:01.040 --> 0:53:03.120 do you do? Right? Device, Well, if we're in a 0:53:03.200 --> 0:53:06.160 world where work is less important, now, how are we 0:53:06.239 --> 0:53:08.440 going to define our meaning? How? What does our purpose? 0:53:08.840 --> 0:53:12.680 So therefore a consequence of AI and a consequence of 0:53:13.239 --> 0:53:16.040 more efficiency and better productivity as a result of AI, 0:53:16.120 --> 0:53:19.000 which is one of the things I should do. Maybe 0:53:19.040 --> 0:53:22.520 government or somebody has now gone an obligation to actually 0:53:22.560 --> 0:53:26.560 help us redefine what we do. So in other words, 0:53:26.560 --> 0:53:28.640 I've only worked three days a week. That's the time 0:53:28.680 --> 0:53:31.479 I was working five days a week. That was my life. 0:53:31.480 --> 0:53:35.160 Now I'm working less than one half of my ev seven. 0:53:35.320 --> 0:53:37.080 By the way, seven days is fiction as well, But 0:53:37.360 --> 0:53:41.080 don't worry. We just assume seven days a week is correct. 0:53:41.600 --> 0:53:43.759 But four of those seven I'm not working. So all 0:53:43.800 --> 0:53:46.080 of a sudden, I'm at home. I'm not doing anything 0:53:46.840 --> 0:53:49.080 I don't think to do. It's government's got to start 0:53:49.080 --> 0:53:53.520 to re educate us to or to encourage us to, 0:53:54.200 --> 0:53:56.879 because we don't really sitting at home doing nothing. Maybe 0:53:56.880 --> 0:53:59.919 we do, but most of we'll get jacked that pretty quick. 0:54:00.640 --> 0:54:04.200 Yes, you're right. I mean I always say to remind people, 0:54:04.200 --> 0:54:06.879 you know, the purpose of education. When we educate people, 0:54:06.880 --> 0:54:08.759 it's not just to prepare them for work. It's to 0:54:08.760 --> 0:54:12.040 prepare them to take advantage of their lives, to be 0:54:12.440 --> 0:54:15.080 good citizens, to be able to do stuff and you know, 0:54:15.600 --> 0:54:16.560 make the world a better play. 0:54:16.600 --> 0:54:18.239 Do you think we're how to enjoy ourselves though, Like 0:54:19.719 --> 0:54:21.399 if we're only going to work three days a week, 0:54:21.440 --> 0:54:24.160 because we identify with ourselves working five days a week. 0:54:24.239 --> 0:54:25.600 And as you say, this is going to happen quickly, 0:54:26.120 --> 0:54:32.560 so quicker than we think. We've got to learn relearn 0:54:32.600 --> 0:54:35.520 how to fill in those extra couple of days a week. 0:54:35.640 --> 0:54:38.520 We are, but I mean, that's the great strength of humans, right, 0:54:38.560 --> 0:54:41.239 we're so quick to there we go back, cast your 0:54:41.280 --> 0:54:45.120 mind back to COVID, and we immediately learned the value 0:54:45.280 --> 0:54:47.680 of being able to come together a bit, being able 0:54:47.719 --> 0:54:49.640 to go out and you know, have a walk in 0:54:49.680 --> 0:54:53.080 the park and so on, those things that we remember. 0:54:53.280 --> 0:54:56.279 Reminded so quickly about the value of those things, and 0:54:56.320 --> 0:54:59.040 we appreciated those things so much more. 0:54:59.360 --> 0:55:01.600 Of course, a lot of people during the COVID period 0:55:01.960 --> 0:55:07.919 became addicted to social Yeah. Do you think is there. 0:55:07.800 --> 0:55:09.960 A very seductive, very addictive. 0:55:10.160 --> 0:55:12.360 Is there a potential problem because that that is AI 0:55:12.560 --> 0:55:14.360 working You either be time why. 0:55:14.200 --> 0:55:15.840 It's going to be super charged by it to be 0:55:16.960 --> 0:55:19.040 because the content is going to be our generated content, 0:55:19.120 --> 0:55:21.480 that it's going to be personalized for you. It's the 0:55:21.640 --> 0:55:25.040 AI algorithm's course that are finding your sweet spots and things. 0:55:24.960 --> 0:55:28.680 And keep seeing you the same stuff and then testing 0:55:28.719 --> 0:55:31.279 where you want to go from there? Do you think 0:55:31.320 --> 0:55:34.920 that therefore, which is why we should. 0:55:34.640 --> 0:55:38.040 Be concerned about the power of the people in Silicon 0:55:38.120 --> 0:55:40.759 Valley who are making decisions to yeah, let's put the 0:55:40.760 --> 0:55:43.480 beauty filters back in yeah, despite the fact that they 0:55:43.600 --> 0:55:45.920 know that we're not going to be able to resist 0:55:46.000 --> 0:55:47.920 using them, that we're all going to be Oh, I 0:55:47.960 --> 0:55:50.279 can't compare myself to these other people have got the 0:55:50.320 --> 0:55:53.440 beauty filters. Are I have to use the beauty filters myself? 