WEBVTT 00:00:00.030 --> 00:00:02.200 Let's start then. 00:00:02.370 --> 00:00:03.490 Good evening. 00:00:03.490 --> 00:00:05.730 Thank you very much for coming. 00:00:05.730 --> 00:00:06.280 My name is Ben. 00:00:06.460 --> 00:00:12.560 I think our time and attention is one of the most valuable things we have 00:00:12.640 --> 00:00:16.880 and you have kindly offered me yours for the next 60 minutes or so. 00:00:16.880 --> 00:00:18.370 So thank you very much. 00:00:18.370 --> 00:00:20.360 I really appreciate it. 00:00:20.360 --> 00:00:26.260 The latest technology promising more of our time back to us is AI. 00:00:26.260 --> 00:00:32.520 Now I know many of you probably are sick of hearing about AI but I'm afraid 00:00:32.520 --> 00:00:34.750 we can't bury our heads in the sand on this one. 00:00:34.940 --> 00:00:40.600 We have to face the consequences it presents to us head-on. 00:00:40.600 --> 00:00:47.480 Many parents I talked to are concerned about AI's impacts on their kids 00:00:47.480 --> 00:00:52.950 or students learning but I'm far more concerned about AI making certain 00:00:53.340 --> 00:00:54.960 learning obsolete. 00:00:54.960 --> 00:01:00.560 I used to implement AI for businesses and I can sort of see this coming 00:01:00.560 --> 00:01:05.150 and that's what we're going to talk about this evening. 00:01:05.150 --> 00:01:08.360 AI's impact on entry-level jobs. 00:01:08.360 --> 00:01:16.140 Now we're going to look at the trends of previous technologies and we'll 00:01:16.140 --> 00:01:21.370 see that the progress seems to follow certain rules or laws and the progress 00:01:21.580 --> 00:01:27.130 is extremely predictable and we're going to see if that those certain trends 00:01:27.130 --> 00:01:29.830 or that progress also applies to AI. 00:01:29.830 --> 00:01:35.040 We're going to look at AI's capabilities because of those trends and see 00:01:35.040 --> 00:01:40.860 if that translates to any impacts showing up in the employment data or labour 00:01:41.090 --> 00:01:42.180 market data. 00:01:42.180 --> 00:01:46.350 And finally we're going to see what to do about it. 00:01:46.350 --> 00:01:52.720 But to give you a hint there's a reason I left my job as an AI engineer 00:01:52.750 --> 00:01:58.960 to found my company Stable Future and talk with you tonight because I think 00:01:58.990 --> 00:02:02.360 those impacts are starting to show up. 00:02:02.360 --> 00:02:07.230 So let's begin. 00:02:07.230 --> 00:02:12.950 The progress of certain technologies is extremely predictable. 00:02:12.950 --> 00:02:18.220 In fact sometimes it's so predictable that when we plot the progress on 00:02:18.220 --> 00:02:23.840 a graph they seem to follow straight lines and the progress follows certain 00:02:24.010 --> 00:02:25.120 rules or laws. 00:02:25.120 --> 00:02:30.040 So to see the future we don't need to believe in science fiction. 00:02:30.040 --> 00:02:34.440 We just need to look at straight lines on graphs. 00:02:34.470 --> 00:02:38.750 The most famous of these straight lines was discovered by Gordon Moore. 00:02:38.750 --> 00:02:39.630 Here he is. 00:02:39.760 --> 00:02:43.260 Discovered it in 1965. 00:02:43.260 --> 00:02:49.550 He saw that the progress and the power of computers was doubling roughly 00:02:49.550 --> 00:02:54.670 every two years and that this doubling every two years will continue for 00:02:54.670 --> 00:02:56.320 at least a decade. 00:02:56.320 --> 00:03:02.300 But it didn't continue for just one decade it continued for six and it's 00:03:02.300 --> 00:03:04.600 still alive to date. 00:03:04.600 --> 00:03:12.190 Now as promised this is a fairly straight line but that's only because we've 00:03:12.190 --> 00:03:14.160 changed the y-axis here. 00:03:14.160 --> 00:03:16.430 It's not the y-axis that we're all used to. 00:03:16.530 --> 00:03:19.850 Here we're measuring computing power on the y-axis. 00:03:19.850 --> 00:03:25.480 Normally it would start at zero and goes from 1,000 to 2,000 to 3,000 add 00:03:25.480 --> 00:03:31.440 up a thousand every time but this one multiplies by a thousand every time 00:03:31.440 --> 00:03:37.520 so it doesn't go it goes from one to a thousand to a million and so on. 00:03:37.520 --> 00:03:40.510 Which means this graph is hiding something. 00:03:40.700 --> 00:03:46.230 It's hiding something that is every tech nerd's favourite word. 00:03:46.230 --> 00:03:49.240 Mine included exponential. 