WEBVTT - How long will Moore's Law continue boosting computing power? (featuring Dr. Adam Becker) 0:00:07.840 --> 0:00:09.800 I feel like an old man whenever I buy a 0:00:09.800 --> 0:00:12.960 new computer. I mean, why does my laptop need sixty 0:00:12.960 --> 0:00:15.680 four gigabytes of memory? When I learned to program on 0:00:15.720 --> 0:00:19.960 a PC that had twenty kilobytes? No joke. Kids these 0:00:20.040 --> 0:00:22.799 days don't understand how hard we had it back in 0:00:22.800 --> 0:00:26.159 the day. Right, But it's also a nice feeling. It 0:00:26.239 --> 0:00:29.560 tells me that we're making progress and that's good. It's 0:00:29.640 --> 0:00:32.479 creating new worlds and new ways of life. It's literally 0:00:32.600 --> 0:00:37.040 saving lives by accelerating science. That's all great stuff, right, 0:00:37.440 --> 0:00:40.319 But how long can it go on? What is the 0:00:40.400 --> 0:00:44.319 engine of this exponential growth in computing power? And can 0:00:44.400 --> 0:00:46.680 we count on it to take us to the stars, 0:00:46.720 --> 0:00:51.159 to cure cancer and to develop self driving toothbrushes. Today 0:00:51.280 --> 0:00:54.120 we'll dive into the physics underlying this trend and ask 0:00:54.600 --> 0:00:57.360 whether there are fundamental limits that could block us from 0:00:57.440 --> 0:01:00.680 achieving our dreams. And we'll talk about whether there's danger 0:01:00.760 --> 0:01:05.200 and assuming technology will solve all of our problems. Welcome 0:01:05.240 --> 0:01:08.000 to Daniel and Kelly's Extraordinary Universe. 0:01:21.319 --> 0:01:25.319 Hello. I am Kelly Wiener Smith. I study parasites and space, 0:01:25.560 --> 0:01:27.720 and I realized when we were starting to do this 0:01:27.760 --> 0:01:30.200 episode that I wasn't one hundred percent clear on what 0:01:30.240 --> 0:01:31.480 mores Law meant exactly. 0:01:31.880 --> 0:01:35.000 Hi, I'm Daniel. I'm a particle physicist and I've been 0:01:35.040 --> 0:01:39.360 programming computers for more than forty years. They get faster 0:01:39.560 --> 0:01:40.560 and I get slower. 0:01:40.760 --> 0:01:44.560 Oh you're not slowing down yet, Daniel, you stab. 0:01:45.600 --> 0:01:48.520 So my question for you today, Kelly, is what was 0:01:48.560 --> 0:01:51.920 your first computer? Let's age Kelly. 0:01:52.080 --> 0:01:55.080 Okay, So later when we talked to Adam Becker in 0:01:55.120 --> 0:01:57.160 our interview, he mentions that there was a while there 0:01:57.200 --> 0:01:59.600 where folks wouldn't get a computer because you'd wait as 0:01:59.680 --> 0:02:01.880 long as you could, because the computers kept getting so 0:02:02.000 --> 0:02:05.320 much better so quickly that if you could wait, your 0:02:05.320 --> 0:02:07.600 computer would be much better. Yeah, and so my family 0:02:07.640 --> 0:02:11.200 waited way too long. We didn't get one until I 0:02:11.320 --> 0:02:15.040 was in like high school, and I know, and I 0:02:15.040 --> 0:02:17.600 don't even remember what it was. But in the meantime 0:02:17.639 --> 0:02:19.520 I had to write my essays on like it was 0:02:19.560 --> 0:02:21.919 like a brother typewriter, but it also had a little 0:02:22.000 --> 0:02:25.960 electronic screen, and so I could very slowly and laboriously 0:02:26.560 --> 0:02:29.120 click through my essays and then I would print it 0:02:29.160 --> 0:02:30.880 and something would be wrong. It would take me forever 0:02:30.960 --> 0:02:33.680 to find where the error was. It was very annoying. 0:02:34.120 --> 0:02:36.519 But what about you, did you have like the first 0:02:36.560 --> 0:02:37.600 Apple computer? Ever? 0:02:38.000 --> 0:02:38.200 Oh? 0:02:38.320 --> 0:02:41.639 Apple was way too advanced. I go way before that. 0:02:41.760 --> 0:02:45.359 What My first computer was a Commodore VIC twenty, which 0:02:45.400 --> 0:02:48.480 I think had twenty kilobytes of RAM, and we stored 0:02:48.520 --> 0:02:51.280 stuff on an audio tape, you know, like you write 0:02:51.320 --> 0:02:54.040 a little program and then you'd stored on these cassette 0:02:54.080 --> 0:02:56.000 tapes that you could later listen to and like, ooh 0:02:56.040 --> 0:02:59.960 what does that sound? So yeah, we were very very early. 