0:55:53.640 --> 0:55:57.160 Because for me, the way I see myself advancing in 0:55:57.200 --> 0:56:01.080 life is to learn more things. But to learn things 0:56:01.440 --> 0:56:03.160 they've got nothing to do with what I already know. 0:56:04.680 --> 0:56:08.120 But the whole of the time for that well, at 0:56:08.120 --> 0:56:10.319 the essence of AIS, it doesn't work that way. It's 0:56:10.320 --> 0:56:11.759 going to say no, I'm going to keep giving you 0:56:11.800 --> 0:56:13.440 what I know you like, especially if it's on a 0:56:13.480 --> 0:56:16.440 social environment. Do you think the conversation needs to be 0:56:16.520 --> 0:56:20.319 had as to how do I encourage Mark when he's 0:56:20.360 --> 0:56:22.200 only working three days a week he said a four days, 0:56:22.440 --> 0:56:26.040 seven days, five days a week. How do I encourage 0:56:26.080 --> 0:56:30.319 Mark this conversation to or how to I encourage Mark 0:56:30.400 --> 0:56:34.880 Zuckerberg or those people control these environments to start to 0:56:34.920 --> 0:56:38.440 feed Mark something different, to encourage him to start to 0:56:38.440 --> 0:56:42.640 think more broadly or laterally about other things. What do 0:56:42.680 --> 0:56:43.840 you think about that conversation? 0:56:44.560 --> 0:56:46.640 As I've seen the reason why we couldn't think of 0:56:46.680 --> 0:56:49.759 that algorithmically, how we couldn't actually say to the algorithms, no, 0:56:49.840 --> 0:56:51.960 actually we don't want more of the same. We want 0:56:52.480 --> 0:56:54.799 If you play my favorite song to me too many times, 0:56:54.800 --> 0:56:56.680 it's going to stop being my favorite song. I'm going 0:56:56.680 --> 0:56:59.120 to get tired of it. So play me other things. 0:56:59.320 --> 0:57:01.440 You know, I remember, you know, I used to go 0:57:01.480 --> 0:57:03.719 to the music shop and browse through the music and 0:57:04.800 --> 0:57:09.719 find things by chance serendipity. It's fantastic for finding musicians 0:57:09.719 --> 0:57:11.960 that I would never have otherwise found, just because they 0:57:11.960 --> 0:57:13.960 happen to be next to where I wanted on the shelf. 0:57:15.200 --> 0:57:19.280 So you know, maybe we howgroythmically can encourage serendipity. But 0:57:19.320 --> 0:57:23.880 the problem here is the perverse incentives that the incentives 0:57:23.880 --> 0:57:27.000 that the media companies the tech companies have to you know, 0:57:27.120 --> 0:57:29.560 keep you there in the short term to keep you 0:57:29.840 --> 0:57:34.600 scrolling infinite scrolling through what their content for the now, 0:57:35.240 --> 0:57:37.479 because you know, obviously they're sending your time, they're selling 0:57:37.520 --> 0:57:40.680 your adverts and everything in and now, which is good 0:57:40.680 --> 0:57:42.680 for their revenues in the short term, but bad for 0:57:42.760 --> 0:57:45.680 you and indeed probably bad for them actually in the 0:57:45.720 --> 0:57:48.760 longer term, which is that you know, maybe they should 0:57:48.800 --> 0:57:52.640 just be saying, you know, chatter gbt'd ever says, you know, tybe, 0:57:52.680 --> 0:57:54.720 you've been asking me questions for three hours. It's three 0:57:54.760 --> 0:57:57.480 o'clock in the morning. Time to go to bed. You know, 0:57:57.640 --> 0:58:00.720 it always ends with an open question because they're trying 0:58:00.720 --> 0:58:01.920 to sell you more token. 0:58:01.800 --> 0:58:04.040 Or three more questions. Yes, yeah, it's funny. You know. 0:58:04.880 --> 0:58:08.200 I remember in two thousand and one, excuse me. In 0:58:08.240 --> 0:58:11.240 two thousand and I got asked by the then Minister 0:58:12.200 --> 0:58:16.880 for Finance Jaie Hockey to share a task force as 0:58:16.920 --> 0:58:28.320 part of Australia's OCD obligations to build a code of 0:58:28.360 --> 0:58:34.919 Practice for consumer protection for Internet vendors. So we're twenty 0:58:34.960 --> 0:58:38.640 two thous and one. It's simple things like spam you 0:58:38.720 --> 0:58:40.560 should you you know, what's the relationship you got to 0:58:40.600 --> 0:58:43.120 have opped in opted out? Just a code of practice, 0:58:43.160 --> 0:58:46.680 not legislation, but a code of practice. Simple things like 0:58:46.760 --> 0:58:51.040 if you know, if you go to a vendors website 0:58:51.880 --> 0:58:54.280 and if you want to buy pajamas for your kid, 0:58:54.760 --> 0:58:57.600 there should be the code of practice has said that 0:58:57.680 --> 0:59:02.479 you should disclose if the pajamas a fly noble or 0:59:02.520 --> 0:59:06.080 returns by something doesn't hit. What's my returns policies? You 0:59:06.120 --> 0:59:08.280 have to have it laid out. This is in two thousand, Well, 0:59:08.320 --> 0:59:11.160 it's not that long ago. It's not very long ago. 0:59:11.640 --> 0:59:14.560 Australia was the