00:03:49.240 --> 00:03:53.910 Computing power progressed exponentially over these 60 years and computing 00:03:53.910 --> 00:03:58.900 power didn't increase a thousand times or 10,000 times it increased 300 00:03:58.900 --> 00:04:05.920 billion times and at the same time as that 300 billion times increase in 00:04:05.920 --> 00:04:09.960 computing power there was a efficiencies in the manufacturing of computers 00:04:09.960 --> 00:04:16.000 also led to decreases in the cost of computing power. 00:04:16.020 --> 00:04:21.940 That was also exponential and it didn't just decrease a thousand times 10,000 00:04:22.150 --> 00:04:25.410 times it decreased 30 trillion times. 00:04:25.560 --> 00:04:32.320 So those forces interacted together which means that today we each have 00:04:32.320 --> 00:04:37.500 super computers in our pockets that are more powerful than the ones that 00:04:37.500 --> 00:04:40.580 took us to the moon in 1969. 00:04:40.580 --> 00:04:43.730 I think that's pretty amazing. 00:04:43.880 --> 00:04:51.220 So the story here is that the cost of computing power decreased exponentially 00:04:51.730 --> 00:04:57.190 at the same time that the computing power increased exponentially. 00:04:57.190 --> 00:05:03.660 So those forces interacted together which led to computing power and computers 00:05:03.660 --> 00:05:08.260 diffusing throughout the economy and being adopted everywhere. 00:05:08.260 --> 00:05:13.800 And again it was adopted, guess what, exponentially. 00:05:13.800 --> 00:05:22.440 So we don't just see that progress or this pattern show up in computing 00:05:22.550 --> 00:05:23.400 alone. 00:05:23.400 --> 00:05:26.010 We also see it in solar panels. 00:05:26.010 --> 00:05:32.660 We see it in wind turbines, DNA sequencing, rechargeable batteries and LED 00:05:33.110 --> 00:05:33.590 lights. 00:05:33.760 --> 00:05:38.350 I think you can see where I'm going with this. 00:05:38.750 --> 00:05:42.440 So AI is the latest technology on everyone's minds. 00:05:42.610 --> 00:05:45.140 We need to measure two things. 00:05:45.140 --> 00:05:49.450 We need to measure the cost of AI and we need to measure the performance 00:05:49.490 --> 00:05:49.920 of AI. 00:05:49.970 --> 00:05:56.770 And if we see similar trends in AI then we can predict with a reasonable 00:05:56.920 --> 00:06:01.370 amount of certainty that AI is going to be used absolutely everywhere in 00:06:01.370 --> 00:06:05.020 the economy over the coming years. 00:06:05.020 --> 00:06:10.230 So for the cost, the cost is quite easy to measure, right. 00:06:10.230 --> 00:06:11.340 It's very quantifiable. 00:06:11.360 --> 00:06:13.400 Just look at the receipt, right. 00:06:13.400 --> 00:06:16.890 Very easy to put a number on the cost. 00:06:16.890 --> 00:06:21.160 But the performance is something different. 00:06:21.160 --> 00:06:26.440 For LEDs it's quite easy, you just look at how much light is emitted for 00:06:26.440 --> 00:06:31.800 a certain amount of energy or for rechargeable batteries like how much energy 00:06:31.940 --> 00:06:34.430 the battery can store over time. 00:06:34.520 --> 00:06:39.240 But how would we measure the performance of AI? 00:06:39.240 --> 00:06:43.870 One way is to just say well we can kind of tell. 00:06:43.870 --> 00:06:46.000 It's kind of obvious, right. 00:06:46.000 --> 00:06:52.640 Here is a video, an AI generated video of Will Smith eating spaghetti in 00:06:53.060 --> 00:06:55.310 2023. 00:06:55.310 --> 00:06:57.480 It's not very realistic. 00:06:57.510 --> 00:06:59.110 Spaghetti going everywhere. 00:06:59.110 --> 00:07:01.280 His mouth's moving everywhere as well. 00:07:01.280 --> 00:07:02.400 It's crazy. 00:07:02.430 --> 00:07:09.890 Here is the latest video model with the same instructions 2026. 00:07:10.360 --> 00:07:13.240 Sound effects included. 00:07:13.240 --> 00:07:25.630 Smooth. 00:07:25.630 --> 00:07:28.420 Smooth. 00:07:28.580 --> 00:07:31.160 So performance is clearly increased. 00:07:31.160 --> 00:07:33.850 We can't put a number on it, right. 00:07:33.980 --> 00:07:36.620 It's not very quantifiable. 00:07:36.620 --> 00:07:41.540 A better way is to use the tests that we already have to measure our own 00:07:42.180 --> 00:07:42.480 performance. 00:07:42.480 --> 00:07:49.160 Things like, you know, tests, exams, competitions or if you're really smart, 00:07:49.160 --> 00:07:55.640 scientific breakthroughs like Newtonian physics or Darwinian evolution. 