How Ants Inspired This Founder To Build A Robotic Vacuum Cleaner

[00:00:04]Well, welcome. We're here today. I'm Carrie Dolan from Forbes here with Rodney Brooks. We're in the HP garage, which is the birthplace of Silicon Valley, where Hullet Packard was founded with Bill Hullet and Dave Packard's first invention. Rodney is here today because he's on our list of America's 250 greatest innovators. And he's on that list with a slightly cryptic and interesting description of inspired by ants. He he created the Roomba vacuum cleaner. So, I'm interested to hear about the ants. Anyway, he's a roboticist, a longtime professor at MIT of robotics. And so, Rodney, thanks for coming coming today.

[00:00:40]>> Thanks for having me.

[00:00:41]>> Yeah. So, let's before we talk about your career in robotics, talk about what got you into robots or was that the first thing when you were a kid growing up? Were you just interested in machines or or robot? I mean, did robots exist? And first of all, you grew up in Australia, right?

[00:00:57]>> Yeah, I grew up in Australia. And my my mother got me these two American books, How My Wonder books. One on electricity where I learned to make simple circuits and one on computers and giant brains I think it was called. And it had robots in it, mostly fictional robots. Um, and so I spent from my age six or seven, I spent a lot of my time in the in the back shed as we called it, a garage >> not too different from this garage.

[00:01:26]about this size, one car, >> uh cuz the car had migrated out to a carport and I was I would try to build stuff and >> I was much better. I wasn't really good at mechanical stuff. So, the robots never got very far and I concentrate on building circuits that could play games or learn or do stuff like that out of many very cheap components cuz I didn't have much of a budget. And it was only later um that I managed to build actual robots that worked and rolled around.

[00:01:56]That was in my late teens.

[00:01:58]>> So then did did that lead to as an undergraduate in college did you study engineering or electrical engineering or >> I went to actually it was pretty brand new university that didn't have um much.

[00:02:08]It had a ma a great mathematics department.

[00:02:11]>> I started there in uh beginning of 1972 and the mathematics department was filled with refugees from the Prague spring. fascinating.

[00:02:20]>> And um so I got a classical Eastern European mathematics education, but the university had a mainframe computer with 16 kilobytes of memory. 16 kilobytes and punch cards. And during the week, there were four full-time operators who fed people's jobs in. But one of the professors uh um professor Kowski from uh Czechoslovakia um managed to think somehow that it would be good for me if I had that computer to myself for 12 hours every Sunday. And so every Sunday from 9:00 a.m. to 9:00 p.m. I got to use with another student, I got to have that machine to myself and taught myself computer science and built AI programs >> and like literally taught yourself. No one was I mean did you know how to use >> shut down books?

[00:03:08]>> Okay. Okay. So you'd create programs what would you do on your Sundays like write write programs on?

[00:03:13]>> Well I wrote a whole language an AI based language and then I the best program I built was one that could do um could do integration but not numerical integration. Here's the formula you want to integrate and do that um uh symbolically.

[00:03:31]>> So you were a math major but you had this sort of computer side hustle thing going.

[00:03:35]>> Yeah. And because it was only 16 kilob for those who know I had to reinvent well I heard about virtual memory so I had to put virtual memory on this computer so that I had enough room to do stuff.

[00:03:45]>> Yeah. I mean if you think about the memory that we have like on an iPhone. I mean it's Yeah. blows that 16 kilob thing away right comparable.

[00:03:53]>> Yeah. But that was a long time ago when their computers were a very different I mean they were huge right? They was before PCs. So then you you came to the United States right out of college or >> uh yeah I did a master's degree in Australia. It was a masters by research and it was on machine learning and I >> So they had machine learning.

[00:04:11]>> Well they didn't have I'd read about it so I wrote it.

[00:04:14]>> You created your own masters. Yeah.

[00:04:15]>> It was it was a really bad master thesis. It was incredibly bad.

[00:04:20]One of the examiners said I should fail.

[00:04:22]The other examiner said I should pass.

