WEBVTT 00:00:00.000 --> 00:00:03.259 So I want you to picture the classic archetype 00:00:03.259 --> 00:00:06.419 of the mad professor. You know, the exact image, 00:00:06.480 --> 00:00:08.720 I mean. Right. The wild hair, maybe a bit of 00:00:08.720 --> 00:00:11.220 a frantic energy. Yeah, exactly. That kinetic 00:00:11.220 --> 00:00:14.259 energy that makes them kind of hard to keep up 00:00:14.259 --> 00:00:16.679 with. They're usually holding some strange contraption 00:00:16.679 --> 00:00:18.559 that looks like it might explode in a shower 00:00:18.559 --> 00:00:20.820 of sparks. And everyone around them just calls 00:00:20.820 --> 00:00:23.960 them Doc. Yes. It's a trope we see in movies 00:00:23.960 --> 00:00:26.579 constantly, right? From Back to the Future to, 00:00:26.640 --> 00:00:28.899 well, those old black and white monster movies. 00:00:29.120 --> 00:00:32.100 It's the standard shorthand for the genius who 00:00:32.100 --> 00:00:35.200 is too busy thinking about the secrets of the 00:00:35.200 --> 00:00:38.219 universe to worry about social norms. Or combing 00:00:38.219 --> 00:00:40.679 their hair. Exactly. But what if I told you that 00:00:40.679 --> 00:00:43.799 this exact character actually existed? And not 00:00:43.799 --> 00:00:47.340 only did he exist, but he was a calm, critical 00:00:47.340 --> 00:00:49.679 force behind winning World War II. And perhaps 00:00:49.679 --> 00:00:52.039 even more surprisingly, he is the reason machines 00:00:52.039 --> 00:00:54.259 can read your handwriting today. It's just wild. 00:00:54.340 --> 00:00:56.380 Today on The Deep Dive, we are unpacking the 00:00:56.380 --> 00:00:59.759 life and work of Alan William Mark Coombs, known 00:00:59.759 --> 00:01:02.799 to his colleagues, and soon to you, simply as 00:01:02.799 --> 00:01:06.209 Doc. And to guide us through this, we've pulled 00:01:06.209 --> 00:01:10.069 together a stack of biographical records, some 00:01:10.069 --> 00:01:12.629 Wikipedia entries regarding his work at the Post 00:01:12.629 --> 00:01:14.890 Office Research Station. Which is way more exciting 00:01:14.890 --> 00:01:17.530 than it sounds, I promise. It really is. And 00:01:17.530 --> 00:01:19.549 we're also looking at historical accounts of 00:01:19.549 --> 00:01:22.170 the Colossus computer. So the mission today isn't 00:01:22.170 --> 00:01:23.909 just to talk about a guy who built a computer. 00:01:24.430 --> 00:01:26.650 We aren't just, you know, listing out specs and 00:01:26.650 --> 00:01:29.730 dates. No, the goal is to move Alan Coombs out 00:01:29.730 --> 00:01:32.450 of the shadows. He often gets listed as a footnote 00:01:32.450 --> 00:01:34.769 next to bigger names like Alan Turing or Tommy 00:01:34.769 --> 00:01:37.689 Flowers. Right. But we want to explore the bridge 00:01:37.689 --> 00:01:40.549 Coombs built. He is the missing link between 00:01:40.549 --> 00:01:43.609 wartime code breaking and the modern automation 00:01:43.609 --> 00:01:46.390 of information. We're connecting the high stakes 00:01:46.390 --> 00:01:48.909 world of cracking the Enigma and Lorenz ciphers 00:01:48.909 --> 00:01:51.709 to the everyday miracle of a machine reading 00:01:51.709 --> 00:01:54.480 a postcode on a letter. Which is a crazy trajectory 00:01:54.480 --> 00:01:56.680 when you think about it, from saving the free 00:01:56.680 --> 00:01:59.879 world to sorting the mail. Yeah. So let's get 00:01:59.879 --> 00:02:01.920 into it. Let's meet the doc. Let's do it. We 00:02:01.920 --> 00:02:03.599 have to start with the stats just to kind of 00:02:03.599 --> 00:02:06.819 ground us in reality here. Alan Coombs, born 00:02:06.819 --> 00:02:09.479 in Bristol, England in 1911. And looking at his 00:02:09.479 --> 00:02:11.439 early life, you see immediately that he didn't 00:02:11.439 --> 00:02:13.620 just stumble into engineering as a hobby. He 00:02:13.620 --> 00:02:15.939 was an academic heavyweight from the very beginning. 00:02:16.180 --> 00:02:18.319 He went to the University of Glasgow, getting 00:02:18.319 --> 00:02:22.539 his Bachelor of Science in 1932. But he didn't 00:02:22.539 --> 00:02:25.580 stop there. He went on to earn a Ph .D. in 1936. 