1 00:00:00,320 --> 00:00:07,240 Speaker 1: The electric cars, cars, electric brain drains with electric frames. 2 00:00:07,520 --> 00:00:11,160 Speaker 1: Our next guest has worked for NASA's Jet Propulsion Laboratory 3 00:00:11,200 --> 00:00:13,640 Speaker 1: for over thirty years, helping advance our exploration of the 4 00:00:13,680 --> 00:00:16,560 Speaker 1: Solar system. He's taking a look at what's going on 5 00:00:16,560 --> 00:00:20,520 Speaker 1: ins a society disruption artificial intelligence. He's got a new 6 00:00:20,520 --> 00:00:23,800 Speaker 1: book out. Peter Scott is author of Crisis of Control, 7 00:00:24,160 --> 00:00:28,479 Speaker 1: How Artificial Superintelligence May destroy or Save the Human race, 8 00:00:28,520 --> 00:00:32,640 Speaker 1: and he joins us on the phone from Victoria, British Columbia. Peter, 9 00:00:32,800 --> 00:00:35,720 Speaker 1: nice to have you here with us. Uh. We spend 10 00:00:35,760 --> 00:00:38,680 Speaker 1: so much time here at Bloomberg talking a lot about 11 00:00:38,720 --> 00:00:41,880 Speaker 1: artificial intelligence. It seems to have found its way into 12 00:00:41,920 --> 00:00:46,800 Speaker 1: just about every industry. Is it friend or foe? Hikay, Well, 13 00:00:46,840 --> 00:00:49,680 Speaker 1: thanks for having me on the program here. When I 14 00:00:49,720 --> 00:00:53,360 Speaker 1: started writing my book, I had to preface every remark 15 00:00:53,520 --> 00:00:56,400 Speaker 1: with that I when I spoke with I noticed is 16 00:00:56,440 --> 00:01:00,400 Speaker 1: going to sound crazy. But events of certainly changed in 17 00:01:00,440 --> 00:01:02,480 Speaker 1: the last year where this is starting to become more 18 00:01:02,480 --> 00:01:08,800 Speaker 1: of a mainstream message and you're echoing a common refrain 19 00:01:08,920 --> 00:01:13,560 Speaker 1: in the battle currently going between Mark Zuckerberg and Elon Musk, 20 00:01:13,640 --> 00:01:17,960 Speaker 1: for instance, and in the short term, AI is definitely 21 00:01:18,120 --> 00:01:22,319 Speaker 1: a friend. In the long term, it's likely to be 22 00:01:22,400 --> 00:01:26,680 Speaker 1: a foe. And the trouble is that the dividing line 23 00:01:26,720 --> 00:01:29,320 Speaker 1: between short and long term could be as little as 24 00:01:29,319 --> 00:01:31,920 Speaker 1: twenty years, and it might take longer than that to 25 00:01:32,400 --> 00:01:35,600 Speaker 1: mitigate those effects. Well, it's you know, when we talk 26 00:01:35,640 --> 00:01:38,160 Speaker 1: about artificial intelligence, I do also feel like we throw 27 00:01:38,160 --> 00:01:43,000 Speaker 1: it around. Um, you know you're someone who understands it intrinsically, 28 00:01:43,280 --> 00:01:45,760 Speaker 1: What exactly are we talking about? I mean, we're basically 29 00:01:45,800 --> 00:01:50,360 Speaker 1: teaching computers what to think like a human. So there 30 00:01:50,440 --> 00:01:54,640 Speaker 1: is actually a very good working definition of off official intelligence, 31 00:01:54,680 --> 00:02:00,560 Speaker 1: but several competing ones, And oftentimes people say artificial intelligence 32 00:02:00,600 --> 00:02:02,800 Speaker 1: is what we haven't learned how to do yet, because 33 00:02:03,320 --> 00:02:06,200 Speaker 1: every time we do achieve something that we thought all 34 00:02:06,240 --> 00:02:09,560 Speaker 1: that takes AI to do it, then we say, well, no, 35 00:02:09,720 --> 00:02:13,080 Speaker 1: it wasn't really that difficult after all, like playing chess. 36 00:02:14,080 --> 00:02:19,760 Speaker 1: The popular conception of artificial intelligence now is the field 37 00:02:19,760 --> 00:02:24,679 Speaker 1: called artificial general intelligence, which we don't have any instantiations 38 00:02:24,760 --> 00:02:29,280 Speaker 1: of yet, where a computer would produce a good facts 39 00:02:29,280 --> 00:02:34,840 Speaker 1: simile of a human reasoning being able to solve general tasks, 40 00:02:35,639 --> 00:02:39,760 Speaker 1: and we're several major breakthroughs away from that happening, but 41 00:02:40,280 --> 00:02:43,120 Speaker 1: that could be as little as twenty years. You have 42 00:02:43,200 --> 00:02:45,960 Speaker 1: a great quote in one of your chapters, Alan Turing, 43 00:02:46,080 --> 00:02:48,600 Speaker 1: if a machine is expected to be infallible, it cannot 44 00:02:48,639 --> 00:02:54,400 Speaker 1: also be intelligent. Um. Yes, I think that we will 45 00:02:55,240 --> 00:03:00,280 Speaker 1: need to get a lot more understanding of the human 46 00:03:01,040 --> 00:03:04,240 Speaker 1: psychology and cognitive processes in order to be able to 47 00:03:04,280 --> 00:03:09,520 Speaker 1: tackle things like this, to have good and ethical artificial 48 00:03:09,639 --> 00:03:14,800 Speaker 1: general intelligence, because the human brain is wide to make 49 00:03:14,840 --> 00:03:19,440 Speaker 1: a lot of mistakes, which computers generally don't, and what 50 00:03:19,680 --> 00:03:22,560 Speaker 1: we need to somehow incorporate that as a feature before 51 00:03:22,639 --> 00:03:26,000 Speaker 1: we get robot friends. Well, you know it's interesting, you know, 52 00:03:26,040 --> 00:03:28,280 Speaker 1: your press release talks about how the book looks at 53 00:03:28,280 --> 00:03:32,000 Speaker 1: both the challenges and opportunities of artificial superintelligence. I think 54 00:03:32,040 --> 00:03:35,119 Speaker 1: about things like the medical profession, where there's so much information. 55 00:03:35,640 --> 00:03:38,800 Speaker 1: I think it's hard for anybody in the medical profession 56 00:03:39,000 --> 00:03:41,600 Speaker 1: to be up to date on everything that's happening in 57 00:03:41,680 --> 00:03:43,960 Speaker 1: terms of R and d UM. But if you were 58 00:03:44,000 --> 00:03:49,080 Speaker 1: able to have a computer kind of assimilate all that 59 00:03:49,160 --> 00:03:52,920 Speaker 1: information so that when somebody has some kind of ailment, 60 00:03:53,040 --> 00:03:56,360 Speaker 1: they're getting the latest and greatest information. It's not you know, 61 00:03:56,440 --> 00:03:58,400 Speaker 1: it's not just for those people who are in major 62 00:03:58,480 --> 00:04:03,440 Speaker 1: cities who have access to the greatest and best hospitals. Um, 63 00:04:03,480 --> 00:04:06,760 Speaker 1: you know, but that everybody has access to that information, 64 00:04:06,880 --> 00:04:10,800 Speaker 1: that's a good thing. Yes, it certainly should be earned. 65 00:04:11,560 --> 00:04:16,040 Speaker 1: For instance, Watson the computer the one jeopardy is now 66 00:04:16,240 --> 00:04:21,000 Speaker 1: solving diagnosing cancer cases. First Sloan Cattering memorial and when 67 00:04:21,000 --> 00:04:24,720 Speaker 1: they first took Watson on board, they thought, well, maybe 68 00:04:24,720 --> 00:04:27,159 Speaker 1: it'll be able to diagnose five the cases that we 69 00:04:27,279 --> 00:04:30,680 Speaker 1: can't and the number was something more like heading upwards. 70 00:04:30,680 --> 00:04:33,440 Speaker 1: And they discovered it was because Watson was up to 71 00:04:33,480 --> 00:04:36,640 Speaker 1: date on literature that they were six months behind in reading, 72 00:04:37,320 --> 00:04:40,440 Speaker 1: and it made that much of a difference. And so 73 00:04:41,240 --> 00:04:46,200 Speaker 1: absolutely we could put together um understanding of not just 74 00:04:46,560 --> 00:04:50,039 Speaker 1: cancer but everything else. So you could have a lot 75 00:04:50,080 --> 00:04:55,080 Speaker 1: of ais that were narrow intelligences, each one specializing. So 76 00:04:55,200 --> 00:04:59,280 Speaker 1: you have one that can specialize in everything, generalized specialist 77 00:04:59,320 --> 00:05:02,680 Speaker 1: in medic and and that would be a good thing. 78 00:05:03,279 --> 00:05:08,640 Speaker 1: It's not one of the the doomsday scenarios there was 79 00:05:08,680 --> 00:05:12,440 Speaker 1: arise from a bug of one form or another. What 80 00:05:12,640 --> 00:05:15,200 Speaker 1: is it that worries you might happen as a result 81 00:05:16,080 --> 00:05:23,000 Speaker 1: of artificial intelligence or super inlligence. Yes, The problem boils 82 00:05:23,040 --> 00:05:26,280 Speaker 1: down to unintended side effects and what we would call 83 00:05:26,320 --> 00:05:30,080 Speaker 1: a bug who wants to be relying on version one 84 00:05:30,120 --> 00:05:34,320 Speaker 1: point oh of anything. And with artificial intelligence, it will 85 00:05:34,400 --> 00:05:37,479 Speaker 1: have the power, by vertue of being connected to the 86 00:05:37,480 --> 00:05:40,000 Speaker 1: world's computer networks that if it goes off in the 87 00:05:40,000 --> 00:05:43,680 Speaker 1: wrong direction, it could read enormous havoc. Just like today, 88 00:05:44,040 --> 00:05:46,040 Speaker 1: termites could throw us back to the Stone Age in 89 00:05:46,080 --> 00:05:49,800 Speaker 1: a week if they got intelligent about it. They just won't, fortunately, 90 00:05:49,839 --> 00:05:53,960 Speaker 1: but they're distributed enough to destroy enough of us to 91 00:05:54,040 --> 00:05:58,159 Speaker 1: do that. If anything that's connected to the world's computer 92 00:05:58,240 --> 00:06:03,440 Speaker 1: networks evolved what we might call an intent in that direction, 93 00:06:03,760 --> 00:06:07,520 Speaker 1: we would have a serious problem. I mean, is it 94 00:06:07,560 --> 00:06:10,840 Speaker 1: that you're worried or that you're thinking that I think 95 00:06:10,880 --> 00:06:14,160 Speaker 1: about our defense systems, that our military systems. Is that 96 00:06:14,200 --> 00:06:17,680 Speaker 1: what could create or you know, artificial superintelligence could reach 97 00:06:17,720 --> 00:06:22,400 Speaker 1: havoc upon It's certainly one vector, and it's a popular 98 00:06:22,400 --> 00:06:24,520 Speaker 1: one because you can't talk about this for long without 99 00:06:24,560 --> 00:06:28,680 Speaker 1: thinking about the terminator, right, and and and and they're 100 00:06:28,720 --> 00:06:34,120 Speaker 1: also far more elegant ways of acting against the human 101 00:06:34,200 --> 00:06:38,279 Speaker 1: race and the nuclear weapons. But the side effects don't 102 00:06:38,279 --> 00:06:44,719 Speaker 1: even require consciousness or malevolence or intent to happen. Nick 103 00:06:44,800 --> 00:06:49,200 Speaker 1: Bostrom of Oxford University, England outlined a whole lot of 104 00:06:49,240 --> 00:06:54,320 Speaker 1: scenarios where we could be destroyed by artificial intelligence just 105 00:06:54,480 --> 00:06:58,080 Speaker 1: trying to do exactly what we told it to. Well, 106 00:06:58,120 --> 00:07:01,520 Speaker 1: that's kind of terrifying. Well it is, and I don't 107 00:07:01,520 --> 00:07:03,520 Speaker 1: want to anyone to take come away from us thinking 108 00:07:03,520 --> 00:07:07,920 Speaker 1: we've got to stop this, because firstly we can't. But secondly, 109 00:07:08,120 --> 00:07:11,000 Speaker 1: there are other problems to which this can solve. If 110 00:07:11,040 --> 00:07:16,680 Speaker 1: we develop AI ethically and compassionately, and that we understand 111 00:07:16,720 --> 00:07:19,640 Speaker 1: what ethics and compassion are the humans on a much 112 00:07:19,680 --> 00:07:23,280 Speaker 1: greater level, then we can ensure that they will be 113 00:07:23,360 --> 00:07:26,200 Speaker 1: on our side and help protect us from our own 114 00:07:26,320 --> 00:07:30,000 Speaker 1: worst impulses, like the possibility of people taking the same 115 00:07:30,520 --> 00:07:36,800 Speaker 1: exponential advances in technology too cause habit with bio terrorism 116 00:07:36,920 --> 00:07:39,520 Speaker 1: launched from their basement, for instance. It sounds like we 117 00:07:39,560 --> 00:07:43,280 Speaker 1: have to be intelligence intelligent, if you will, about the 118 00:07:43,360 --> 00:07:47,400 Speaker 1: artificial intelligence that's you know, increasingly evading or invading I 119 00:07:47,400 --> 00:07:49,920 Speaker 1: should say our world. Um, I'm taking this book on 120 00:07:50,000 --> 00:07:52,920 Speaker 1: vacation with me because I think it's fascinating. Peter Scott, 121 00:07:52,960 --> 00:07:55,920 Speaker 1: thank you so much, author his books crisis of control, 122 00:07:55,960 --> 00:08:00,400 Speaker 1: How artificial super intelligence may destroy or save the human race. 123 00:08:00,800 --> 00:08:04,720 Speaker 1: You listen to Bloomberg Markets. I'm Carol Master in Bloomberg 124 00:08:04,760 --> 00:08:07,920 Speaker 1: eleven three oh Studio. Big news today, President Trump ousting 125 00:08:07,960 --> 00:08:11,880 Speaker 1: Anthony Scaramucci of another shake up UH in the administration. 126 00:08:11,960 --> 00:08:13,480 Speaker 1: You are listening to Bloomberg Radio.