00:07:55.640 --> 00:08:07.140 So let's use those tests and let's start off easy with primary school maths. 00:08:07.460 --> 00:08:13.990 Models in the middle of 2023, sorry, the middle of 2020. 00:08:13.990 --> 00:08:18.000 This was a model from the makers of ChatGPT called OpenAI. 00:08:18.030 --> 00:08:22.230 Middle of 2020 they're getting just under 10% on primary school maths. 00:08:22.580 --> 00:08:24.070 Not very good. 00:08:24.200 --> 00:08:28.480 Thing to understand about AI is that they are computer brains. 00:08:28.480 --> 00:08:33.060 They are neural networks, similar and based off the same architecture as 00:08:33.060 --> 00:08:34.000 our own brains. 00:08:34.030 --> 00:08:36.570 So it's just a computer brain. 00:08:36.570 --> 00:08:42.490 The researchers made the computer brain AI more efficient and with those 00:08:42.490 --> 00:08:49.010 efficiencies after two and a half years of research, the next best model 00:08:49.010 --> 00:08:51.160 scored 50%. 00:08:51.160 --> 00:08:54.570 Good improvement but two and a half years. 00:08:54.660 --> 00:08:58.010 and we're only on primary school maths. 00:08:58.010 --> 00:09:04.100 So instead they made the computer brain ten times bigger with better computers, 00:09:04.100 --> 00:09:10.550 bigger data centres and finally the test was declared effectively solved 00:09:10.690 --> 00:09:13.760 when that model scored 90%. 00:09:13.760 --> 00:09:17.330 So we're gonna have to find a new test. 00:09:17.330 --> 00:09:23.740 But researchers at this point thought the progress was a bit too 00:09:23.740 --> 00:09:24.350 slow. 00:09:24.570 --> 00:09:28.540 They needed a new way to progress these models. 00:09:28.910 --> 00:09:33.630 So what they did was they taught the models to reason through the answer 00:09:33.650 --> 00:09:35.350 before they gave it. 00:09:35.440 --> 00:09:40.650 So you give it a maths question, it breaks down, like we do, the maths question, 00:09:40.660 --> 00:09:44.760 into steps, reasons through it and then produces its answer. 00:09:44.760 --> 00:09:48.000 And this, this is called the reasoning paradigm. 00:09:48.000 --> 00:09:53.720 You can think of it as the models, the reasoning and thinking for themselves. 00:09:53.720 --> 00:10:00.540 This turned out to be a breakthrough because only two years after this point, 00:10:00.540 --> 00:10:07.220 models that were released summer of last year, Gemini and ChatGPT won a 00:10:07.220 --> 00:10:11.640 gold at the International Maths Olympiad. 00:10:11.790 --> 00:10:18.870 So this is at the level of the brightest young maths talent in the world. 00:10:18.990 --> 00:10:25.000 Over the same time, over that two-year period to summer of last year, they 00:10:25.000 --> 00:10:31.750 also beat PhD level experts in their own fields in physics, chemistry and 00:10:31.980 --> 00:10:32.510 biology. 00:10:32.520 --> 00:10:37.490 So physics questions, PhD answered physics questions right in their own 00:10:37.660 --> 00:10:45.950 field, they got 65% models today get 95% significantly outperforming PhD 00:10:46.200 --> 00:10:46.760 experts. 00:10:46.940 --> 00:10:53.120 If we were to measure their IQ, they were significantly below the human 00:10:53.250 --> 00:10:58.340 average, now they're significantly above even the top 1%. 00:10:58.340 --> 00:11:04.090 And however we measure this progress, whether it's in reading a clock, which 00:11:04.340 --> 00:11:09.300 they're still below human, strangely enough, or general common sense, fixing 00:11:09.710 --> 00:11:15.340 software bugs, no matter what it is, progress is only going in one direction, 00:11:15.340 --> 00:11:18.600 is only going up and to the right. 00:11:18.600 --> 00:11:24.180 So this is up to, again, summer of last year. 00:11:24.180 --> 00:11:29.630 And up to this point, there are questions about whether AI could reason 00:11:29.690 --> 00:11:34.710 and generalise beyond its training data, or whether it's just a parrot, 00:11:34.860 --> 00:11:39.310 just regurgitating the answers we give it in the training data. 00:11:39.350 --> 00:11:43.690 So there are two sides to this debate. 00:11:43.900 --> 00:11:49.190 And this year, we found out the answer, because just a few weeks ago, the 00:11:49.380 --> 00:11:55.240 latest AI models, again, bigger computers, more efficiencies in the computer 00:11:55.450 --> 00:11:59.710 brains, they solved a Millennium Prize problem in maths. 