00:02:25.879 --> 00:02:27.900 Can we just pause on that for a second? A Ph 00:02:27.900 --> 00:02:30.419 .D. in the 1930s? That feels pretty significant. 00:02:30.719 --> 00:02:32.800 Oh, it was incredibly rare compared to today. 00:02:33.060 --> 00:02:35.960 In the 30s, engineering was often viewed as a 00:02:35.960 --> 00:02:38.509 practical trade. You learned by... doing, you 00:02:38.509 --> 00:02:40.629 know, by apprenticing. Right. Getting your hands 00:02:40.629 --> 00:02:43.449 dirty. Exactly. To have a doctorate in it meant 00:02:43.449 --> 00:02:45.530 you were operating at a theoretical level that 00:02:45.530 --> 00:02:47.729 most people just didn't touch. You were dealing 00:02:47.729 --> 00:02:50.169 with the physics of electricity, the deep math 00:02:50.169 --> 00:02:53.169 of signal processing, not just soldering wires. 00:02:53.530 --> 00:02:55.810 Right. And he wasn't done accumulating letters 00:02:55.810 --> 00:02:58.310 after his name either. No. In that same year, 00:02:58.310 --> 00:03:02.270 1932, he also received an ARTC award, which was 00:03:02.270 --> 00:03:05.009 later designated as the ARCST from the University 00:03:05.009 --> 00:03:07.919 of Strathclyde. So by the time he is 25 years 00:03:07.919 --> 00:03:09.919 old, he is highly credentialed. He understands 00:03:09.919 --> 00:03:12.300 the math, the physics, and the mechanics. He's 00:03:12.300 --> 00:03:15.539 a double threat. But here is the contrast that 00:03:15.539 --> 00:03:17.879 I love, and it's what makes him such a great 00:03:17.879 --> 00:03:21.139 character to dive into. Despite all those degrees, 00:03:21.419 --> 00:03:23.639 despite being one of the smartest guys in any 00:03:23.639 --> 00:03:27.080 room, nobody called him Dr. Coombs or Professor. 00:03:27.400 --> 00:03:30.699 No, he was universally known as Doc. Doc. It's 00:03:30.699 --> 00:03:32.960 affectionate, isn't it? Yeah. But it also implies 00:03:32.960 --> 00:03:35.509 a certain kind of personality. It does. And the 00:03:35.509 --> 00:03:37.889 descriptions of him are so vivid. He wasn't your 00:03:37.889 --> 00:03:40.610 stiff academic in a tweed suit. He had this notable 00:03:40.610 --> 00:03:42.810 facial tick. Yeah. And looking at the source 00:03:42.810 --> 00:03:44.530 material, there was a really interesting note 00:03:44.530 --> 00:03:46.830 on this. The biographers mentioned that today 00:03:46.830 --> 00:03:49.909 this behavior would likely be classed under Tourette's 00:03:49.909 --> 00:03:52.849 syndrome. But back in the 30s and 40s, they didn't 00:03:52.849 --> 00:03:54.729 really have that diagnosis handy, or at least 00:03:54.729 --> 00:03:56.569 it wasn't socially understood in the same way. 00:03:56.750 --> 00:03:59.449 Exactly. So to his colleagues, it just became 00:03:59.449 --> 00:04:01.800 part of the package. Combined with his intellect 00:04:01.800 --> 00:04:04.319 and his intense, rapid -fire way of speaking, 00:04:04.560 --> 00:04:07.599 it just contributed to this mad professor vibe. 00:04:07.960 --> 00:04:10.919 It wasn't seen as a disability. It was seen as 00:04:10.919 --> 00:04:13.659 a quirk of high intelligence. It makes him feel 00:04:13.659 --> 00:04:17.019 so real to me. You can just imagine him in the 00:04:17.019 --> 00:04:19.939 lab, maybe twitching slightly, mind racing a 00:04:19.939 --> 00:04:22.439 mile a minute, holding a soldering iron and muttering 00:04:22.439 --> 00:04:24.889 about voltage. And I think acknowledging that 00:04:24.889 --> 00:04:28.009 persona adds a really vital layer of humanity 00:04:28.009 --> 00:04:30.649 to the machines we are going to talk about. We 00:04:30.649 --> 00:04:32.610 often think of the people at Bletchley Park or 00:04:32.610 --> 00:04:35.689 Dollis Hill as these stoic robotic figures. The 00:04:35.689 --> 00:04:38.589 stiff upper lip. Right. But here you have Doc 00:04:38.589 --> 00:04:41.389 with his tics and his brilliance leading a team. 00:04:41.610 --> 00:04:44.709 