first country, so we produced a paper 0:59:14.600 --> 0:59:17.280 in a document that was the first country to first 0:59:17.360 --> 0:59:21.360 over CD country that actually put out this code of practice, 0:59:21.920 --> 0:59:24.960 and we're all quite proud of ourselves. But of course 0:59:24.960 --> 0:59:28.720 a lot of that stuff's now embedded in legislation. Do 0:59:28.760 --> 0:59:33.200 you think that the world is now at a point 0:59:33.560 --> 0:59:37.760 where we either need a code of practice or or 0:59:38.040 --> 0:59:42.240 legislation because then it was directed at millions and millions 0:59:42.240 --> 0:59:44.920 of indors. Do you think we will have the same 0:59:44.960 --> 0:59:48.480 equivalent for the twelve or fifteen powerful people who run 0:59:48.520 --> 0:59:54.120 all these AI platforms to say, dudes, this is a 0:59:54.120 --> 0:59:57.760 code of practice, or even stronger, this is the legislation, 0:59:58.040 --> 0:59:59.920 because code practice always ends up getting code of fi 1:00:01.280 --> 1:00:04.680 into law, which is what happened. Do you think this 1:00:04.800 --> 1:00:06.280 sort of thing should be done? And is there any 1:00:06.280 --> 1:00:08.680 inquiries into this stuff going on at the moment anywhere 1:00:08.680 --> 1:00:09.120 in the world. 1:00:09.520 --> 1:00:12.320 Yeah, I mean, I think policy personal politicians are working 1:00:12.360 --> 1:00:15.040 up to that they could do something. There was certainly, 1:00:15.040 --> 1:00:17.160 I think a feeling of impotence that, you know, how 1:00:17.240 --> 1:00:20.560 can we push back against these powerful companies. They're wealthier 1:00:20.560 --> 1:00:25.120 than nations, they operate outs outside our national borders and 1:00:25.280 --> 1:00:28.200 pay tax. They're pay tax. You know, the rules don't 1:00:28.200 --> 1:00:29.120 seem to apply to them. 1:00:29.240 --> 1:00:30.320 Yeah, well they're not people. 1:00:31.000 --> 1:00:33.400 Why they're not people? So you know, we're just going 1:00:33.440 --> 1:00:35.240 to have to accept it. And I think now there's 1:00:35.240 --> 1:00:38.600 a growing realization and a huge pressure from the public 1:00:38.640 --> 1:00:41.160 as well to do something. And we saw the success 1:00:41.200 --> 1:00:45.640 of the social media you know, age delay laws demonstrating 1:00:45.680 --> 1:00:47.640 that we can we can protect young people from this, 1:00:47.720 --> 1:00:50.640 and there's being in the press other legislation around, you know, 1:00:51.200 --> 1:00:54.800 making it criminal to distribute new defied pictures of people, 1:00:54.880 --> 1:00:57.280 because that was becoming a problem in our schools. 1:00:58.480 --> 1:01:01.280 As in fake fake images of faking a friend your 1:01:01.360 --> 1:01:03.600 friend and then spread around the whole school. 1:01:05.520 --> 1:01:09.080 So yeah, I think there's we've demonstrated we can do things, 1:01:09.320 --> 1:01:11.520 and there's a real appetite now to do things. I 1:01:11.560 --> 1:01:14.600 really do feel that with the new things that we're 1:01:14.600 --> 1:01:16.040 going to be able to do with AI that werey're 1:01:16.040 --> 1:01:18.960 going to be new harms. We talked about how people 1:01:19.000 --> 1:01:21.479 are having relationships with AIS, how people are using AI 1:01:21.600 --> 1:01:26.600 the therapists, people are committing suicides, and the chat pot 1:01:26.600 --> 1:01:30.400 is encouraging that behavior. Like, I think there's we've also 1:01:30.440 --> 1:01:35.800 seen that the tech companies themselves, Oh you know, there's 1:01:35.840 --> 1:01:37.920 perverse incentives been played for them not to do the 1:01:38.000 --> 1:01:41.720 right thing, not to worry necessary so much about our safety. 1:01:41.840 --> 1:01:44.920 Do you think we might end up having an AI policeman? 1:01:45.000 --> 1:01:48.400 So you know, like I'm on one of the platforms 1:01:48.400 --> 1:01:54.680 and I'm doing what I'm doing, but somehow I allow 1:01:55.400 --> 1:01:58.840 or permit an II policeman to always enter into my 1:01:58.920 --> 1:02:02.920 world and say, Mark, you're going too far. You need 1:02:02.960 --> 1:02:06.080 to go to sleep outside of outside of, for example, 1:02:06.080 --> 1:02:11.800 whatever I'm using. So it's actually an artificial intelligence policeman 1:02:12.160 --> 1:02:17.680 police person that actually not policing but effectively monitors what 1:02:17.760 --> 1:02:20.760 I'm doing. But across the board on everything I would I. 1:02:20.760 --> 1:02:22.480 Wouldn't call it a policeman. I call it a guardian. 