0:03:00.120 --> 0:03:02.200 In fact, I remember hanging out with my dad in 0:03:02.240 --> 0:03:04.880 grad school while he was doing his research and he 0:03:04.919 --> 0:03:08.400 was literally feeding punch cards into those punch cards machines. 0:03:08.480 --> 0:03:12.000 So I feel like I've personally experienced a huge fraction 0:03:12.200 --> 0:03:16.200 of the transformation of computers into the basically supercomputers we 0:03:16.240 --> 0:03:19.320 have today. I mean, my smartphone is so much more 0:03:19.360 --> 0:03:23.040 powerful than anything my dad ever used in his research. 0:03:23.320 --> 0:03:25.560 That is absolutely amazing. I didn't even know that we 0:03:25.560 --> 0:03:28.760 were storing data on like cassette tapes. Oh yeah, that's 0:03:28.840 --> 0:03:29.520 amazing to me. 0:03:29.720 --> 0:03:31.640 Yeah, before magnetic floppies for sure. 0:03:31.760 --> 0:03:34.280 So when you like are using your are you a 0:03:34.320 --> 0:03:34.960 MacBook guy? 0:03:35.160 --> 0:03:35.720 I am? Yeah. 0:03:35.760 --> 0:03:37.560 It's like when you're using your Mac Do you every 0:03:37.640 --> 0:03:39.680 day think I am so lucky I'm not doing this 0:03:39.760 --> 0:03:42.080 on punch cards or you just do you take it 0:03:42.080 --> 0:03:42.839 for granted? Now? 0:03:43.280 --> 0:03:46.400 I think it's awesome. It's incredible. I mean every time 0:03:46.440 --> 0:03:49.040 I get a new MacBook, I'm like, while, this drive 0:03:49.160 --> 0:03:51.600 is ten times bigger than anything I've ever seen, and 0:03:52.040 --> 0:03:55.320 the memory is just shocking. And it's also then incredible 0:03:55.320 --> 0:03:59.200 to me how rapidly our computational ambitions grow. You know, 0:03:59.280 --> 0:04:01.920 my group does a lot of computation, and we're basically 0:04:02.000 --> 0:04:05.440 limited by computation, and so every time we get more 0:04:05.440 --> 0:04:08.880 powerful computers, we scale up our ambitions and solve bigger, 0:04:08.880 --> 0:04:12.360 harder problems, and so we're always at the edge of 0:04:12.360 --> 0:04:14.480 what the computers can do, right, Like, we have an 0:04:14.480 --> 0:04:18.080 infinite number of questions we could ask with harder computers. 0:04:18.279 --> 0:04:21.760 So yeah, I'm in awe of the MacBook, not just 0:04:21.839 --> 0:04:24.080 because it's so much more powerful than anything I've used, 0:04:24.120 --> 0:04:26.599 but it's so reliable. I mean, I spend hours and 0:04:26.640 --> 0:04:28.440 hours a day in front of this thing. It almost 0:04:28.520 --> 0:04:32.960 never gives me problems. So yeah, it's incredible. What engineers 0:04:32.960 --> 0:04:34.720 have provided it is incredible. 0:04:34.800 --> 0:04:37.000 And today we're going to talk about one way in 0:04:37.040 --> 0:04:40.279 which that incredible ability has been expanded, which has to 0:04:40.279 --> 0:04:43.000 do with Moore's Law, which I thought was about how 0:04:43.080 --> 0:04:46.440 much data you can store on your computer, and I 0:04:46.440 --> 0:04:47.520 think it's much more than that. 0:04:47.960 --> 0:04:50.120 You're right, it's much more than that. It's also about 0:04:50.120 --> 0:04:52.800 how things are growing over time and how long that 0:04:52.880 --> 0:04:57.200 will continue. And there's this lore about Moore's Law which 0:04:57.279 --> 0:05:00.600 is permeated Silicon Valley and broader culture. So in a minute, 0:05:00.600 --> 0:05:03.359 we're gonna also talk to Adam Becker about how this 0:05:03.400 --> 0:05:07.760 has impacted philosophy and politics and policy and how it 0:05:07.839 --> 0:05:10.520 might affect our future. But first we wanted to know 0:05:10.640 --> 0:05:13.760 how long people thought Moore's Law might continue to make 0:05:13.839 --> 0:05:16.760 all of our computers faster. So I went out there 0:05:16.800 --> 0:05:19.480 and I talked to our group of volunteers. Here's what 0:05:19.520 --> 0:05:22.279 they had to say about the future of Moore's law. 0:05:22.920 --> 0:05:26.040 So I'll give it six years and then I'm gonna 0:05:26.120 --> 0:05:32.119 sell my nvidious stock. But quand computing will change the game. 0:05:32.680 --> 0:05:35.960 We are restricted by things like housmow. 