00:11:59.710 --> 00:12:03.500 Millennium Prize problem, there's only seven of them in the world, they 00:12:03.500 --> 00:12:07.000 each have a million dollar bounty for solving them. 00:12:07.000 --> 00:12:12.290 And some of these problems, they've stood for decades, or a century, or 00:12:12.820 --> 00:12:14.360 more. 00:12:14.360 --> 00:12:15.220 And it solved one of them. 00:12:15.240 --> 00:12:18.350 This one in particular is called the Navier-Stokes problem, and it's to 00:12:18.350 --> 00:12:24.520 do with fluid dynamics, the way water moves in a container, for example. 00:12:24.520 --> 00:12:28.440 So it's making scientific breakthroughs. 00:12:29.190 --> 00:12:34.670 And it's important to note, this isn't like a maths researcher that's sort 00:12:34.670 --> 00:12:39.480 of leveraging an AI model to, you know, go back and forth and iterate on 00:12:39.480 --> 00:12:39.830 ideas. 00:12:39.840 --> 00:12:45.400 This in particular, and many other maths problems solved by AI, people just 00:12:45.400 --> 00:12:51.760 say, do a breakthrough, and try really hard, and believe in yourself, and 00:12:51.760 --> 00:12:55.250 the models go away and solve some of these problems. 00:12:55.440 --> 00:12:59.730 It doesn't happen every time, but that's one of the methods researchers 00:13:00.500 --> 00:13:04.460 are using to make frontier breakthroughs like this one. 00:13:05.350 --> 00:13:11.330 But this is a talk on entry-level jobs. 00:13:11.330 --> 00:13:12.850 That's what you're here for. 00:13:12.880 --> 00:13:15.010 There's not many maths researchers in the world. 00:13:15.010 --> 00:13:18.170 You might be having a bit of trouble, but it's not really representative 00:13:19.210 --> 00:13:21.880 of the economy and real-world jobs. 00:13:21.880 --> 00:13:29.230 So what we'll need is a test that tests for exactly that. 00:13:29.230 --> 00:13:36.200 And my last benchmark, my last exam, will show exactly this. 00:13:36.250 --> 00:13:41.450 But what you've got to understand is that we've moved on from the reasoning 00:13:41.450 --> 00:13:45.680 paradigm or the reasoning era to the agentic era. 00:13:45.680 --> 00:13:49.360 You might have heard of AI agents or agentic AI. 00:13:49.360 --> 00:13:55.760 These are AI models that they still break down the task that you give them 00:13:55.970 --> 00:13:58.470 into steps, but now they achieve goals. 00:13:58.470 --> 00:14:04.430 So instead of asking a model like ChatGPT, instead of asking them a question, 00:14:04.430 --> 00:14:05.740 you give them a goal. 00:14:05.750 --> 00:14:08.360 You say, you know, sort out this spreadsheet. 00:14:08.360 --> 00:14:12.750 It will break that goal or that task down into steps. 00:14:12.750 --> 00:14:18.060 It will complete each step and therefore complete the goal. 00:14:18.060 --> 00:14:26.280 So these models, ChatGPT, anything you do on a computer, it can now also 00:14:26.280 --> 00:14:27.480 do. 00:14:27.480 --> 00:14:29.640 This is the agentic era. 00:14:29.640 --> 00:14:34.210 My last exam is called agents last exam. 00:14:34.260 --> 00:14:38.270 It's across all of these domains. 00:14:38.270 --> 00:14:42.840 There's 1,500 tasks in this benchmark. 00:14:42.870 --> 00:14:47.780 It's across, you know, engineering, agriculture, legal, health and medicine, 00:14:47.780 --> 00:14:49.210 you name it, it's on there. 00:14:49.280 --> 00:14:53.240 This is a representation of what I like to call the laptop class, which 00:14:53.380 --> 00:14:57.330 I am a part of or the, you know, the knowledge economy. 00:14:57.480 --> 00:15:04.700 Each one of these 1,500 tasks was handpicked by an expert in their domain 00:15:04.860 --> 00:15:07.960 and it was a past project that they did. 00:15:07.960 --> 00:15:11.530 And it often took some, it took them an average of days or weeks for them 00:15:11.550 --> 00:15:13.440 to complete this project. 00:15:13.440 --> 00:15:19.380 And if you had to add up all the time it takes to do these 1,500 tasks, 00:15:19.380 --> 00:15:25.870 it would take someone their entire career to do them, to work through them. 00:15:26.010 --> 00:15:28.280 Here's just one example. 00:15:28.280 --> 00:15:29.980 It's climate science. 