1 00:00:04,849 --> 00:00:08,689 And now, work. What will you be doing as a job in the future? 2 00:00:09,070 --> 00:00:12,490 Well, 50 years ago, the world of work was pretty easy to understand. 3 00:00:12,490 --> 00:00:16,929 You either did manual work, which basically meant supervising machines, 4 00:00:17,230 --> 00:00:20,510 or you worked in an office, which basically meant doing a lot of typing 5 00:00:20,510 --> 00:00:23,070 or getting somebody else to do a lot of typing for you. 6 00:00:23,550 --> 00:00:28,789 Bottom line, people were integral to the workforce, but no more. 7 00:00:29,149 --> 00:00:32,509 Because in the 1970s, machines got clever. 8 00:00:33,009 --> 00:00:34,609 Car plants were filled with robots, 9 00:00:35,030 --> 00:00:37,130 helplines were answered by computers 10 00:00:37,130 --> 00:00:40,310 and almost all bank clerks became extinct. 11 00:00:40,689 --> 00:00:42,210 But I suspect that most of you are saying, 12 00:00:42,350 --> 00:00:44,789 well, a robot couldn't possibly take my job. 13 00:00:45,109 --> 00:00:46,450 But are you sure? 14 00:00:46,990 --> 00:00:48,130 Have a look at this. 15 00:00:50,090 --> 00:00:53,530 We sent Dr Zoe Williams to check out a piece of software 16 00:00:53,530 --> 00:00:56,630 which is rumoured to diagnose illnesses faster 17 00:00:56,630 --> 00:01:00,250 and more accurately than human medical professionals. 18 00:01:00,250 --> 00:01:08,109 so should I be worried as a GP the thought that a robot or artificial intelligence could take 19 00:01:08,109 --> 00:01:14,049 my job just seems crazy I mean I've spent six years at medical school ten years practicing 20 00:01:14,049 --> 00:01:23,280 as a doctor now surely all of that can't be boiled down to a few lines of code Babylon 21 00:01:23,280 --> 00:01:29,819 health are a medical tech company they've just received sixty million dollars of funding to 22 00:01:29,819 --> 00:01:37,680 develop an AI doctor the system works by asking questions but anyone can ask 23 00:01:37,680 --> 00:01:42,620 questions if it's going to replace me I really want to put it through its paces 24 00:01:42,620 --> 00:01:47,760 I'm gonna pose as a patient and give myself an imaginary condition and I'm 25 00:01:47,760 --> 00:01:54,900 not going to tell anybody I'm just gonna write it down and then we can see just 26 00:01:54,900 --> 00:02:00,780 how accurate the machine really is may I ask please what's troubling you today 27 00:02:00,780 --> 00:02:08,879 I'm feeling tired all the time so as well as feeling tired I've been feeling 28 00:02:08,879 --> 00:02:17,419 kind of weak let's tell the computer that and I've also been feeling a bit of 29 00:02:17,419 --> 00:02:22,860 dizziness is it okay to ask do you have painful 30 00:02:22,860 --> 00:02:31,740 periods there we go I said so painful periods as well do you get breathless on exertion yes I do 31 00:02:31,740 --> 00:02:39,960 thanks I've noted this so I've given the computer all of my symptoms now and it's come up with a 32 00:02:39,960 --> 00:02:47,759 diagnosis so let's see if it's correct here's my bit of paper from earlier and you can see that I 33 00:02:47,759 --> 00:02:52,780 I have put down fibroids and the computer has said 34 00:02:52,780 --> 00:02:56,759 uterine leiomyoma, which is actually the same thing. 35 00:02:57,560 --> 00:03:00,580 That's impressive, but how's it done? 36 00:03:01,319 --> 00:03:03,360 Time to face down the evil genius 37 00:03:03,360 --> 00:03:05,979 hell-bent on replacing me with my laptop. 