0:05:36.000 --> 0:05:40.240 We can make stuff, so I think it's not true anymore. 0:05:40.800 --> 0:05:44.120 My first thought was to sign no, but I know 0:05:44.480 --> 0:05:48.440 very little bit quantuine computing. I would say until the 0:05:48.440 --> 0:05:52.000 computer speed, which just be the light maybe ten years, 0:05:52.320 --> 0:05:53.880 another good decade or so. 0:05:54.680 --> 0:05:56.440 My understanding is that it's already done. 0:05:56.560 --> 0:06:01.480 I don't think we are doubling in raw processing speed. 0:06:01.839 --> 0:06:05.400 In my understanding, Moore's law kind of slowed down for 0:06:05.600 --> 0:06:08.720 laptop desktop chips a number of years back, but has 0:06:08.800 --> 0:06:11.960 continued with mobile just because they were a little behind. 0:06:12.240 --> 0:06:15.440 But then you also have graphics and neural processing units 0:06:15.520 --> 0:06:19.120 that power our AI platforms of today. So can we 0:06:19.160 --> 0:06:21.440 go into the future. I think we can go for 0:06:21.520 --> 0:06:26.039 a number more years, given the innovations in transistor stacking. 0:06:26.400 --> 0:06:29.400 I thought Moore's law had to do with cost decreasing 0:06:29.480 --> 0:06:32.200 as speed increased. We do seem to be close to 0:06:32.200 --> 0:06:34.840 a tipping point with electrons being too large for the 0:06:35.240 --> 0:06:39.680 tiny circuitry. However, it seems that optical circuitry might be 0:06:39.720 --> 0:06:40.840 a good replacement for that. 0:06:41.279 --> 0:06:47.360 They're currently reaching the lower limits of workability before they 0:06:47.360 --> 0:06:50.719 start reaching quantum effects. With silicon. 0:06:51.240 --> 0:06:54.760 We are getting closer to the particle level and that 0:06:55.640 --> 0:06:56.320 stops us. 0:06:56.440 --> 0:06:59.400 Honestly, I thought it had already stopped. So do you 0:06:59.440 --> 0:07:01.919 think these are optimistic or pessimistic? 0:07:02.279 --> 0:07:04.919 I mean I think they're realistic, which seems to be 0:07:04.920 --> 0:07:07.800 the option that always gets left out. So I think 0:07:07.839 --> 0:07:09.880 there were a lot of people who said, you know, 0:07:09.920 --> 0:07:11.760 I thought we've already reached the limits. So we're getting 0:07:11.760 --> 0:07:13.880 close to reaching the limits. And I'll admit that I 0:07:14.040 --> 0:07:16.320 did not realize we were getting close to reaching the limits. 0:07:16.320 --> 0:07:18.239 But it seems like a lot of our listeners are 0:07:18.600 --> 0:07:19.720 on top of this trend. 0:07:20.040 --> 0:07:22.560 Yeah, and so we're not giving financial advice. So I 0:07:22.600 --> 0:07:24.440 won't tell you whether or not to buy or sell 0:07:24.440 --> 0:07:29.000 in Vidia stock. But you know, sometimes I wonder about 0:07:29.040 --> 0:07:31.559 these tech companies because their stocks also seem to follow 0:07:31.560 --> 0:07:34.840 Moore's law, Like, how can Google just keep getting more valuable? 0:07:35.400 --> 0:07:37.240 I keep missing out on buying Google. 0:07:38.000 --> 0:07:39.960 I can tell you that if you can make a 0:07:40.000 --> 0:07:41.880 time machine, one of the first things you should do 0:07:42.000 --> 0:07:43.880 is go buy in VideA and Google stuff. 0:07:44.360 --> 0:07:46.720 All right, So let's dig into it. What in the 0:07:46.880 --> 0:07:51.920 end is Moore's law? So Moore's law was something postulated 0:07:52.000 --> 0:07:53.000 by Gordon Moore. 0:07:53.320 --> 0:07:54.600 He has a first name. 0:07:58.480 --> 0:08:01.520 And he was one of the found of Intel, so 0:08:01.760 --> 0:08:05.400 a big dude in like, you know, semiconductors and electronics. 