00:15:30.040 --> 00:15:35.000 So a climate scientist would have to look through images and figure out 00:15:35.000 --> 00:15:39.880 how a lake in Greenland got drained or how it disappeared due to ice caps 00:15:40.670 --> 00:15:43.160 melting and things like that. 00:15:43.160 --> 00:15:49.420 6,000 images, you have to sift through three gigabytes of data. 00:15:49.420 --> 00:15:52.950 So let's see the progress. 00:15:53.140 --> 00:15:55.110 Let's have a look at the progress. 00:15:55.110 --> 00:16:00.630 Those models around the same time was that International Maths Olympiad gold 00:16:00.630 --> 00:16:04.420 last year, they got 10%. 00:16:04.550 --> 00:16:08.470 Not that good, I mean kind of impressive because it is, you know, completing 00:16:08.650 --> 00:16:11.320 these tasks but it's only 10%. 00:16:11.350 --> 00:16:16.600 Again, bigger computers, more efficiencies. 00:16:16.600 --> 00:16:23.960 Three months ago the latest models got 50%, 50%, half of a representation 00:16:23.960 --> 00:16:27.320 of the entire knowledge economy. 00:16:27.320 --> 00:16:36.690 Now, how long did it take this model to complete all of these 1,500 tasks? 00:16:36.720 --> 00:16:38.430 Well, half of them. 00:16:38.430 --> 00:16:44.020 It attempted, it attempted all 1,500 but obviously only got 50%. 00:16:44.020 --> 00:16:47.310 So how long did it take for the model to work through these problems? 00:16:47.580 --> 00:16:50.350 It took less than a month. 00:16:50.350 --> 00:16:55.900 And how much did it cost to achieve 50%? 00:16:55.900 --> 00:16:58.730 Cost 2,000 pounds. 00:16:58.880 --> 00:17:00.180 Exactly. 00:17:00.390 --> 00:17:07.920 And if you think that progress has stopped and that we might have hit a 00:17:07.920 --> 00:17:13.190 wall and there's no more progress to be had, well, a few weeks ago the latest 00:17:13.190 --> 00:17:15.400 model got 60%. 00:17:15.400 --> 00:17:21.050 And in fact, last week, I've not added it to this because it's hard to keep 00:17:21.070 --> 00:17:24.760 up with all the progress but last week a model got 63%. 00:17:24.770 --> 00:17:31.220 And so now we're no longer measuring progress in years, not even measuring 00:17:31.510 --> 00:17:33.580 progress in months, we're now measuring it in weeks. 00:17:33.670 --> 00:17:41.720 So progress is occurring exponentially, performance is exponentially increasing. 00:17:41.720 --> 00:17:44.840 So that's what's happening with performance. 00:17:45.050 --> 00:17:50.400 With costs, last year an international maths Olympiad gold would cost 2,000 00:17:50.480 --> 00:17:53.900 pounds, this year it would cost 20 pounds. 00:17:54.060 --> 00:18:00.670 So it depends on the domain and the sector that you're measuring but costs 00:18:00.810 --> 00:18:05.690 are decreasing about 10 times a year, something that cost 1,000 pounds, 00:18:05.690 --> 00:18:09.400 last year it cost 100 pounds, this year 10 pounds, next year. 00:18:09.400 --> 00:18:13.970 That is quite predictable progress. 00:18:13.970 --> 00:18:22.650 Now, you might be thinking, okay, it's getting 60% on agents last exam, 00:18:22.650 --> 00:18:26.240 the job market has to be in freefall, right? 00:18:26.370 --> 00:18:29.170 But that's not what we see. 00:18:29.170 --> 00:18:34.430 Like I said, it takes time for technology to diffuse throughout the economy 00:18:34.590 --> 00:18:38.990 and probably some of you here haven't heard of these results, which means 00:18:39.390 --> 00:18:42.840 business owners probably haven't heard of these results and they've not 00:18:42.870 --> 00:18:43.410 caught on yet. 00:18:43.410 --> 00:18:48.320 So again, it takes time for the technology to be adopted. 00:18:48.320 --> 00:18:54.640 So let's look at current job market impacts. 00:18:54.640 --> 00:19:01.100 The best, the most in-depth studies, what they do is they measure a sector's 00:19:01.100 --> 00:19:02.710 exposure to AI. 00:19:02.710 --> 00:19:09.550 So that's just like, you know, you take a job, you see which tasks AI can 00:19:09.550 --> 00:19:15.840 do and the more tasks AI can do, the more exposed that job is to AI, right? 00:19:16.180 --> 00:19:22.390 The best, the most in-depth studies show that employment in the most AI 00:19:22.470 --> 00:19:28.780 exposed sectors is actually up since the release of ChatGPT, which is really 00:19:28.990 --> 00:19:29.470 strange. 00:19:29.800 --> 00:19:35.100 You think it would be the opposite but those same studies look to the effect 00:19:35.100 --> 00:19:35.940 of age. 00:19:35.940 --> 00:19:38.790 So this comes with a huge caveat. 