[00:04:24]Fortunately I passed. But I knew I had to go somewhere else. So I applied um to uh the top well arrogantly to the top places in the US and I happened to get into Stanford >> for a PhD in computer science.

[00:04:39]>> Yeah. And I tell people that I was lucky it was 1977 because by 1979 or so people realized that computer science was important and I wouldn't have got in by then.

[00:04:50]>> You never know with my background you were smart enough to know that this is what you wanted to study and you were fascinated with it. So I mean yeah I mean so at the time in the late '7s computers were these huge huge machines right that it was way before the PC.

[00:05:03]>> Yeah.

[00:05:03]>> So then how did you go from computer science PhD at Stanford to robot professor robotics professor at MIT and a few things in between.

[00:05:12]>> Well when I was at Stanford I was a gopher for another graduate student named Hans Maravic. People may have heard of the Maraveca paradox about um how um moving, grasping is hard, but the intellectual stuff is relatively easy in terms of building AI systems. So I I was an office mate of his. He had a robot that uh he was trying to get across a room. It was a 60t long room. And after everyone else had gone home at night when the main frame wasn't very occupied, I would help him set up and then his robot would move about a yard or so every 15 minutes. And by 6:00 a.m.

[00:05:52]I' got across the room if things were lucky.

[00:05:54]>> And what was the form factor of that robot?

[00:05:55]>> It was a four-w wheeled. It was like a a a card table on wheels. Mhm.

[00:06:00]>> So I'd seen him do that and I I I uh whereas my my PhD was just on pure um machine vision looking at satellite images and trying to interpret what was in the satellite image. And by the way just to show how things different were then during my whole PhD I processed three images. Three frames that was it >> because that's all you could get.

[00:06:22]>> It took weeks on on the main frame to get to each image anyway. So I went to MIT as a postoc in 1981 and joined a robotics group there >> because you got an interest in robotics because of your office mate >> and started doing robotics then came back to MIT on the faculty at Stanford on the faculty and then back to MIT on the P.

[00:06:44]>> Wow. And then along the way you've you've been you've also been an entrepreneur you've co-founded another number of companies. Yeah, actually when when I was on the faculty at Stanford, right as I was leaving, uh we a group of us found a company uh to build AI software uh on the first workstation, Sun Workstations.

[00:07:01]>> Oh, I remember Sun Workstations. Yeah.

[00:07:03]>> Yeah. And so that that that company lasted about 8 years.

[00:07:07]>> Mhm.

[00:07:08]>> And it was actually software for expert systems.

[00:07:12]>> People may not remember, but in the 80s expert systems were the big thing. They were going to take over the world. Um, >> guess that didn't really work out, huh?

[00:07:20]>> They're still around. Okay, they're component, but they're not like everyone thought. And that's a good lesson because there's been wave after wave after wave of it's going to do everything.

[00:07:30]>> And inevitably, we've discovered that there's other things we need besides what >> whatever the hot new thing is. Yeah.

[00:07:37]Forbes mentioned that you were inspired by ants and then you came up with the Roomba. What's the connection there? I was sitting in Thailand um in a very remote part of the country because I had married a classmate from Australia who was a Thai woman and we had kids together. I was with the kids out where she uh grew up and largely everyone was speaking Thai and I wasn't very good at Thai. So, I was sitting around a lot and I was just watching insects do their stuff and ants and it was it was fascinating and then I realized that they were doing a lot of things that much better than our big mobile robots that we had at the time and they're tiny and they you know didn't have 100,000 neurons. So, how were they doing that compared to our big computers that we were trying to get things to move around and they they were feeding they were hiding they were going down holes.

[00:08:26]They're doing all sorts of stuff. So it made me think okay I got to rethink how intelligence might be organized and that got me to thinking about the basic stuff that all animals do upon which for us or some other animals intelligence is built but what's that stuff down below and that's where I spent the next few years working and the Roomba came the way we controlled the room came out of that work.

[00:08:50]>> Got it. So, and then the Roomba has sensors, right, that tell it, okay, you've hit a wall or you're going here and you can control where it goes.