38 00:03:06,900 --> 00:03:10,340 So you start off with a knowledge base, 39 00:03:10,620 --> 00:03:12,680 and this is essentially a medical database 40 00:03:12,680 --> 00:03:15,020 which contains hundreds of millions of medical concepts. 41 00:03:15,240 --> 00:03:17,199 That's kind of like being at medical school 42 00:03:17,199 --> 00:03:20,039 and all the knowledge that is inputted into the brain. 43 00:03:20,120 --> 00:03:22,099 Exactly, so this might be all of the textbooks 44 00:03:22,099 --> 00:03:23,319 which you've read at medical school, 45 00:03:23,460 --> 00:03:25,240 all of the papers which you've read at medical school, 46 00:03:25,439 --> 00:03:27,539 and then apply it to all of that information. 47 00:03:27,919 --> 00:03:31,460 We'll apply a set of methods known as machine learning methods. 48 00:03:33,520 --> 00:03:36,180 Machine learning is the ability of computers 49 00:03:36,180 --> 00:03:40,039 to take vast amounts of data and make sense of it themselves. 50 00:03:41,919 --> 00:03:44,080 Like this network of medical information 51 00:03:44,080 --> 00:03:46,740 which the computer uses to make a diagnosis. 52 00:03:47,199 --> 00:03:53,479 What those circles represent are diseases, symptoms and risk factors, and what those 53 00:03:53,479 --> 00:03:58,280 lines represent are the relationships between those. So based on that the computer has taught 54 00:03:58,280 --> 00:04:03,639 itself actually how strongly related those diseases, symptoms and risk factors are. 55 00:04:03,639 --> 00:04:08,800 OK, so that's how it determines the probability is from looking at past real life cases? 56 00:04:08,800 --> 00:04:15,210 Absolutely, and that's why this is machine learning. 57 00:04:15,210 --> 00:04:20,689 the network learns about more and more symptoms and diseases, it's tested and refined by a team 58 00:04:20,689 --> 00:04:27,670 of doctors and programmers. It's early days, but the company sees a big future for their virtual 59 00:04:27,670 --> 00:04:33,329 medic. We will do with healthcare what, say, Google did with information. It'll be in your 60 00:04:33,329 --> 00:04:38,490 phone, it'll be in the devices you understand, you carry with you. Do you think that a machine 61 00:04:38,490 --> 00:04:41,829 could ever replace my role as a GP? 62 00:04:42,310 --> 00:04:46,170 I don't think this is a competition between machines and humans. 63 00:04:46,529 --> 00:04:49,470 This is machines being an aid to humans. 64 00:04:49,930 --> 00:04:52,730 Half of the world's population has no access 65 00:04:52,730 --> 00:04:56,009 or very, very little access to doctors, right? 66 00:04:56,129 --> 00:04:58,370 Imagine if you could see so many more, 67 00:04:58,529 --> 00:05:01,750 because the machines do the easier part, they save your time. 68 00:05:02,129 --> 00:05:05,870 But can a machine put its hand on your shoulder and say, 69 00:05:05,870 --> 00:05:08,509 Trust me, I look after you. That's a different story. 70 00:05:10,509 --> 00:05:13,889 It's not just in medicine that software's on the march. 71 00:05:14,569 --> 00:05:18,850 In banking, AI is approving or not approving loan applications 72 00:05:18,850 --> 00:05:21,629 and even making investment decisions. 73 00:05:22,509 --> 00:05:25,269 And with autonomous vehicles on the horizon, 74 00:05:25,850 --> 00:05:28,910 many who drive for a living will soon be superseded. 75 00:05:30,850 --> 00:05:32,529 What you need to know about the future 76 00:05:32,529 --> 00:05:36,629 is that no job is immune from the influence of artificial intelligence. 77 00:05:37,069 --> 00:05:38,389 And if it doesn't take your job, 78 00:05:38,730 --> 00:05:41,490 then it's likely to change the way in which you do it.