0:08:06.000 --> 0:08:10.720 And he suggested initially that the number of transistors you 0:08:10.760 --> 0:08:14.360 could squeeze onto a chip would double every year. And 0:08:14.400 --> 0:08:16.280 it's a little bit more complicated than that. He also 0:08:16.360 --> 0:08:18.720 was talking about the power usage and the cost, but 0:08:19.080 --> 0:08:22.720 roughly speaking, he was talking about the density of transistors 0:08:22.720 --> 0:08:26.200 on chips getting higher every single year, which means the 0:08:26.200 --> 0:08:29.040 speed of these computers is growing very quickly. 0:08:29.520 --> 0:08:33.160 So transistors are about speed and not storage, or are 0:08:33.160 --> 0:08:33.920 they about both? 0:08:34.120 --> 0:08:37.880 They're about both. The transistor fundamentally is a tiny programmable switch. 0:08:38.559 --> 0:08:41.640 And the reason that computers got small and got fast 0:08:41.800 --> 0:08:44.400 is because we were able to make transistors small and 0:08:44.480 --> 0:08:47.000 make them fast, which allows us to have lots and 0:08:47.040 --> 0:08:49.720 lots and lots of switches in a small area, which 0:08:49.760 --> 0:08:51.680 is what allows the computer to be complex and to 0:08:51.760 --> 0:08:54.840 be fast. And so essentially it's saying we can make 0:08:54.880 --> 0:08:58.680 computers denser every year, and that makes computers faster and 0:08:58.720 --> 0:08:59.360 more powerful. 0:08:59.559 --> 0:09:02.079 Okay, so this has to do with why we went 0:09:02.080 --> 0:09:05.160 from computers that took up entire rooms to something you 0:09:05.160 --> 0:09:06.720 can now stick in your bag and take with you. 0:09:06.960 --> 0:09:09.520 Yeah, exactly, And we'll dig into that in a minute. 0:09:09.800 --> 0:09:12.600 But the history here is that in nineteen sixty five 0:09:13.160 --> 0:09:16.080 More predicted this and then ten years later he revised it. 0:09:16.080 --> 0:09:18.280 He was like, well every year, maybe that's too optimistic. 0:09:18.360 --> 0:09:21.280 Let's go for every two years. And so that was 0:09:21.320 --> 0:09:24.440 the prediction in seventy five, and you'll see that it 0:09:24.480 --> 0:09:27.640 mostly held up until fairly recently. It's sort of an 0:09:27.679 --> 0:09:31.040 extraordinary prediction in that sense, though. You know, anytime there's 0:09:31.040 --> 0:09:33.360 a prediction that holds up, you got to wonder, like, well, 0:09:33.400 --> 0:09:36.160 what were the other predictions this person made, Like you 0:09:36.200 --> 0:09:40.160 just spew predictions constantly, Eventually you're goind of get one Ris. 0:09:39.800 --> 0:09:42.280 Yeah, yeah, well, especially he gave himself another decade to 0:09:42.360 --> 0:09:45.600 like fit the trend line. That was pretty generous to himself. 0:09:45.679 --> 0:09:48.319 But okay, exactly, So let's dick into what a transistor 0:09:48.480 --> 0:09:51.600 is and why it allows computers to be faster, because 0:09:51.640 --> 0:09:54.760 that's crucial to understand why More's law has worked, how 0:09:54.800 --> 0:09:56.679 we've made it work, and whether it's going to work 0:09:56.679 --> 0:10:00.200 in the future. Basically, a transistor is a programmable switch 0:10:00.559 --> 0:10:03.880 like computers operate on digital logic. I have a number 0:10:03.880 --> 0:10:07.280 in the computer, the number four. They store it in binary. 0:10:07.720 --> 0:10:10.520 But to store things in binary you need a physical system. 0:10:10.640 --> 0:10:13.319 But that can store a zero or a one right 0:10:13.360 --> 0:10:15.160 the way you can like write a digit on a 0:10:15.160 --> 0:10:18.080 piece of paper. That's like I'm representing the number four 0:10:18.160 --> 0:10:22.120 by scratching this graphite onto this sheet of paper. I 0:10:22.120 --> 0:10:25.080 want to store things on my computer and zeros and 0:10:25.160 --> 0:10:28.960 ones because binary is the code for computers, and physically 0:10:28.960 --> 0:10:31.440 that means a switch, you know, as you can imagine 0:10:31.720 --> 0:10:34.640 either just like literally like a light switch, but here 0:10:34.679 --> 0:10:36.280 we're doing an electronic switch. 