1:02:22.840 --> 1:02:24.000 Okay, that's a better word. 1:02:24.040 --> 1:02:27.439 Yeah, AI guardian. And I'm pretty convinced on your phone 1:02:27.440 --> 1:02:29.520 you're going to have an AI guardian that sits there 1:02:29.560 --> 1:02:33.040 watching what's going on and says, Mark, be careful. This 1:02:33.080 --> 1:02:37.080 looks like it's a scam advert lock. This company's trying 1:02:37.080 --> 1:02:39.280 to steal some data out of you here, and. 1:02:39.440 --> 1:02:41.439 Well, Mark, you might be spent too much time. Yeah, 1:02:41.520 --> 1:02:44.960 and also just tells me howmuch, says hey, like actually 1:02:45.160 --> 1:02:48.880 prompts me with a notification or a nudge, because the 1:02:48.920 --> 1:02:51.160 nudging is still very powerful. Yes, I mean it, which, 1:02:51.240 --> 1:02:52.760 by the way, is what most of the platform is 1:02:52.760 --> 1:02:55.360 doing to you all the time. But it's so subtle 1:02:56.160 --> 1:02:58.560 you don't even feel the nudge. But something is a 1:02:58.600 --> 1:02:59.760 little less subtle, and. 1:03:00.320 --> 1:03:02.760 The nudges that they're giving are noticary for your benefit. 1:03:02.880 --> 1:03:06.200 No, no, no, they're pushing to get. 1:03:06.000 --> 1:03:08.040 You to spend more time on the platform. Or to 1:03:08.040 --> 1:03:10.960 buy more stuff totally. Whereas you know, we want to 1:03:10.960 --> 1:03:13.920 be nudged in ways to help for us to say, like, okay, 1:03:14.800 --> 1:03:18.600 you know, make you a happier, more content rounded person. 1:03:18.800 --> 1:03:21.640 You know, it could be nearly an education program for 1:03:21.760 --> 1:03:23.840 you too, that could sort of saying, well, the reason 1:03:23.920 --> 1:03:26.600 you need this or the reason this might be a 1:03:26.600 --> 1:03:31.080 good idea this guardian ship for adults as well as kids, 1:03:31.680 --> 1:03:33.360 is because we want to make you a happy person. 1:03:33.400 --> 1:03:36.440 We don't want you to become sort of living in 1:03:36.440 --> 1:03:41.080 this weird world where it's not real because the definition 1:03:41.200 --> 1:03:44.600 real will become brought into question. We won't know, But 1:03:45.200 --> 1:03:47.040 I think that's look for me, I reckon, that's a 1:03:47.040 --> 1:03:50.440 great idea. I would actually I would actually allow that 1:03:50.440 --> 1:03:54.720 app to live in my world. But not not not 1:03:54.720 --> 1:03:57.280 not to wrap me out of the knuckles. But maybe 1:03:57.760 --> 1:04:00.000 if it's you know, maybe you can you can enable 1:04:00.000 --> 1:04:02.000 a lot of empowering. You say, listen. If I don't 1:04:02.040 --> 1:04:03.800 listen to you, and I continue to straighten no matter 1:04:03.840 --> 1:04:06.440 what you tell me, please advise my partner, or please 1:04:06.480 --> 1:04:10.360 advise my oldest son or daughter, or my mom or dad, 1:04:10.440 --> 1:04:12.880 or whatever the case may be, you opt into that 1:04:13.600 --> 1:04:15.680 it's not automatic you can opt in. In other words, 1:04:15.720 --> 1:04:18.720 you're still in control. Is that it? But is that 1:04:18.880 --> 1:04:21.400 being done anywhere? Is anyone doing anything like that? Sounds 1:04:21.440 --> 1:04:23.600 like a fantastic I was just saying, we were just thinking. 1:04:23.600 --> 1:04:26.480 I think we might have just invented something. Yes, just 1:04:26.480 --> 1:04:30.000 talk about your book. So I have the These books 1:04:30.000 --> 1:04:32.440 are both the same. But this book's called the Shortest 1:04:32.480 --> 1:04:34.280 History of AI but it's obviously by you. And this 1:04:34.320 --> 1:04:36.680 is the English version of this strain version. I like 1:04:36.680 --> 1:04:39.600 this one because you signed this one. What and you've 1:04:39.600 --> 1:04:42.680 done a number of books. You've done produced number of books, 1:04:42.680 --> 1:04:44.080 so as well as you know you speak at lots 1:04:44.080 --> 1:04:47.040 of events. You know it's particularly you know you're not 1:04:47.040 --> 1:04:50.120 a nations, which is pretty cool. But what is the shortest? 1:04:50.800 --> 1:04:54.320 Was never in my bucket list. I never expected, never wanted. 1:04:55.120 --> 1:04:58.400 But you know, if you know, you see the harms, 1:04:58.440 --> 1:05:00.000 you think, well, if I could do something about it 1:05:00.120 --> 1:05:03.280 to I will you know. I know what my daughter's 1:05:03.440 --> 1:05:05.560 term me when she's grown up and say, Dad, you 1:05:05.640 --> 1:05:06.840 had a chance to do something. 1:05:06.920 --> 1:05:07.600 Yeah, why didn't you? 1:05:07.680 --> 1:05:08.200 Why didn't you? 1:05:08.480 --> 1:05:13.160 She's your guardian. So this says six the six essential 1:05:13.320 --> 1:05:18.160 ideas that animated obviously that's AI. What do you mean 1:05:18.200 --> 1:05:20.240 by that? The six essential ideas. 