00:19:39.070 --> 00:19:41.360 Senior employment is up. 00:19:41.360 --> 00:19:42.790 Some of their wages. 00:19:42.790 --> 00:19:45.120 Seniors are having a great time. 00:19:45.120 --> 00:19:50.620 Even the most exposed seniors, employment is up compared to the least exposed 00:19:50.620 --> 00:19:51.610 seniors. 00:19:51.760 --> 00:19:57.320 But when we look at junior employment, we see the opposite. 00:19:57.450 --> 00:20:02.400 Junior employment in the most exposed sectors is down. 00:20:02.400 --> 00:20:04.560 Why is this? 00:20:04.560 --> 00:20:10.110 Now there are a few theories, but I think the one that makes the most sense 00:20:10.110 --> 00:20:15.960 is that seniors are the task setters, juniors are the task executors. 00:20:16.130 --> 00:20:20.060 And AI has become an excellent executor of tasks. 00:20:20.220 --> 00:20:28.500 It's now better, often better, faster, cheaper and safer than a junior. 00:20:28.500 --> 00:20:32.640 So normally a senior would set the task, give it to the junior, junior would 00:20:32.690 --> 00:20:38.000 produce an output and the senior gives feedback, iterates on the output 00:20:38.000 --> 00:20:40.400 and then verifies, yep that's good. 00:20:40.430 --> 00:20:41.330 Off you go. 00:20:41.380 --> 00:20:47.480 AI is now replacing that role of juniors of executing tasks. 00:20:48.350 --> 00:20:51.240 This is a standard corporate pyramid. 00:20:51.240 --> 00:20:54.180 You've got seniors at the top, juniors at the bottom. 00:20:54.260 --> 00:20:59.440 What we're seeing now is this in the most exposed sectors. 00:20:59.440 --> 00:21:03.710 AI is beginning to take entry-level roles. 00:21:03.710 --> 00:21:09.920 I believe in the coming years, it remains to be seen, but I think with a 00:21:10.060 --> 00:21:15.200 fair amount of certainty we'll start to see this AI climbing the corporate 00:21:15.200 --> 00:21:18.230 pyramid taking more entry-level roles. 00:21:18.230 --> 00:21:24.350 Now whether or not it can climb any further than this, whether or not it 00:21:24.780 --> 00:21:30.350 starts setting the tasks for perhaps even AI to do, remains to be seen. 00:21:30.350 --> 00:21:36.050 We're not sure, but certainly this will start happening. 00:21:36.050 --> 00:21:38.270 It already is. 00:21:38.270 --> 00:21:45.940 So the number of juniors entering the entry-level job market through things 00:21:45.960 --> 00:21:50.330 like apprenticeships or degrees, that's sort of fixed for the next few years. 00:21:50.330 --> 00:21:54.740 People are already in university, people have already started apprenticeships. 00:21:54.970 --> 00:21:58.800 However the number of opportunities is shrinking. 00:21:58.830 --> 00:22:02.760 So that means competition is rising. 00:22:02.760 --> 00:22:07.480 There is a bar to land a job in these exposed sectors. 00:22:07.480 --> 00:22:13.480 Above the bar you are leveraging AI, and below the bar you are replaced 00:22:14.310 --> 00:22:15.760 by AI. 00:22:15.760 --> 00:22:20.800 The more exposed the job, the more you will need AI skills because you'll 00:22:20.800 --> 00:22:25.950 be leveraging AI, grades to stand out from the competition, work experience 00:22:25.990 --> 00:22:30.760 to show that you know about the sector, and network, which is something 00:22:30.760 --> 00:22:37.340 people often miss, because AI is now flooding online applications. 00:22:37.480 --> 00:22:41.920 People use AI to make their applications more efficient, and now employers 00:22:42.320 --> 00:22:45.290 can't keep up with the number of applications that they've got. 00:22:45.290 --> 00:22:49.900 And so they're going through people they know or recruiters to fill the 00:22:50.260 --> 00:22:51.050 roles. 00:22:51.050 --> 00:22:55.460 And lastly, evidence of enthusiasm. 00:22:55.530 --> 00:22:59.920 Like all I mean by this is a personal project in the sector to show that 00:22:59.920 --> 00:23:02.480 you're genuinely interested in the work. 00:23:02.480 --> 00:23:06.860 All of that will require work ethic and stress tolerance because it's on 00:23:06.860 --> 00:23:11.170 top of the expectations that you already have as a student, on top of the 00:23:11.180 --> 00:23:15.690 grades, on top of the degree that you meant to get, on top of the apprenticeship 00:23:15.690 --> 00:23:16.680 and so on. 00:23:17.540 --> 00:23:22.560 Now, hands up, who can drive? 00:23:22.560 --> 00:23:26.430 Okay, that's pretty much everyone. 