0:10:36.600 --> 0:10:40.000 Okay, And so just to if we switched from using 0:10:40.559 --> 0:10:45.439 transistors to things like DNA to store data or quantum computing, 0:10:46.360 --> 0:10:48.720 could you still apply More's law, like if we switched 0:10:48.840 --> 0:10:52.240 to some other method or is More's law specifically about 0:10:52.720 --> 0:10:55.440 the transistors that we're talking about now, Yeah. 0:10:55.400 --> 0:10:58.360 Great question. You're talking about fundamental changes in how we 0:10:58.400 --> 0:11:02.120 do computing. So currently computing operates on bits, zeros and ones, 0:11:02.360 --> 0:11:05.079 and we're saying those are represented by transistors, which is 0:11:05.120 --> 0:11:07.760 like a physical implementation of that bit. You switch to 0:11:07.880 --> 0:11:11.120 quantum computing, the fundamental element of that is a cubit, 0:11:11.480 --> 0:11:14.160 which isn't necessarily a zero one. It has the probability 0:11:14.160 --> 0:11:16.400 to be in several different states, and so it requires 0:11:16.520 --> 0:11:18.800 a different physical system to model that. We don't use 0:11:18.840 --> 0:11:23.400 transistors or not even like quantum transistors. In fact, transistors 0:11:23.440 --> 0:11:27.480 are already relying deeply on quantum mechanics, so quantum transistor 0:11:27.559 --> 0:11:30.560 is redundant. But yeah, cubit, there's no guarantee that you 0:11:30.600 --> 0:11:33.280 can like build cubits and then build them more densely 0:11:33.320 --> 0:11:36.280 and more rapidly. There's certainly no More's law for quantum 0:11:36.320 --> 0:11:40.520 computing that's a guarantee. And biological computing like DNA is 0:11:40.600 --> 0:11:43.240 super awesome as an idea, but there you have like 0:11:43.400 --> 0:11:47.200 four possibilities, right, DNA is basically base four, and so 0:11:47.320 --> 0:11:49.880 it's a question of like how do you encode numbers 0:11:49.920 --> 0:11:52.040 into DNA? Do you use all four bases? Do you 0:11:52.160 --> 0:11:55.319 group them into two to make binary? The technology is 0:11:55.320 --> 0:11:58.760 fundamentally different, So again you wouldn't expect necessarily further to 0:11:58.800 --> 0:12:00.719 be more solved, but you might get some other law 0:12:00.720 --> 0:12:03.559 which could be better. So but yeah, Mores law reflects 0:12:03.640 --> 0:12:07.120 the details of the technology we're using to represent the 0:12:07.120 --> 0:12:10.160 fundamental element of computing, which is a zero or one, 0:12:10.240 --> 0:12:13.720 and then crucially the logic that operates on those zeros 0:12:13.720 --> 0:12:14.160 and ones. 0:12:14.400 --> 0:12:15.400 Let's get into that logic. 0:12:15.600 --> 0:12:17.800 Yeah, because what you want to do is represent like 0:12:17.880 --> 0:12:19.720 numbers in your computer. I want to put the number 0:12:19.800 --> 0:12:22.120 four in but also I want to calculate stuff. I 0:12:22.120 --> 0:12:23.959 don't just want to write four into my computer. I 0:12:24.040 --> 0:12:25.920 want to be able to add four to two. I 0:12:25.920 --> 0:12:28.520 want to be able to compare four and seven. Right, 0:12:28.559 --> 0:12:30.760 That's what allows you to program a computer for it 0:12:30.800 --> 0:12:34.120 to do useful computation. And if you know something about computing, 0:12:34.120 --> 0:12:36.560 you know like the basics of computation is a Turing 0:12:36.600 --> 0:12:39.959 machine which can like read in numbers and write numbers 0:12:40.000 --> 0:12:44.240 onto this infinite tape. And so in order to do logic, 0:12:44.360 --> 0:12:46.480 you need to be able to have things that respond 0:12:46.559 --> 0:12:50.559 to different inputs. So in logic you have things like gates. 