1:05:20.160 --> 1:05:23.040 So it's a short history of AI six chapters. Each 1:05:23.160 --> 1:05:26.040 chapter is is one idea that some of the magic 1:05:26.120 --> 1:05:28.480 behind art of intelligence. So one of the chapters is 1:05:28.480 --> 1:05:31.520 about neural networks and chat gipt and how how they work. 1:05:31.600 --> 1:05:33.640 Tries to explain and explains the history, where did they 1:05:33.680 --> 1:05:35.800 come from? Where did that how did that technology come? 1:05:35.840 --> 1:05:37.480 And some of the colorful characters. 1:05:37.280 --> 1:05:39.920 Because there's a lot of let's call it jargon in 1:05:40.040 --> 1:05:45.000 AI world, neural networks, large language models, and all these 1:05:45.080 --> 1:05:47.320 sorts of things, and it's nearly like it's to me, 1:05:47.440 --> 1:05:52.680 it looks like it's designed to the average person to say, look, 1:05:54.040 --> 1:05:57.280 just accept it's too complicated for you. Just go along 1:05:57.320 --> 1:06:01.680 with it. That this nearly it is nearly a new language. 1:06:02.280 --> 1:06:05.880 It is. But let's forget we're trying to make machines 1:06:05.920 --> 1:06:09.200 that are intelligent, and you know intelligence is you know, 1:06:09.360 --> 1:06:11.680 our brain is the most complex thing in the universe. Yep, 1:06:11.920 --> 1:06:14.919 by far. Yeah, nothing approaches the complexity, so we're trying 1:06:14.920 --> 1:06:16.800 to match that. So it's not surprising that it gets 1:06:16.800 --> 1:06:20.520 a bit technical and it does seem a bit magical 1:06:20.640 --> 1:06:24.040 to us. And we're easily taken in when the machines 1:06:24.080 --> 1:06:26.760 start talking back to us, because no one else has 1:06:26.800 --> 1:06:29.760 ever only how the humans, but the only thing that 1:06:29.800 --> 1:06:32.320 ever talked back to us were other humans, other intelligent beings. 1:06:32.360 --> 1:06:36.000 So it's easy to be fooled and seduced by the 1:06:36.080 --> 1:06:38.760 machines when they started talking to us because our only 1:06:38.840 --> 1:06:41.600 experience of things that talk back to us are other humans, 1:06:42.800 --> 1:06:44.720 all of their intelligence and everything else that goes with 1:06:44.800 --> 1:06:45.360 being humans. 1:06:45.520 --> 1:06:47.800 You own a book called twenty sixty two The World 1:06:47.840 --> 1:06:49.600 that AOI made. This is looking into the future. 1:06:49.960 --> 1:06:51.760 Yes, the year twenty sixty two. 1:06:51.840 --> 1:06:54.680 Yeah, the year twenty sixty two. So why they're year? 1:06:55.000 --> 1:06:57.960 Well I gave a talk in North Sydney and this 1:06:58.280 --> 1:07:00.560 old lady turned up at the talk and she said, oh, 1:07:00.680 --> 1:07:02.720 so it's nothing to do with my post code. Then 1:07:06.040 --> 1:07:08.600 I'm very sorry. It's got you here on the forced protector. 1:07:08.600 --> 1:07:09.520 It's all about the year. 1:07:09.480 --> 1:07:11.120 Twenty sixty or why twenty sixty two? 1:07:11.560 --> 1:07:14.520 Well? I surveyed three hundred colleagues other experts around the 1:07:14.520 --> 1:07:17.280 world in AI. I asked them the simple question, when 1:07:17.320 --> 1:07:21.040 would computers match humans in their intelligence? And the average 1:07:21.080 --> 1:07:23.680 answer they said this was ten years ago. It was 1:07:23.720 --> 1:07:26.200 twenty sixty two. I think I suspect if I did 1:07:26.240 --> 1:07:29.240 the survey today, they might say twenty forty two. Oh really, yes, 1:07:29.320 --> 1:07:32.080 I think the date has moved forwards quite significantly. 1:07:32.280 --> 1:07:34.720 And what do you mean by match our intelligence? 1:07:34.800 --> 1:07:37.400 To do everything that we can do? That requires intelligence? 1:07:38.720 --> 1:07:43.000 And what do you mean by intelligence to be able 1:07:43.040 --> 1:07:48.120 to write a poem, do a tax return, answer a 1:07:48.200 --> 1:07:50.440 maths question? They can do this, plan a holiday. 1:07:50.640 --> 1:07:54.720 He can do that now though, yes, why twenty forty two? 1:07:55.200 --> 1:07:57.959 What is it that do you think that a computer? Sorry, 1:07:58.200 --> 1:08:00.720 I I will that we can do that. I I 1:08:00.880 --> 1:08:01.760 can't do anything. 