[00:08:57]>> There's a whole bunch of sensors. Um, there's some down the first version, some downward looking infrareds. Um, some touch sensors. We had to look a little out because it >> did not want the room to ever fall downstairs because then there's a heavy thing coming down and >> that work. We had a triply redundant system, one of which didn't even use the onboard microprocessor. Um, so multiple sorts of senses to know what was going on. And the first rooms were wander around randomly. They they would knock and they But if they found dirt, we made them go in a circle cuz >> they really get all that dirt up. Huh.

[00:09:34]Yeah.

[00:09:34]>> But what we didn't understand, we had built a fake um apartment in a highway, you know, to try it out in all the different rooms.

[00:09:43]>> Real people's houses or apartments have a lot of dirt under the bed. So the robot would go under the bed. When we finally got it out, the real customers were selling it. There were complaints because we made it so that when when it filled up its dirt compartment, it would stop.

[00:10:01]>> So we go to the bed, see some dirt or find some dirt, start going round and round with a sensor that was measuring that the bumpiness in the air channel cuz that's how it knew there was dirt.

[00:10:12]Mhm.

[00:10:12]>> go round and round and round and run out run down its battery and then or or either run down its battery or fill up the container and just stop where it was which was in the middle of under the bed and no one could reach that room. So we had to make some changes.

[00:10:26]>> So you basically tell it don't be quite so efficient under the bed.

[00:10:29]>> Uh >> or come back out every once in a while.

[00:10:31]>> Come back out and we and we tripled the size of the dirt bin.

[00:10:35]>> Okay, there you go. So it could catch catch >> cuz in our fake apartment that we were cleaning all the time, we just completely underestimated how dirty some places got.

[00:10:42]>> Yeah. Well, that makes me think about the dust underneath my bed, too. I could use a Roomba for that. That's the I mean, that's the nice thing about the Roomba is they can get under the bed in a way that normal vacuum cleaners don't.

[00:10:52]>> Yeah. And we had to do that because if people don't notice, but in the kitchen, there's a kick space under the against the closet and >> had to get into there. So, I had to go under that kick space. So you couldn't have the Roomba be that very tall, right? Yeah. It had to be Yeah.

[00:11:06]Fascinating. So you co-ounded the Roomba with a couple of other folks from MIT.

[00:11:12]Is that correct?

[00:11:13]>> Yeah. Two two students um um Colin Angel and Helen >> Graham. And you had the the they were originally manufacturing the Roombas in >> Oh, the Roombas came we we released the Rimbas 12 years in. We've been in doing a lot of other stuff.

[00:11:27]>> Right. Right. So you but they came out in what 2002. Right. And the company is now you're not involved with the company.

[00:11:34]>> No, I I stepped off the board back in 2011 >> and it's now owned by the is it did I read correctly that the the folks who were manufacturing the >> Yeah. There was a whole thing where Amazon wanted to buy it for quite a good price but then there was a >> um um you know some worry about >> uh antitrust.

[00:11:56]>> Antitrust and stuff and and it it just >> Yeah. I but I was not involved. So I know none of the details.

[00:12:02]>> Yeah. So you've gone on to do a number of other robot companies. You're doing what you've doing now is called robust.

[00:12:09]>> Robust AI.

[00:12:10]>> Robust AI. What is robust AI doing?

[00:12:12]>> We build um intelligent carts. Sounds incredibly boring, but they're incredibly important. Um during CO was leading up to CO, but during CO we accelerated buying stuff that gets delivered to our house. Mhm.

[00:12:28]>> Those packages we get every day um um put together in the fulfillment center where people put stuff into the box and where do they get the stuff? They get the stuff from shelves. Um Amazon had bought a company uh called uh Kever in Boston in the 2000s where the shelves the robots bring the shelves to the person who does the pick.

[00:12:51]>> That person is called a picker. But that's only a very small percentage of all the warehouses. Most of the warehouses, human pickers go out with carts, pick the stuff, put it into different totes on the cart, and push that heavy cart around, push it to the packout >> place, gets packed out, etc. And those workers are doing over 30,000 steps a day.