0:12:50.600 --> 0:12:52.640 Like a not gate is something which if you give 0:12:52.640 --> 0:12:54.520 it to zero, it responds a one. If you give 0:12:54.520 --> 0:12:56.600 it a one, it responds to zero. It's like a 0:12:56.679 --> 0:13:00.880 logical map from inputs to outputs, or an a gate. Right, 0:13:00.920 --> 0:13:03.240 an and gate gives you a one if both inputs 0:13:03.280 --> 0:13:06.640 are one and a zero otherwise. Or the converse of 0:13:06.679 --> 0:13:10.240 that is a nand gate NA n D, which is 0:13:10.280 --> 0:13:12.680 the combination of an N gate and a knot gate. 0:13:13.320 --> 0:13:15.600 And the really cool thing is that if you can 0:13:15.640 --> 0:13:18.760 build a nand gate, you can build any logical map 0:13:19.280 --> 0:13:22.200 nandgates are like the basis function of logic. So if 0:13:22.240 --> 0:13:24.560 you have nands, people have shown that you can build 0:13:24.640 --> 0:13:27.760 any map from inputs to outputs and essentially any sort 0:13:27.760 --> 0:13:31.800 of computer logic. So you can build knot gates and 0:13:31.800 --> 0:13:35.440 and gates out of transistors. Transistors are like this digital switch, 0:13:35.840 --> 0:13:37.400 and we'll go into the detail of the physics of 0:13:37.440 --> 0:13:40.120 how they work, but essentially they're a programmable switch. You 0:13:40.160 --> 0:13:42.760 can turn them on or off in response to other stuff. 0:13:43.160 --> 0:13:45.320 So from that you can build logic, and from that 0:13:45.360 --> 0:13:47.400 you can build nandgates, and from that you can build 0:13:47.559 --> 0:13:51.440 literally anything like adders and comparitors and anything you need 0:13:51.559 --> 0:13:54.840 in computers. So this is like the basically the smallest 0:13:54.840 --> 0:13:59.199 little lego brick of computing is a switch, a programmable 0:13:59.240 --> 0:14:00.880 switch that goes from zero to one. And that's what 0:14:00.920 --> 0:14:03.720 a transistor is an implementation of. And it didn't have 0:14:03.760 --> 0:14:05.640 to be a transistor. It could have been something else. 0:14:05.679 --> 0:14:07.520 It could have been DNA, it could have been whatever. 0:14:07.920 --> 0:14:10.760 But this is like the best, fastest, smallest thing that 0:14:10.800 --> 0:14:14.400 we have invented, and this is what revolutionized our society. 0:14:14.640 --> 0:14:16.080 What does the transistor look like? 0:14:17.000 --> 0:14:19.160 Yeah, what does the transistor look like? It looks like 0:14:19.200 --> 0:14:22.520 nothing because it's super duper tiny, right, Like the ones 0:14:22.520 --> 0:14:26.560 that we're building these days are order nanometers right, so 0:14:26.680 --> 0:14:28.880 like you put one on your finger, you can't see it. 0:14:29.320 --> 0:14:32.640 The number of transistors on a typical chip is billions, 0:14:33.080 --> 0:14:35.000 so you can't see an individual one. There used to 0:14:35.040 --> 0:14:36.720 be able to, like when they were first building them 0:14:36.720 --> 0:14:39.360 in the fifties, you would like make one, you know, 0:14:39.960 --> 0:14:41.880 on a bench. You could think of it as sort 0:14:41.880 --> 0:14:44.720 of like three wires coming together. You have a source, 0:14:45.160 --> 0:14:48.400 a drain, and then a gait and the gate basically 0:14:48.440 --> 0:14:51.080 decides do I connect the source and the drain Do 0:14:51.160 --> 0:14:54.040 I open or close this switch? And so it's sort 0:14:54.040 --> 0:14:55.840 of like a wire with a lever in it that 0:14:55.880 --> 0:14:58.320 you know you can open or close, and then another 0:14:58.400 --> 0:15:01.720 wire that determines what whether or not that's open or closed. 