1:08:01.840 --> 1:08:05.640 Well, So they don't have our creativity, they don't have 1:08:05.720 --> 1:08:08.919 our emotional intelligence, they don't have our social intelligence. There's 1:08:08.960 --> 1:08:12.640 still still a very what's called jagged intelligence. You know, 1:08:12.720 --> 1:08:15.640 there's really good at something so that they can answer 1:08:15.800 --> 1:08:18.960 international maths olympiad questions. But you know better than almost 1:08:19.000 --> 1:08:21.439 anyone on the planet. And yet you can ask them, 1:08:21.439 --> 1:08:23.439 as you know, a simple puzzle, you know, like, here's 1:08:23.439 --> 1:08:28.599 a simple one that defeats chattypt You know, I've got 1:08:28.640 --> 1:08:31.479 a two liter jug and a five liter jug. How 1:08:31.520 --> 1:08:35.680 do I measure out three liters of liquid? And you know, 1:08:35.760 --> 1:08:38.200 the answer right, you feel the five liter jug to 1:08:38.320 --> 1:08:40.479 the top, you pour it into the two liter jug, 1:08:40.520 --> 1:08:43.400 and you've got three liters left to it. But you 1:08:43.520 --> 1:08:46.559 ask CHATTYPT and it says, it gives you this twelve 1:08:46.600 --> 1:08:50.160 step plan of pouring water backwards and fords between the jugs, 1:08:51.520 --> 1:08:54.639 and eventually it says, and now in the two liter jug, 1:08:54.720 --> 1:08:58.200 you've got three liters of liquid. Batshit crazy. It's not 1:08:59.320 --> 1:09:00.960 deeply under earning the problem that you. 1:09:01.000 --> 1:09:05.960 Are is that because it's taken that answer from somewhere else, 1:09:06.080 --> 1:09:07.280 or it actually deducted and. 1:09:07.600 --> 1:09:10.040 It's been trained. There's lots of example where it puzzles 1:09:10.080 --> 1:09:13.080 on the internet where people have talked about jug questions 1:09:13.080 --> 1:09:15.519 where you pourp but it and it's sort of just 1:09:15.600 --> 1:09:19.160 giving you a blurred average of those problems where it's 1:09:19.360 --> 1:09:21.479 lots of pouring liquids backwards. And for us, it's not 1:09:21.600 --> 1:09:25.400 really sitting there thinking through logically what the steps are. 1:09:25.520 --> 1:09:29.200 So would like as someone who's like you're a mathematicism, 1:09:29.240 --> 1:09:32.519 but as someone like you think, would you say, therefore, 1:09:33.280 --> 1:09:35.879 artificial intelligence is not really making a deduction. 1:09:37.040 --> 1:09:39.160 Yes, it's not reasoning the way that we're reasing. 1:09:39.240 --> 1:09:41.120 Yeah, the way we reason I should say the way 1:09:41.200 --> 1:09:42.200 we would make it deduction. 1:09:42.520 --> 1:09:45.519 Yes, and yes, I mean we're working on how we 1:09:45.560 --> 1:09:49.120 can actually add more deductive capabilities to these language models. 1:09:49.160 --> 1:09:52.240 But they're they're you know, retrieving the answers from you know, 1:09:52.320 --> 1:09:54.559 what they were trained on more than they are reasoning 1:09:54.600 --> 1:09:55.200 about the answer. 1:09:55.280 --> 1:09:57.720 What do you expect for twenty twenty six? That's the 1:09:57.760 --> 1:09:59.960 opposite of twenty sixty two? But like, what do you 1:10:00.080 --> 1:10:02.400 what are you looking forward to in twenty twenty six 1:10:02.680 --> 1:10:04.559 this year when it comes to artificial intelligence? 1:10:04.760 --> 1:10:06.960 Well, I mean, the exciting thing happening now is is 1:10:07.000 --> 1:10:10.920 that we have these AI agents that instead of having 1:10:11.040 --> 1:10:12.880 these chatbots which sit there and wait for you to 1:10:12.920 --> 1:10:16.360 ask a question, they actually have some agencies, some ability 1:10:16.400 --> 1:10:19.080 to go and do stuff. So, as an example, when 1:10:19.120 --> 1:10:21.479 you've got a new employee, there's a bunch of stuff 1:10:21.479 --> 1:10:23.000 you've got to do. You've got to give you know, 1:10:23.080 --> 1:10:24.840 you've got to give them a log in, You've got 1:10:24.960 --> 1:10:26.800 to they've got to do the health and safety on 1:10:26.920 --> 1:10:31.280 board training. They've got to get a path for the building, 1:10:31.880 --> 1:10:33.640 they've you know, they've got to you've got to get 1:10:33.640 --> 1:10:37.519 their financial data so you can pay their pay their 1:10:37.560 --> 1:10:39.519 wages at the end of the first week, there's a 1:10:39.600 --> 1:10:42.040 bunch of you know, tasks you've got to do. Well. Now, 1:10:42.160 --> 1:10:44.400 instead of you having to go through and do all 1:10:44.479 --> 1:10:46.599 those things yourself, you can just tay to the agent, okay, 1:10:47.360 --> 1:10:49.200 can you on board this new person, and it breaks 1:10:49.240 --> 1:10:51.160 the task down into the different things, and then it 1:10:51.240 --> 1:10:53.080 takes each of those tasks and break those down into 1:10:53.080 --> 1:10:54.759 the things you have to do. To get the people's 1:10:54.800 --> 1:10:57.599 financial information, you've got to get some you know, identity documents, 1:10:57.640 --> 1:10:59.800 you've got to get their bank co ordinance. You've got 1:10:59.840 --> 1:11:01.720 to just to them in the accountancy system. You've got 1:11:01.760 --> 1:11:03.960 to set up the payroll. And then it said, you know, 1:11:04.080 --> 1:11:05.680 to register the payroll. It's got to get you know, 1:11:05.720 --> 1:11:08.360 all the information their date of birth, the tax number 1:11:09.120 --> 1:11:10.519 and so on, and it goes through all those things. 