[00:13:14]>> Wow. That's a lot.

[00:13:15]>> It's a lot. It's really tough on the body. And and um there's just an incredible labor shortage. people don't like to do it for a long time in general.

[00:13:23]>> So, our robots reduce the number of steps they take by a lot because once the cart is full, it just goes off by itself. And these >> or you can basically just say cart go and it'll just figure out where it knows that it's full.

[00:13:35]>> It knows that every pick's been done.

[00:13:37]And you know these when you when you take off from the airport, you see those buildings, those flat buildings with lots of trucks, they're fulfillment centers, typically a million square ft.

[00:13:46]So you can walk a thousand feet in any direction and often that's what you have to do with the cart. So now the robots just do that. Also the robots often lead the people. They say come over here. you know, >> we already know where this part is that you need to go pick up >> and indicate where the part is. And so it makes it a lot easier than what's happening right now where people have a little tiny green screen emulating a fixed size terminal from the 80s where they have to read a 10 character code match that >> and then figure out what part to get.

[00:14:19]Yeah.

[00:14:19]>> But where we we put the robot right there, we've got an arrow, they pick it, they scan it, and then on the robot it lights up where to put it. So, >> and so you're and you're in this I mean where are you in the company development? Have you released product to customers?

[00:14:32]>> We're all over the country and in Mexico. How long has it been? Several years of sales.

[00:14:39]>> Seven years to get there. People underestimate how long it takes to get hardware really operating. So that this has been our act one.

[00:14:46]>> And do you lease them or sell sell them?

[00:14:48]>> We sell them as a service.

[00:14:49]>> Okay.

[00:14:50]>> Um so and our biggest customer >> So then if you have an upgrade then they get the upgrade and Yeah. Exactly.

[00:14:56]>> Yeah. So we um people probably don't you know wouldn't know this but the biggest operator of third party logistics which is what this is called the some company will operate >> the warehouse for a brand >> right >> and um um the biggest one in the world is DHL uh supply chain >> and so that's our biggest customer but we're also working with a lot of other customers.

[00:15:22]>> Wow. There's um I'd say there's 10 major operators in the US.

[00:15:28]>> And how long did it take to develop this robust AI robot? You said seven years of development.

[00:15:33]>> Seven years since we started.

[00:15:34]>> Yeah. And um and do you have competition? Like are there other >> There are other other companies. Yeah.

[00:15:40]Yeah. But we we of course believe that we are doing things a lot better because we came later. So, we're using more um we're using more of the bounty of the um uh deep learning revolution which has put GPUs. GPUs is what Nvidia has. I I sort of think Nvidia was the luckiest company in the world because they had graphical processing units, GPUs. It was for making games faster and it turns out to work just great for these uh neural net um computations.

[00:16:13]>> And so, everyone wants more GPUs, more GPUs. The older companies that did this before didn't have that. So they use different techniques. We're using GPUs and we can change the models. We're putting in new models every every couple of months and getting a lot better.

[00:16:26]>> Fascinating.

[00:16:27]>> So we can keep going faster.

[00:16:28]>> Fascinating. And it's a private ventureback company.

[00:16:31]>> Yeah, a private venture.

[00:16:32]>> So you when you developed the Roomba, were you were you guys the first one to come up with a robot vacuum?

[00:16:38]>> We were not. Electrolux in um in Europe had had one um called the trilobyte um and it was actually taller than the under parts of the uh kitchen closets and it cost €2,000 um which was a lot of money and it's over well over $2,000 then and um >> which is a lot for a vacuum >> and we knew that that was not going to be a way to sell cuz we had previously had a partnership with Hasbro where we had started building Um, actually it was I could call them humanoids. Ah, my real baby. It was a doll that you could play with and had facial expressions and moods and you fed it and you tickled its feet, etc. So, we built that very low cost um and in China. So, we knew how to do lowcost manufacturing. So, we set out to build a much lower cost than €2,000.