0:15:02.160 --> 0:15:04.960 So that's not a physical description of what they look like. 0:15:05.560 --> 0:15:07.760 We can get into like the semiconductors in a minute, 0:15:07.800 --> 0:15:10.200 but that's sort of the logical construction. And when I 0:15:10.200 --> 0:15:12.720 think about more's low, I think, well, what is exactly 0:15:12.800 --> 0:15:16.920 is the connection between more transistors and speed? Like it's 0:15:17.000 --> 0:15:19.080 cool to have things small because then you can put 0:15:19.080 --> 0:15:22.840 a computer in your watch or whatever. But why do 0:15:22.960 --> 0:15:26.440 smaller computers operate faster? Because that's really the crucial key. 0:15:26.480 --> 0:15:29.120 When you sit down at your laptop, you're not like, wow, 0:15:29.240 --> 0:15:33.600 the transistors are super dense. You're like, wow, you know 0:15:33.720 --> 0:15:36.360 word opened in a bill a second instead of you know, 0:15:36.440 --> 0:15:39.880 spinning my beach ball forever. Yeah, So it's the speed 0:15:39.920 --> 0:15:42.880 that's really crucial, and that's really transformed society. Right, it's 0:15:42.920 --> 0:15:49.040 computational power and miniaturization means faster operation for a few reasons. 0:15:49.520 --> 0:15:52.000 Number one, things just don't have to go as far. Right, 0:15:52.080 --> 0:15:55.520 Electronics is limited by the speed of light. It's not instantaneous. 0:15:55.760 --> 0:15:58.920 You close a switch, the electrons don't move instantly. Right, 0:15:58.920 --> 0:16:02.360 the current doesn't change instantly, and so we are still 0:16:02.400 --> 0:16:05.080 limited by the speed of light. And so if the 0:16:05.160 --> 0:16:08.560 distances between the transistors are smaller and the transistors themselves 0:16:08.640 --> 0:16:11.720 are smaller, things just happen faster because there's a speed 0:16:11.760 --> 0:16:14.120 limit to information in the universe. 0:16:14.480 --> 0:16:17.560 That's awesome. I guess I hadn't imagined that as a 0:16:17.640 --> 0:16:20.520 limiting factor. Okay, super cool, what's next. 0:16:20.720 --> 0:16:23.720 Yeah, that's one. The other is you can have wider 0:16:23.800 --> 0:16:26.880 data paths, like instead of just using thirty two bits 0:16:26.920 --> 0:16:29.840 to store your numbers. You can use sixty four bits, right. 0:16:30.120 --> 0:16:32.840 Remember bits are this essential element of binary numbers, and 0:16:32.880 --> 0:16:35.400 so if you have like a two bit number, you 0:16:35.400 --> 0:16:38.040 can only store between zero and four. If you have 0:16:38.080 --> 0:16:40.600 an eight bit number, you can store many more numbers. 0:16:40.600 --> 0:16:42.880 You have thirty two these days computing a sixty four 0:16:42.920 --> 0:16:44.920 one hundred and twenty eight bit If you hear about 0:16:44.960 --> 0:16:47.600 these numbers as the sort of the core the computing 0:16:47.760 --> 0:16:51.040 of your CPU or your operating system, that's what it describes, 0:16:51.080 --> 0:16:54.920 like what size numbers are we operating on? And this 0:16:55.000 --> 0:16:57.920 is important because basically it's how much your computer can 0:16:57.960 --> 0:17:01.040 do in parallel. Like if you can add two hundred 0:17:01.040 --> 0:17:03.400 and twenty eight bit numbers, it's really one hundred and 0:17:03.440 --> 0:17:07.560 twenty eight bit wise operations done in parallel instead of 0:17:07.840 --> 0:17:10.360 if you're doing sixty four bit numbers, then you're only 0:17:10.359 --> 0:17:14.040 doing sixty four operations in parallel, and so you can 0:17:14.080 --> 0:17:16.920 do more operations in parallel, You can pass more data 0:17:16.960 --> 0:17:20.720 at the same time, and so data flows more quickly. 0:17:20.920 --> 0:17:23.439 Another thing that really limits the speed of computers is 0:17:23.720 --> 0:17:26.000 how long does it take to get the data into 0:17:26.080 --> 0:17:29.440 the actual CPU. Right, Like you have these numbers in memory, 0:17:29.440 --> 0:17:31.200 you want to do some calculation on you got to 0:17:31.240 --> 0:17:34.560 slurp them from memory and put them into the registers 0:17:34.560 --> 0:17:37.119 in your CPU. They're actually doing the comparisons or the 0:17:37.160 --> 0:17:41.080 adding or the subtracting or whatever. And so the wider 0:17:41.080 --> 0:17:43.080 the data path, the faster the data gets loaded, and 0:17:43.160 --> 0:17:44.720 the faster the computation happens. 