1:11:10.640 --> 1:11:11.760 It does all that stuff for you. 1:11:12.960 --> 1:11:15.640 And do you see for the twenty twenty six so 1:11:15.960 --> 1:11:18.320 I mean you just see just more efficiency games or 1:11:18.360 --> 1:11:20.920 do you see something large coming out of all it's 1:11:20.960 --> 1:11:22.840 something really disruptive. 1:11:24.160 --> 1:11:27.120 Well, the disruptive thing that you have there is suddenly 1:11:27.280 --> 1:11:32.400 that the computers are doing stuff. They're changing your files there, 1:11:32.800 --> 1:11:36.760 maybe spending money, maybe the ordering stuff and that's going 1:11:36.800 --> 1:11:37.360 to go wrong. 1:11:37.439 --> 1:11:39.879 Sometimes that will be disruptive. 1:11:40.000 --> 1:11:41.760 That's going to be disrupted, and it's like, who are 1:11:41.760 --> 1:11:43.839 we going to blame because we can't blame the computer, 1:11:44.439 --> 1:11:47.479 but what will do? Also, things will go wrong at 1:11:47.520 --> 1:11:49.760 speed because this is a computer speed, not human speed. 1:11:49.840 --> 1:11:52.519 So if it's trying to reverse, it could be hard 1:11:52.560 --> 1:11:53.040 to reverse. 1:11:53.240 --> 1:11:55.559 I mean, there was a story in the paper yesterday 1:11:55.600 --> 1:11:58.680 where the head of safety at open AI had all 1:11:59.000 --> 1:12:02.840 email was deleted by one of these agents. She's the 1:12:02.920 --> 1:12:06.040 head of safety, like a rookie mistake. She gave it 1:12:06.120 --> 1:12:09.559 the power to clean up her mailbox. You know, I'd 1:12:09.640 --> 1:12:12.120 love to have someone who could go through my mailbox 1:12:12.160 --> 1:12:14.720 and clean out all the junk, but also include out 1:12:14.760 --> 1:12:16.120 all the stuff that she wanted to keep. 1:12:16.320 --> 1:12:20.320 Because the AI agent had agency and made the decisions. 1:12:19.960 --> 1:12:21.080 Yes, made the decisions. 1:12:21.280 --> 1:12:24.400 So that that means the instructions and the rules that 1:12:24.520 --> 1:12:30.320 have been given to the AAI agent were not not 1:12:30.439 --> 1:12:32.280 complete or maybe not accurately. 1:12:32.520 --> 1:12:34.080 It wasn't just that the rules were in complete, but 1:12:34.120 --> 1:12:38.040 also the AI is not perfect. You know, we talked 1:12:38.040 --> 1:12:41.600 about how they hallucinate, They make mistakes, you know, just 1:12:41.680 --> 1:12:43.839 like humans are not perfect. These are not perfect either. 1:12:44.040 --> 1:12:46.479 So we be, and I guess that's sort of part 1:12:46.479 --> 1:12:49.880 of it too. I think we should learn to expect imperfection. 1:12:50.200 --> 1:12:52.920 They are, we should, But I mean, the interesting thing 1:12:53.040 --> 1:12:56.160 is people have higher standards for machines than they'd have 1:12:56.439 --> 1:12:56.800 to have. 1:12:56.960 --> 1:12:57.559 That's weird. 1:12:57.880 --> 1:13:00.840 So you say to people, you know, we've got self 1:13:00.920 --> 1:13:04.479 driving cars now, and they're roughly the level of humans 1:13:04.680 --> 1:13:07.000 in terms of, you know, the number of mistakes they make, 1:13:07.080 --> 1:13:10.200 and some ways they're better than humans they make fewer mistakes, 1:13:10.200 --> 1:13:14.000 but people are still very reluctant to give them responsibility 1:13:14.080 --> 1:13:16.880 to drive. So you say, well, okay, well how much 1:13:16.960 --> 1:13:19.280 better than humans? If they do, they have to be 1:13:19.360 --> 1:13:23.000 ten times better than humans before we can entrust them 1:13:23.040 --> 1:13:27.360 with the responsibility of driving without human oversight. And then 1:13:27.400 --> 1:13:29.160 at some point, you know, okay, so maybe we say 1:13:29.200 --> 1:13:32.960 it's ten Well, what happens when they're a hundred times 1:13:33.120 --> 1:13:37.240 better than humans? Do we let humans still drive? Because 1:13:37.439 --> 1:13:40.360 humans are going to be killing people. Machines are provably 1:13:40.439 --> 1:13:42.080 not going to be killed. They can't get drunk. They 1:13:42.080 --> 1:13:44.320 can't get drunk, they're going to be tired, they overtired. Yeah, 1:13:44.320 --> 1:13:45.960 they're not going to be speeding, They're not going to 1:13:46.000 --> 1:13:50.160 be breaking the road rules. At what point do we say, humans, 1:13:50.280 --> 1:13:53.439 you know you're killing you know, your fellow citizens, maybe 