It suggests that the intense discipline required 00:04:44.709 --> 00:04:47.149 for his work was balanced by a very distinct, 00:04:47.370 --> 00:04:50.410 very human personality. It disrupts the idea 00:04:50.410 --> 00:04:53.279 that you have to be boring to be brilliant. Terrifically 00:04:53.279 --> 00:04:55.519 said. So we have the man. Now let's put him in 00:04:55.519 --> 00:04:57.939 the history books. We are moving the timeline 00:04:57.939 --> 00:05:00.399 to World War II, and we have to talk about his 00:05:00.399 --> 00:05:03.500 workplace, the post office research station at 00:05:03.500 --> 00:05:06.040 Dulles Hill. Now, if you're not British, post 00:05:06.040 --> 00:05:08.040 office research station sounds like a place where 00:05:08.040 --> 00:05:10.279 they, I don't know, test envelope glue or... 00:05:10.589 --> 00:05:13.449 Design new stamps. It sounds incredibly mundane, 00:05:13.529 --> 00:05:16.009 I know. But Dulles Hill was essentially the British 00:05:16.009 --> 00:05:19.529 equivalent of Bell Labs or a secret DARPA facility 00:05:19.529 --> 00:05:22.709 today. It was a hub of high -level electronics 00:05:22.709 --> 00:05:24.889 work. Because the post office ran the telephone 00:05:24.889 --> 00:05:27.490 network. Exactly. So they had the best experts 00:05:27.490 --> 00:05:30.430 on valves, vacuum tubes, and signal processing 00:05:30.430 --> 00:05:33.610 in the entire country. If you wanted to build 00:05:33.610 --> 00:05:36.029 something electronic in the 1940s, this is where 00:05:36.029 --> 00:05:38.269 you went. And this brings us to the big one, 00:05:38.389 --> 00:05:41.720 the Colossus computer. Specifically, the Mark 00:05:41.720 --> 00:05:44.160 II. Now, this is a crucial distinction we need 00:05:44.160 --> 00:05:46.199 to make. Usually when you hear about Colossus, 00:05:46.199 --> 00:05:48.560 you hear the name Tommy Flowers. He designed 00:05:48.560 --> 00:05:51.220 the original Colossus, the Mark I. Which was 00:05:51.220 --> 00:05:54.300 the world's first programmable electronic digital 00:05:54.300 --> 00:05:57.480 computer. It was. But the Mark lie was essentially 00:05:57.480 --> 00:06:00.379 the prototype. It was the proof of concept. It 00:06:00.379 --> 00:06:02.439 proved you could use thousands of vacuum tubes 00:06:02.439 --> 00:06:04.639 to process data without them all burning out 00:06:04.639 --> 00:06:07.399 instantly. Which, by the way, was the prevailing 00:06:07.399 --> 00:06:09.500 wisdom at the time. People thought it couldn't 00:06:09.500 --> 00:06:12.000 be done. But Coombs comes in for the Mark II. 00:06:12.220 --> 00:06:15.300 Yes. Coombs was a principal designer of the Mark 00:06:15.300 --> 00:06:17.860 II, which was the production version of the machine 00:06:17.860 --> 00:06:20.660 used at Bletchley Park for code breaking. What 00:06:20.660 --> 00:06:22.259 was the difference? Was it just a software update? 00:06:22.420 --> 00:06:25.500 Oh, no. It was a massive hardware leap. The Mark 00:06:25.500 --> 00:06:29.459 I had about 1 ,600 valves. The Mark II, which 00:06:29.459 --> 00:06:31.680 Coombs helped engineer and eventually took charge 00:06:31.680 --> 00:06:35.259 of, had 2 ,400 valves. It was five times faster 00:06:35.259 --> 00:06:37.899 than the original. We are talking about a machine 00:06:37.899 --> 00:06:40.740 that could read paper tape at 5 ,000 characters 00:06:40.740 --> 00:06:45.279 per second. 5 ,000 characters per second in 1944. 00:06:45.959 --> 00:06:49.139 That is just mind -boggling. To give you a physical 00:06:49.139 --> 00:06:51.639 sense of that, the tape was moving so fast, about 00:06:51.639 --> 00:06:54.139 30 miles per hour, that the edges of the paper 00:06:54.139 --> 00:06:56.959 tape acted like a saw. If the tape broke and 00:06:56.959 --> 00:06:59.100 hit you, it would cut you open. Okay, that is 00:06:59.100 --> 00:07:01.079 actually terrifying. It was a beast of a machine. 