0:17:45.119 --> 0:17:53.040 And CPU probably means senorebditis pyro wetting underwater. What does 0:17:53.080 --> 0:17:53.760 CPU mean? 0:17:54.280 --> 0:17:57.239 CPU means central processing unit. It's a thing on your 0:17:57.240 --> 0:18:00.920 computer that does the actual crunching, does the adding or 0:18:00.960 --> 0:18:04.440 subtracting or comparing, or loading or unloading or writing to 0:18:04.560 --> 0:18:08.200 memorates the closest thing we have to a digital brain. Okay, 0:18:08.280 --> 0:18:11.040 but there's another sort of mechanical element to like why 0:18:11.200 --> 0:18:14.280 speed means faster computers. And you know, back in the 0:18:14.320 --> 0:18:17.600 nineteen fifties, people were doing electronics, and they're doing it 0:18:17.680 --> 0:18:19.600 sort of the way you might do it in your garage. 0:18:19.800 --> 0:18:22.520 You got resistors, you got capacitors, you solder them together, 0:18:22.600 --> 0:18:25.480 you make these big sort of physical circuits. But in 0:18:25.520 --> 0:18:29.399 the late nineteen fifties people invented what's called the integrated circuit. 0:18:29.800 --> 0:18:32.240 Integrated circuit is just like you know, it's a big 0:18:32.400 --> 0:18:35.240 green board and it's got the whole circuit printed onto it. 0:18:35.280 --> 0:18:38.560 You don't like solder the components together, and this really 0:18:38.600 --> 0:18:41.640 allows for like the embedding of these transistors and other 0:18:41.680 --> 0:18:46.760 components inside these protective layers, which enhance their reliability. And 0:18:46.800 --> 0:18:49.080 so that means you can make them smaller, you can 0:18:49.119 --> 0:18:51.439 make more complex circuits that you didn't have to like 0:18:51.480 --> 0:18:55.800 wire together yourself with dripping hot bits of solder, and 0:18:55.880 --> 0:18:57.679 so this makes them more reliable, so you don't need 0:18:57.720 --> 0:19:00.960 as much error correction, et cetera. And that allows things 0:19:01.000 --> 0:19:03.920 to be smaller and to be faster. So you've got 0:19:04.000 --> 0:19:07.200 integrated circuits, you got wider data paths, you got shorter 0:19:07.320 --> 0:19:10.159 distances to travel, and you have faster switching. All these 0:19:10.160 --> 0:19:13.160 things are why more transistors means faster computing. 0:19:13.520 --> 0:19:16.399 Okay, And so when did we get our first transistor? 0:19:16.520 --> 0:19:19.080 Yeah, so the transistor was invented in Bell Labs in 0:19:19.200 --> 0:19:22.320 nineteen forty seven, I think it was. And there was 0:19:22.400 --> 0:19:25.040 a lot of research in the forties different kinds of 0:19:25.040 --> 0:19:28.560 technologies for transistors. Try this, try that, try the other thing. 0:19:28.600 --> 0:19:31.080 But the basic concept was invented in the late forties 0:19:31.080 --> 0:19:33.480 in Bell Labs and you know, Bell Labs is one 0:19:33.480 --> 0:19:37.480 of these like elements of another era, an institution that 0:19:37.560 --> 0:19:40.280 I really miss. You know, it's a privately funded research 0:19:40.400 --> 0:19:43.080 lab that did basic research. You know, this is an 0:19:43.200 --> 0:19:46.399 arm of the telephone company. But they just like gave 0:19:46.600 --> 0:19:50.680 nerds money and said, hey, play around, figure stuff out, 0:19:50.880 --> 0:19:53.720 and they came up with things like the transistor, which 0:19:53.760 --> 0:19:57.760 is I think a trillion dollar idea would be underestimating it, right, Like, 0:19:57.880 --> 0:20:01.760 it's literally the foundation of our entire economy. Its transformed 0:20:01.760 --> 0:20:02.560 the way we live. 0:20:02.760 --> 0:20:03.119 Wow. 0:20:03.359 --> 0:20:06.679 And so I think even if every other piece of 0:20:06.720 --> 0:20:10.439 science was a waste of money, this one brings the 0:20:10.600 --> 0:20:15.040 average up like this one idea like means all of 0:20:15.080 --> 0:20:19.200 science has been worthwhile just from a purely economical, cynical 0:20:19.240 --> 0:20:22.280 point of view. And that's the way science works, right, 0:20:22.359 --> 0:20:26.040 like a lot of fuzzes out and occasionally a huge, 0:20:26.640 --> 0:20:30.240