00:23:26.700 --> 00:23:33.640 Keep your hand raised if you believe you are better than the average driver. 00:23:33.640 --> 00:23:37.120 Is that hand going up? 00:23:37.120 --> 00:23:41.200 It's humble man, humble man. 00:23:41.230 --> 00:23:46.320 90% of people believe they are better than the average driver. 00:23:46.450 --> 00:23:49.250 That's mathematically impossible, right? 00:23:49.460 --> 00:23:54.640 Only 50% can be above average, and it's fortunately the other 50% below 00:23:54.640 --> 00:23:55.440 average. 00:23:55.440 --> 00:24:01.920 So as competition is rising, bear that in mind, you know, if you're a student, 00:24:01.920 --> 00:24:03.360 bear that in mind. 00:24:03.360 --> 00:24:09.410 We overestimate our own capabilities and we underestimate the capability 00:24:09.420 --> 00:24:10.640 of others. 00:24:10.640 --> 00:24:16.960 Okay, so this is, yeah, this is kind of a bit depressing, you know. 00:24:16.960 --> 00:24:22.460 If the jobs aren't in the laptop class, part of the laptop class and knowledge 00:24:22.460 --> 00:24:24.610 economy, where are they going to be? 00:24:24.610 --> 00:24:30.470 In other words, how can we find the jobs of the future? 00:24:30.550 --> 00:24:35.010 I think the one way, one way to look at this, the one that makes the most 00:24:35.010 --> 00:24:39.510 sense, is to look at it through a lens of supply and demand. 00:24:39.510 --> 00:24:42.920 I'm sure we've all heard of supply and demand, but just so we're all on 00:24:42.920 --> 00:24:48.050 the same page, if I'm in my kitchen and I'm really thirsty, and someone 00:24:48.050 --> 00:24:52.890 offers me a bottle of water for a thousand pounds, I'm not gonna take that. 00:24:52.890 --> 00:24:57.520 It's a bad bet, it's a bad deal, because the supply of water from my kitchen 00:24:57.620 --> 00:24:58.530 tap is massive. 00:24:58.530 --> 00:25:00.510 I can get it for next to nothing. 00:25:00.510 --> 00:25:06.590 Put me in a desert, though, where my demand is still really high, because 00:25:06.620 --> 00:25:07.640 I'm really thirsty. 00:25:07.670 --> 00:25:14.000 That offer of a thousand pounds for that water starts to look really tempting, 00:25:14.750 --> 00:25:17.720 because the supply of water is so low. 00:25:17.720 --> 00:25:26.710 So what we want to look for are sectors, where the demand for work is massive, 00:25:26.710 --> 00:25:33.490 but the supply of people that can do that work is low. 00:25:33.970 --> 00:25:39.510 So here are some sectors, not an exhaustive list, but just to give you an 00:25:39.510 --> 00:25:40.170 idea. 00:25:40.300 --> 00:25:45.740 The most obvious place where this dynamic is taking place is that we have 00:25:45.740 --> 00:25:50.400 an ageing population, which by definition means the demand for health care 00:25:50.400 --> 00:25:52.110 is massive, right? 00:25:52.120 --> 00:25:57.480 It's increasing over time, because we have an elderly population who want 00:25:57.480 --> 00:26:02.180 health care, but at the same time, the supply of people who can fulfil 00:26:02.360 --> 00:26:08.500 that demand, you know, health care workers, is shrinking, because less people 00:26:08.640 --> 00:26:10.040 of a working age. 00:26:10.250 --> 00:26:16.050 So, see things like medicine and dentistry, I think, is a really good option. 00:26:16.110 --> 00:26:17.210 It already was. 00:26:17.210 --> 00:26:22.090 We have physiotherapists, paramedics, adult social care, and nursing. 00:26:22.110 --> 00:26:25.480 I think these are all good options. 00:26:25.610 --> 00:26:30.550 Sadly, climate change is a problem we're gonna have to deal with, and the 00:26:30.740 --> 00:26:32.820 green transition needs hands. 00:26:32.880 --> 00:26:38.040 So, we'll need solar and wind turbine technicians and engineers. 00:26:38.040 --> 00:26:43.250 We'll need to electrify the grid to move away from fossil fuels. 00:26:43.460 --> 00:26:48.450 We'll also need the retrofitting of buildings and houses, for things like 00:26:48.450 --> 00:26:52.150 heat pumps, air conditioning, and insulation. 00:26:52.150 --> 00:26:58.190 And one thing that's not so, well, one sector that's not as obvious is flood, 00:26:58.200 --> 00:27:05.720 coastal, and drainage engineers, because we're gonna have more extreme weather. 00:27:05.720 --> 00:27:08.120 So, that's climate change. 00:27:08.120 --> 00:27:10.240 Next one, geopolitical tensions. 00:27:10.240 --> 00:27:12.430 Sadly, again, this is on there. 