00:07:01.259 --> 00:07:03.800 It generated so much heat from those 2400 valves 00:07:03.800 --> 00:07:05.639 that people worked in their undershirts even 00:07:05.639 --> 00:07:07.980 in the dead of winter. And here is the pivot 00:07:07.980 --> 00:07:10.620 point in the story. Tommy Flowers, the visionary 00:07:10.620 --> 00:07:12.699 behind the concept, moved on to other projects. 00:07:12.920 --> 00:07:16.019 He left the Colossus work. He did. He left to 00:07:16.019 --> 00:07:19.000 focus on other things. And who took over leadership 00:07:19.000 --> 00:07:23.040 of the project? Docums. He became the guy keeping 00:07:23.040 --> 00:07:25.139 these massive heat generating machines running. 00:07:25.199 --> 00:07:27.259 And remember, they weren't just building one. 00:07:27.399 --> 00:07:29.319 They were trying to build a fleet of them to 00:07:29.319 --> 00:07:32.139 defeat Nazi encryption before D -Day. I want 00:07:32.139 --> 00:07:34.480 to pause on the why here. We throw around terms 00:07:34.480 --> 00:07:37.360 like Nazi encryption. Yeah. What were they actually 00:07:37.360 --> 00:07:39.540 trying to do? Because most people know the Enigma 00:07:39.540 --> 00:07:42.360 story, thanks to all the movies. Was this Enigma? 00:07:42.680 --> 00:07:46.079 No, and that's a very common misconception. Enigma 00:07:46.079 --> 00:07:48.620 was used for tactical stuff, tanks talking to 00:07:48.620 --> 00:07:51.600 submarines, that sort of thing. Coombs and Colossus 00:07:51.600 --> 00:07:54.500 were attacking the Lorenz Cipher. Which was different. 00:07:54.720 --> 00:07:57.199 It was even more complex than Enigma. The Lorenz 00:07:57.199 --> 00:07:59.279 was used for high -level communication between 00:07:59.279 --> 00:08:02.420 Hitler and his top generals. It used 12 different 00:08:02.420 --> 00:08:05.110 wheels to scramble the message. And Colossus 00:08:05.110 --> 00:08:07.949 didn't read the message directly. It used statistical 00:08:07.949 --> 00:08:10.310 analysis to find the starting positions of those 00:08:10.310 --> 00:08:12.850 wheels. It was essentially looking for patterns 00:08:12.850 --> 00:08:16.629 in the noise. So Coombs is managing a team that 00:08:16.629 --> 00:08:19.790 is building these massive electronic brains to 00:08:19.790 --> 00:08:23.470 hunt for statistical patterns in German radio 00:08:23.470 --> 00:08:26.089 signals. Precisely. And if they got it right, 00:08:26.269 --> 00:08:28.670 the Allies knew exactly what Hitler was planning 00:08:28.670 --> 00:08:31.290 before his own generals did. I want to capture 00:08:31.290 --> 00:08:33.830 the atmosphere of that time because Coombs actually 00:08:33.830 --> 00:08:36.370 wrote about this later on. And the quote he gave, 00:08:36.370 --> 00:08:38.789 it gives me chills every time I read it. He said, 00:08:38.870 --> 00:08:42.809 no member of our team could ever forget the fellowship, 00:08:42.970 --> 00:08:45.730 the sense of purpose and above all. The breathless 00:08:45.730 --> 00:08:48.090 excitement of those days. Breathless excitement. 00:08:48.149 --> 00:08:51.250 It's so evocative, isn't it? Usually war stories 00:08:51.250 --> 00:08:54.470 are about grim determination or fear or just 00:08:54.470 --> 00:08:56.389 doing your duty. It really is unique. You have 00:08:56.389 --> 00:08:58.590 to remember, this was a secret bunker, essentially. 00:08:58.850 --> 00:09:01.289 They were working on technology that didn't officially 00:09:01.289 --> 00:09:04.830 exist. If they failed, convoys sank. If they 00:09:04.830 --> 00:09:06.850 failed, the invasion of Europe might fail. And 00:09:06.850 --> 00:09:09.669 yet he calls it breathless excitement. It sounds 00:09:09.669 --> 00:09:11.809 like the thrill of discovery. Like they knew 00:09:11.809 --> 00:09:13.529 they were building the future, even if they couldn't 00:09:13.529 --> 00:09:16.190 tell anyone about it. That is the key. They were 00:09:16.190 --> 00:09:18.769 inventing the digital age in the dark. They were 00:09:18.769 --> 00:09:21.289 solving problems no human had ever faced before, 00:09:21.470 --> 00:09:24.529 like how to synchronize 2000 electronic switches 00:09:24.529 --> 00:09:28.389 perfectly. The fellowship he mentions, that camaraderie, 00:09:28.389 --> 00:09:30.789 must have been incredibly tight. They were the 00:09:30.789 --> 00:09:32.350 only people on earth who understood what they 00:09:32.350 --> 00:09:35.039 were doing. And that secrecy, that is the double 00:09:35.039 --> 00:09:37.299 -edged sword of this whole story, because that 00:09:37.299 --> 00:09:40.059 secrecy didn't end when the war did. No. In fact, 00:09:40.080 --> 00:09:42.580 it got tighter. Winston Churchill ordered the 00:09:42.580 --> 00:09:45.559 Colossus machines to be dismantled, destroyed. 