1:13:53.439 --> 1:13:54.200 we shouldn't allow this. 1:13:54.439 --> 1:13:56.040 Well, a friend of mine the other day told me 1:13:56.120 --> 1:14:01.080 that Tom she was in San Francisco and the incinet 1:14:01.160 --> 1:14:05.360 went down, stopped and they all stopped, which is the 1:14:05.400 --> 1:14:07.200 stop stop there and then, which is the safe thing 1:14:07.240 --> 1:14:10.280 to do totally. But there were other non waymos on 1:14:10.400 --> 1:14:14.599 the road. Yes, humans people driving cars. I'm now called 1:14:14.640 --> 1:14:18.560 to non non and you're driving and and like I 1:14:18.640 --> 1:14:21.200 all of a sudden the things because literally they're just yeah, 1:14:22.040 --> 1:14:23.880 they're not used, they're not going very fast because they're 1:14:23.880 --> 1:14:27.719 all governed. But nonetheless they just stopped because something happened 1:14:27.720 --> 1:14:32.320 in obviously and yeah, and we keep forgetting that. The 1:14:33.280 --> 1:14:40.559 the Achilles heel of all artificial intelligence is that is power. Yes, 1:14:41.640 --> 1:14:43.920 I mean that's what drives everything. If you turn off 1:14:43.960 --> 1:14:49.200 the switch, not literally, if you know what I mean metaphorically, 1:14:49.200 --> 1:14:52.080 if you just turn the power off this thing that 1:14:52.120 --> 1:14:55.519 we're really worried about, or somehow you can if the 1:14:55.600 --> 1:14:59.599 thing is battery based, like it's some other way it's powered, 1:15:00.080 --> 1:15:02.880 not not by electricity with something, not by put money 1:15:02.880 --> 1:15:05.360 in a wall. If somebody you can disconnect the power 1:15:06.000 --> 1:15:07.880 thing stops, it does. 1:15:07.960 --> 1:15:10.800 But that's the problem as well. I am forced to 1:15:10.840 --> 1:15:14.040 write this fantastic science fiction story one hundred years ago 1:15:14.680 --> 1:15:19.160 titled When the Machine Stops, about how wig becoming ever 1:15:19.280 --> 1:15:22.439 more dependent upon the machines and the machines are becoming 1:15:22.720 --> 1:15:29.400 ever more when we're losing touch with actually how they work. 1:15:31.600 --> 1:15:33.519 And the short story is about what happens, you know, 1:15:33.600 --> 1:15:37.160 when the machines eventually stop break, no one knows how 1:15:37.160 --> 1:15:39.080 to fix them, and the world grinds to a halt. 1:15:40.120 --> 1:15:40.360 Yeah. 1:15:40.520 --> 1:15:42.120 So it's not that you know that we turn the 1:15:42.200 --> 1:15:44.960 machines off. It's problems. But when the machine stops, what happens? 1:15:45.520 --> 1:15:48.080 Could we still do anything? And we're already completely dependent. 1:15:48.240 --> 1:15:50.840 If the Internet goes down, the stock market stops, the 1:15:51.240 --> 1:15:56.840 air traffic you know, you know in the university, you know, 1:15:56.880 --> 1:15:59.400 the internet goes down, people say, well, time to go home. 1:16:00.000 --> 1:16:02.519 I think we can do yeah go home? Yeah, Well 1:16:03.040 --> 1:16:03.519 I'm wishing. 1:16:03.720 --> 1:16:05.720 And now even that is problematic because lots of people 1:16:05.760 --> 1:16:08.920 have got smart doorbells and so on, and they can't 1:16:08.960 --> 1:16:10.000 get into their homes anymore. 1:16:10.640 --> 1:16:13.559 Yeah, because it's either thing of printer or some sort 1:16:13.560 --> 1:16:17.240 of biometric you can't get out. We were in an 1:16:17.280 --> 1:16:20.800 old fashion key hidden somewhere exactly somewhere, and that's your 1:16:20.840 --> 1:16:21.920 redundancy process. 1:16:23.080 --> 1:16:24.840 My wife was saying, should we get some of those 1:16:24.840 --> 1:16:29.240 smart doorbells and things. I said, no, keys are fantastic. 1:16:29.320 --> 1:16:33.120 They work totally the redundancy and you can have and 1:16:33.240 --> 1:16:35.519 you can have a second key. If you lose that key, 1:16:35.560 --> 1:16:37.479 you can have another key hidden somewhere, which is what 1:16:37.560 --> 1:16:42.760 most people do. I really appreciate this conversation. I've had 1:16:42.760 --> 1:16:44.240 a lot of fun talking to you about it, but 1:16:44.720 --> 1:16:46.560 equally on the funds off on the other side of 1:16:46.640 --> 1:16:50.320 fund is a serious topic. Yeah, a very serious topic topic, 1:16:50.360 --> 1:16:51.960 and I wish you the best when you go to 1:16:51.960 --> 1:16:55.320 the United Nations next week and hopefully, you know, we 1:16:55.439 --> 1:16:59.560 can start to get some better road rules around or 1:17:00.160 --> 1:17:02.400 get the conversation going about what the road rules are 1:17:02.840 --> 1:17:05.639 for the future. Indeed, thank you very much. 1:17:05.760 --> 1:17:06.400 Been a pleasure by