00:27:12.480 --> 00:27:16.530 I'm not about to suggest that everyone go and join the Army or Navy or RAF, 00:27:16.640 --> 00:27:20.310 but again, they'll need engineers, technicians. 00:27:20.450 --> 00:27:23.820 We need to build submarines as well. 00:27:23.920 --> 00:27:27.320 Drones, munitions, and so on. 00:27:27.480 --> 00:27:32.780 Finally, last one, AI and robotics. 00:27:33.230 --> 00:27:37.880 Robotics is about where AI was five years ago. 00:27:37.880 --> 00:27:41.640 It's sort of known to be the next big thing, but people haven't quite caught 00:27:41.640 --> 00:27:42.220 on yet. 00:27:42.220 --> 00:27:47.720 I think becoming a robotics engineer, if that's your jam, that's a very 00:27:47.720 --> 00:27:48.320 good option. 00:27:48.860 --> 00:27:54.790 And AI, like I say, if you believe that you are capable enough to rise above 00:27:54.790 --> 00:27:59.970 the competition and you're really interested in the work, then leveraging 00:28:00.220 --> 00:28:04.540 AI skills to go into something like law, software development, banking and 00:28:04.600 --> 00:28:09.040 finance, so on, is a very good option as well. 00:28:09.050 --> 00:28:14.490 So again, not an exhaustive list, just to give you some ideas. 00:28:14.490 --> 00:28:21.550 We spend 80,000 hours in our careers. 00:28:21.550 --> 00:28:26.680 Picking the right career was already one of the most important decisions 00:28:26.680 --> 00:28:27.710 of our lives. 00:28:27.710 --> 00:28:32.920 AI now makes that decision even more important. 00:28:32.920 --> 00:28:39.800 Taking the time to make even a 1% improvement in that decision can yield 00:28:39.800 --> 00:28:43.440 a massive return on investment. 00:28:43.610 --> 00:28:46.760 So how do you choose well? 00:28:46.790 --> 00:28:50.410 Step one, pick a direction. 00:28:51.340 --> 00:28:58.470 Students should align their careers with their interests, skills, and preferences, 00:28:58.470 --> 00:29:04.040 and they should find the best ways to enter those career paths, whether 00:29:04.310 --> 00:29:08.880 that's through a degree, an apprenticeship, or straight into work. 00:29:08.880 --> 00:29:14.490 But that's not nearly as important as step two, which is developing valuable 00:29:14.490 --> 00:29:15.590 skills. 00:29:15.590 --> 00:29:20.950 You need to understand which parts of your job AI automates and which parts 00:29:21.620 --> 00:29:27.210 remain distinctly human and then focus on the parts that have the highest 00:29:27.240 --> 00:29:29.110 skill ceiling. 00:29:29.140 --> 00:29:37.850 PwC estimates that employees with AI skills earn an average of 60% more 00:29:38.000 --> 00:29:41.360 than employees without AI skills. 00:29:41.390 --> 00:29:48.230 Even if you are on UK minimum wage, that means you're an extra 10,000 pounds 00:29:48.270 --> 00:29:51.090 better off every year after tax. 00:29:51.420 --> 00:29:57.360 That is the value and importance of developing the right skills in the AI 00:29:57.360 --> 00:29:57.800 era. 00:29:58.140 --> 00:30:02.640 And finally, step three, you need backup plans. 00:30:02.670 --> 00:30:08.920 You should develop an A-B-Z career plan in case of economic shocks or extreme 00:30:10.110 --> 00:30:10.500 scenarios. 00:30:10.500 --> 00:30:14.040 So your plan A is your preferred career path. 00:30:14.040 --> 00:30:16.840 It could be somewhat exposed to AI. 00:30:16.840 --> 00:30:22.110 Plan B is a viable alternative you're still very happy with and that is ideally 00:30:22.370 --> 00:30:23.560 less exposed to AI. 00:30:23.560 --> 00:30:26.470 And your plan Z is your lifeboat career. 00:30:26.470 --> 00:30:31.530 So no matter what happens, you've got something and that should be minimally 00:30:31.670 --> 00:30:33.360 exposed to AI. 00:30:33.360 --> 00:30:37.960 I left AI engineering to found stable future. 00:30:37.990 --> 00:30:44.280 We've mapped over 1,000 UK jobs to help young people implement exactly those 00:30:44.670 --> 00:30:45.700 three steps. 00:30:45.700 --> 00:30:48.930 The direction, the skills and the backups. 00:30:49.220 --> 00:30:55.110 So you as a parent can stop worrying, start acting and make sure your child 00:30:55.110 --> 00:30:57.900 is financially stable in the AI era. 00:30:58.070 --> 00:31:04.470 So if you want to put them in the best possible position visit stablefuture.uk. 00:31:04.610 --> 00:31:08.510 Thank you very much for your time.