00:09:45.700 --> 00:09:48.340 That hurts to hear, just smashing up history. 00:09:48.600 --> 00:09:51.039 It was the dawn of the Cold War. He wanted to 00:09:51.039 --> 00:09:52.779 keep the fact that the British could read these 00:09:52.779 --> 00:09:55.320 codes a secret from the Soviets, who were using 00:09:55.320 --> 00:09:58.100 similar technology. So for decades, the work 00:09:58.100 --> 00:10:01.000 Coombs and his team did was classified, top secret. 00:10:01.259 --> 00:10:03.740 Imagine building the first computer. Literally 00:10:03.740 --> 00:10:06.000 changing the course of human history. And not 00:10:06.000 --> 00:10:08.500 being able to tell anyone for 30 years. Not your 00:10:08.500 --> 00:10:11.899 wife. Not your kids. It is almost unimaginable 00:10:11.899 --> 00:10:13.919 today where we post our achievements instantly. 00:10:14.440 --> 00:10:17.100 But that was the reality. Coombs went back to 00:10:17.100 --> 00:10:19.340 work and the world went on thinking the Americans 00:10:19.340 --> 00:10:22.779 invented the computer with Enon S in 1946. But 00:10:22.779 --> 00:10:25.360 Coombs knew better. He knew. But he kept his 00:10:25.360 --> 00:10:29.159 oath. This silence lasted until 1976. Enter Professor 00:10:29.159 --> 00:10:32.220 Brian Randell. This is the big turning point. 00:10:32.279 --> 00:10:35.340 Right. So jump forward to the mid 70s. Brian 00:10:35.340 --> 00:10:37.659 Randall is a computer scientist researching the 00:10:37.659 --> 00:10:40.659 history of the field. And he keeps hearing whispers, 00:10:40.679 --> 00:10:43.440 rumors about a British machine. He starts digging 00:10:43.440 --> 00:10:45.200 around because he's prepping for a conference 00:10:45.200 --> 00:10:47.559 at the Los Alamos National Laboratory in New 00:10:47.559 --> 00:10:50.340 Mexico. The home of the atomic bomb. Kind of 00:10:50.340 --> 00:10:52.519 a fitting place for revealing massive government 00:10:52.519 --> 00:10:55.759 secrets. Exactly. And he manages to get something 00:10:55.759 --> 00:10:58.669 incredible. the British government finally decides 00:10:58.669 --> 00:11:00.950 to loosen the grip. Just a little bit. Just a 00:11:00.950 --> 00:11:03.809 crack. In late 1975, they release a series of 00:11:03.809 --> 00:11:05.509 captioned photographs from the public record 00:11:05.509 --> 00:11:08.190 office, and Randall obtained permission to present 00:11:08.190 --> 00:11:10.789 a paper on the wartime development of the colossi 00:11:10.789 --> 00:11:13.470 at the post office research station. So picture 00:11:13.470 --> 00:11:18.250 the scene. It's 1976, Los Alamos. A room full 00:11:18.250 --> 00:11:21.629 of top -tier scientists and historians. Randall 00:11:21.629 --> 00:11:25.080 presents this paper. And the room just explodes. 00:11:25.440 --> 00:11:27.700 It was a total rock star moment for engineers. 00:11:27.899 --> 00:11:29.620 You have to understand, these people thought 00:11:29.620 --> 00:11:31.659 they knew the history of their field inside and 00:11:31.659 --> 00:11:33.899 out. Suddenly, Randell is standing there saying, 00:11:34.059 --> 00:11:36.639 no, actually, there was a programmable electronic 00:11:36.639 --> 00:11:39.899 computer two years before ENIAC. The interest 00:11:39.899 --> 00:11:41.940 was so high that they actually had to organize 00:11:41.940 --> 00:11:44.799 a special evening meeting just to handle all 00:11:44.799 --> 00:11:46.840 the questions. And guess who was there to answer 00:11:46.840 --> 00:11:49.179 them? Doc Coombs and Brian Randell. That has 00:11:49.179 --> 00:11:52.159 to be one of the most satisfying moments. I would 00:11:52.159 --> 00:11:54.639 imagine so, standing there, finally able to say, 00:11:54.799 --> 00:11:57.940 yes, we did this. Yes, it worked. Randall later 00:11:57.940 --> 00:11:59.899 published an article called The First Electronic 00:11:59.899 --> 00:12:02.320 Computer. This was the moment the history books 00:12:02.320 --> 00:12:04.460 were fundamentally rewritten. And Coombs was 00:12:04.460 --> 00:12:07.539 finally able to bask in that revelation. After 00:12:07.539 --> 00:12:10.120 30 years of holding his tongue, he could finally 00:12:10.120 --> 00:12:12.940 explain what that breathless excitement was all 00:12:12.940 --> 00:12:16.059 about. It's a powerful validation for him. But 00:12:16.059 --> 00:12:18.240 what I find even more interesting, and this is 00:12:18.240 --> 00:12:20.100 where this deep dive really gets into the weeds, 00:12:20.220 --> 00:12:22.919 is that Coombs didn't just retire after the war. 00:12:23.139 --> 00:12:25.500 He didn't just sit around waiting for the Secrecy 00:12:25.500 --> 00:12:28.919 Act to expire. No, he kept building. And this 00:12:28.919 --> 00:12:30.740 is where the story gets really surprising for 00:12:30.740 --> 00:12:34.019 me. We go from cracking Nazi codes to sorting 00:12:34.019 --> 00:12:36.240 the mail. It sounds like a massive downgrade, 00:12:36.259 --> 00:12:38.200 doesn't it? Like going from being James Bond 00:12:38.200 --> 00:12:41.179 to working at the DMV. It really does. Yeah. 00:12:41.259 --> 00:12:44.669 Why did he make that shift? Well, because technologically 00:12:44.669 --> 00:12:47.529 it was actually a massive leap forward. OK, let's 00:12:47.529 --> 00:12:50.389 unpack that. The war ends. Coombs is still at 00:12:50.389 --> 00:12:52.830 Dulles Hill. First, he worked on something called 00:12:52.830 --> 00:12:55.769 the Amosaic Computer. It stands for Ministry 00:12:55.769 --> 00:12:58.029 of Supply Automatic Integrator and Computer. 00:12:58.309 --> 00:13:00.450 It's a bit of a neglected fragment of history, 00:13:00.629 --> 00:13:02.789 but it was mainly used for calculating radar 00:13:02.789 --> 00:13:05.730 data and aircraft trajectories. So it shows he 00:13:05.730 --> 00:13:07.509 stayed on the cutting edge of pure computing. 00:13:07.929 --> 00:13:10.590 But then comes the big shift. Yes. He takes over 00:13:10.590 --> 00:13:13.909 the scientific side of R -14. R -14. Sounds like 00:13:13.909 --> 00:13:16.490 a droid from Star Wars. It was the division working 00:13:16.490 --> 00:13:19.629 on postal mechanization. Specifically, optical 00:13:19.629 --> 00:13:22.610 character recognition, or OCR. Okay, so the problem 00:13:22.610 --> 00:13:25.870 is people write addresses on envelopes, and sorting 00:13:25.870 --> 00:13:28.509 them by hand is just way too slow. Precisely. 00:13:29.019 --> 00:13:31.820 The post office was drowning in mail. They desperately 00:13:31.820 --> 00:13:34.519 needed a machine to read the envelope and sort 00:13:34.519 --> 00:13:37.399 it into the right bin automatically. Which honestly 00:13:37.399 --> 00:13:39.940 sounds harder than cracking Enigma. Have you 00:13:39.940 --> 00:13:41.580 seen some people's handwriting? Oh, absolutely. 00:13:41.879 --> 00:13:43.840 My doctor's handwriting is basically encryption. 00:13:44.490 --> 00:13:47.149 It is exponentially more difficult in many ways, 00:13:47.309 --> 00:13:49.570 and that is the connection here. If you think 00:13:49.570 --> 00:13:52.029 about it, Coombs went from looking for patterns 00:13:52.029 --> 00:13:54.629 in encrypted data, identifying the statistical 00:13:54.629 --> 00:13:57.490 anomalies in the Lurin cipher, to looking for 00:13:57.490 --> 00:14:00.350 patterns in human squiggles on an envelope. Oh, 00:14:00.350 --> 00:14:02.889 wow. It's all pattern recognition. Exactly. In 00:14:02.889 --> 00:14:05.070 crypto, the noise is the encryption algorithm 00:14:05.070 --> 00:14:08.629 hiding the message. In OCR, the noise is the 00:14:08.629 --> 00:14:11.409 messy ink, the bad handwriting, the coffee stain 00:14:11.409 --> 00:14:13.590 on the envelope. Coombs had to figure out how 00:14:13.590 --> 00:14:15.419 to teach. machine to look at a squiggle and say, 00:14:15.539 --> 00:14:18.259 that is statistically likely to be the letter 00:14:18.259 --> 00:14:21.279 A. And he's doing this in the 1950s. There are 00:14:21.279 --> 00:14:24.200 no neural networks, no AI as we know it today. 00:14:24.340 --> 00:14:27.460 No cloud computing, just hardwired logic and 00:14:27.460 --> 00:14:30.419 analog circuits. He moved to the new BT Research 00:14:30.419 --> 00:14:33.679 Center at Martlesham in Suffolk and led the work 00:14:33.679 --> 00:14:35.759 that resulted in early postcode reading machines. 00:14:36.259 --> 00:14:38.879 He had to figure out how to interpret lines and 00:14:38.879 --> 00:14:41.960 curves electronically. So every time a machine 00:14:41.960 --> 00:14:44.860 automatically sorts my mail today, or I scan 00:14:44.860 --> 00:14:47.240 a document on my phone and it magically turns 00:14:47.240 --> 00:14:49.759 into text, I can trace a little bit of that back 00:14:49.759 --> 00:14:52.320 to Doc Coombs. You certainly can. He pioneered 00:14:52.320 --> 00:14:54.539 the theory of how computers interact with the 00:14:54.539 --> 00:14:57.919 messy, unstructured real world. He moved computing 00:14:57.919 --> 00:15:00.360 from calculation, just adding numbers to perception. 00:15:00.679 --> 00:15:02.039 And he wasn't just building these things in a 00:15:02.039 --> 00:15:04.240 dark room. He was teaching people how they worked. 00:15:04.840 --> 00:15:07.139 Which brings us to the flash gun and the caltrop. 00:15:07.220 --> 00:15:09.559 Yes. Coombs was a frequent lecturer, and true 00:15:09.559 --> 00:15:12.419 to his doc persona, he was not boring. He wanted 00:15:12.419 --> 00:15:15.019 people to understand how we see so we could understand 00:15:15.019 --> 00:15:17.559 how machines might see. I really wish I had a 00:15:17.559 --> 00:15:20.220 professor like this. Describe his favorite trick 00:15:20.220 --> 00:15:23.220 for us. So he wanted to demonstrate feature detection 00:15:23.220 --> 00:15:26.600 in the human visual system. To do this, he would 00:15:26.600 --> 00:15:30.039 use a flash gun. basically a high intensity strobe 00:15:30.039 --> 00:15:32.379 light like you'd see on an old camera, but much, 00:15:32.460 --> 00:15:34.700 much more powerful. I can see the mad professor 00:15:34.700 --> 00:15:37.840 vibes returning in full force. Oh, yeah. He would 00:15:37.840 --> 00:15:40.720 darken the room, flash this intense light. And 00:15:40.720 --> 00:15:42.940 because of the persistence of vision in the human 00:15:42.940 --> 00:15:45.519 eye, the way an image lingers on your retina 00:15:45.519 --> 00:15:48.039 for a split second, the audience would literally 00:15:48.039 --> 00:15:51.740 see a character like a letter or a shape disintegrate 00:15:51.740 --> 00:15:54.500 fragment by fragment. That is so cool. So he's 00:15:54.500 --> 00:15:56.659 breaking down vision into bits. He was proving 00:15:56.659 --> 00:15:59.129 that our eyes don't. take a static photo, they 00:15:59.129 --> 00:16:02.809 process features, lines, curves, angles. And 00:16:02.809 --> 00:16:04.950 he used this to explain a concept he called the 00:16:04.950 --> 00:16:07.970 caltrop. The caltrop? Like the spiky weapon you 00:16:07.970 --> 00:16:10.470 throw on the ground to pop tires? Yes. The shape 00:16:10.470 --> 00:16:13.190 is the key there. A caltrop has four points. 00:16:13.210 --> 00:16:15.629 No matter how you drop it, one point is always 00:16:15.629 --> 00:16:18.389 sticking up. Coombs used this physical object 00:16:18.389 --> 00:16:21.169 to explain multidimensional space. Okay, you're 00:16:21.169 --> 00:16:22.809 going to have to explain that to me like I'm 00:16:22.809 --> 00:16:25.669 five. Because multidimensional space sounds like 00:16:25.669 --> 00:16:28.860 pure science fiction. Fair enough. Imagine every 00:16:28.860 --> 00:16:31.240 letter isn't just a picture, but a collection 00:16:31.240 --> 00:16:34.159 of scores. A score for curviness, a score for 00:16:34.159 --> 00:16:36.279 straight lines, a score for holes in the shape. 00:16:36.539 --> 00:16:39.570 If you map those scores in a 3D graph, All the 00:16:39.570 --> 00:16:42.029 A's cluster in one area and all the B's cluster 00:16:42.029 --> 00:16:44.590 in another area. It's about mapping features 00:16:44.590 --> 00:16:47.669 in space. So he was trying to teach people that 00:16:47.669 --> 00:16:50.029 machines could see by converting shapes into 00:16:50.029 --> 00:16:52.389 math. Exactly. He was deconstructing vision to 00:16:52.389 --> 00:16:54.269 reconstruct it digitally. He's trying to get 00:16:54.269 --> 00:16:56.230 people to visualize what we now call a feature 00:16:56.230 --> 00:16:59.169 space, which is totally fundamental to how modern 00:16:59.169 --> 00:17:01.529 AI works. It's amazing to think about him standing 00:17:01.529 --> 00:17:03.610 in front of a chalkboard, flashing lights, twitching 00:17:03.610 --> 00:17:05.829 with excitement, trying to explain to a room 00:17:05.829 --> 00:17:07.990 of confused people how a machine could learn 00:17:07.990 --> 00:17:10.740 to see. And doing it decades before computer 00:17:10.740 --> 00:17:13.579 vision became a buzzword in Silicon Valley, he 00:17:13.579 --> 00:17:16.279 was laying the theoretical groundwork for everything 00:17:16.279 --> 00:17:19.319 from facial recognition to self -driving cars. 00:17:19.599 --> 00:17:22.099 So what does this all mean for us? We've gone 00:17:22.099 --> 00:17:24.420 from a Ph .D. in Glasgow to the high -stakes 00:17:24.420 --> 00:17:27.000 secrecy of Bletchley Park to the flashing lights 00:17:27.000 --> 00:17:29.000 of lecture halls and automated mail sorters. 00:17:29.289 --> 00:17:32.509 Sadly, Alan Coombs passed away on January 30th, 00:17:32.509 --> 00:17:36.950 1995 in Yelton, Devon. He was 83 years old. But 00:17:36.950 --> 00:17:39.250 he left quite a mark on the world. He represents 00:17:39.250 --> 00:17:41.769 the unsung technical brilliance that underpins 00:17:41.769 --> 00:17:44.509 our modern life. We love the stories of the lone 00:17:44.509 --> 00:17:47.609 genius, the Turing, the Einstein, but Coombs 00:17:47.609 --> 00:17:49.720 was the guy who made things work. He didn't just 00:17:49.720 --> 00:17:52.480 build the hardware with Colossus. He pioneered 00:17:52.480 --> 00:17:54.859 the theory of pattern recognition. He bridged 00:17:54.859 --> 00:17:57.079 the gap between a machine that calculates numbers 00:17:57.079 --> 00:17:59.740 and a machine that actually understands the world 00:17:59.740 --> 00:18:03.140 around it. Exactly. Without that bridge, without 00:18:03.140 --> 00:18:06.039 his work on OCR and pattern recognition, our 00:18:06.039 --> 00:18:07.940 digital world would look very, very different. 00:18:08.099 --> 00:18:10.440 He helped computers open their eyes. I want to 00:18:10.440 --> 00:18:12.339 leave you with a thought to mull over. We often 00:18:12.339 --> 00:18:15.359 think of computer vision or AI reading text as 00:18:15.359 --> 00:18:17.519 a modern invention, something from the last 10 00:18:17.519 --> 00:18:20.190 or 20 years. But consider Doc Coombs in that 00:18:20.190 --> 00:18:23.490 bunker in 1944, or in his lab in the 50s. It 00:18:23.490 --> 00:18:26.049 raises a fascinating question. Was that breathless 00:18:26.049 --> 00:18:28.210 excitement he felt just about winning the war? 00:18:28.869 --> 00:18:31.390 Or did he already see a future where machines 00:18:31.390 --> 00:18:34.950 could read, see, and understand us? Maybe the 00:18:34.950 --> 00:18:37.289 mad professor wasn't mad at all. Maybe he just 00:18:37.289 --> 00:18:39.569 saw the future before the rest of us did. I think 00:18:39.569 --> 00:18:42.029 he certainly did. Thanks for listening to this 00:18:42.029 --> 00:18:44.509 deep dive into the life of Alan Coombs. It's 00:18:44.509 --> 00:18:45.910 been a pleasure. See you next time.