[00:17:28]>> And how did you figure out was the right cost?

[00:17:31]>> Well, we we asked people we asked people we didn't ask them about vacuum cleaners. We asked them at malls how much could they use, how much could they spend on a sort of impulse purchase without checking with their spouse and remain married. And it looked like $200 was a good thing to aim for. So that's what we did and that was the price that we started with. So, we had a spreadsheet with $200 on the bottom right corner and backed out of that, you know, uh, shipping, advertising, profit of the re retailer, uh, went all the way back to the CODs of cost of goods sold COGS before that the bomb bill of materials. So, I said how much we could spend on the parts and it was around $50 out of 200. So, now >> we had these little engineering fights, you know, but I need this part. You can't. Well, that puts a penny over.

[00:18:24]>> We can't got to go down here. Yeah.

[00:18:26]>> Oh my gosh.

[00:18:27]>> And so it was a it was a real that hard line was what made the profit.

[00:18:31]>> Yeah. I mean because if it's too expensive, people weren't going to buy it, right? Yeah. So that's No. And it really was popular. That's great. One of the things that I know you have spoken publicly about I mean we we're now in this period where there's kind of a little bit of a boom in building humanoid robots. Tesla is building them.

[00:18:50]There's a company called Figure AAI.

[00:18:51]There's a bunch of other ones. You've come out and said that you're pretty skeptical about humanoid robots. And maybe you could talk talk a little bit about why.

[00:19:00]>> Yeah. I'll tell you one one other piece of data.

[00:19:02]>> There's 140 Chinese companies building humanoid robots.

[00:19:06]>> Oh, yes. I didn't I didn't talk about those. That's a lot.

[00:19:08]>> That's a lot. Yeah. Um I'm skeptical. I um at MIT once we'd spun out I robot and we were going to be working on the the insect ideas I didn't want my graduate students at MIT to think they were doing stuff for my company. So I switched and in 1992 started building humanoids the first humanoids in the US. So uh we built a series of about seven humanoids starting with one called Cog sort of like a gear and cognition. Um and um then after I left MIT in 2010, I started a humanoid company and I built uh about 4,000 humanoid robots. Built them in in uh New Hampshire and uh Massachusetts and our biggest market was actually in China in factories. They didn't have legs. They were upper torso, but they were meant to do the jobs that uh a human could do with two arms reach. And they had very long arms because we didn't have hips and we use our hips a lot to reach. Uh but the arms were also very um um force sensitive. So uh a person come up just grab the arm at any time, move it out of the way. And I used to test whether we things were safe enough by and everyone knew this which may have been bad bad idea but I used to test the robots as I've been developed by going putting my head in the way and letting the robot hit my head and um so people knew that I was serious about safety. I had to be safe.

[00:20:40]>> You never got any scratches or anything robots. Okay, that's good. That's good.

[00:20:44]They were smart robots.

[00:20:45]>> So I bu I've worked with humanoids a long time. M um and I see the problems the distance still to go to make dextrous robots. Um last October I wrote a 20,000word piece on why current humanoids were not going to learn to be dextrous in my view and it's because they're largely basing it on >> um looking at how people move when they do something.

[00:21:13]>> But that's not what the important thing is. The important thing is it's not the motion. The motion is determined by the forces that people feel and and a lot of the companies are just leaving out force and touch in our hands. We have 18 different families of touch neurons. 18 families of touch neurons in our hands and then we feel forces and without that I think that we're not going to get anything like human dexterity. I mean there's famous experiments where you um you make the just the tips of people's fingers numb and then ask them to do a simple task that they could do 10 seconds before they can't do it anymore.

[00:21:50]>> We don't even realize how all these neurons all this stuff and how this is enabling.

[00:21:54]>> So the good news out of that is a bunch of little companies forming um uh with people working on >> force sensing in in the robot hands and touch sensing and and then all coming to me and saying will you be an adviser? So I'm advising a bunch of them now and it's great.

[00:22:10]>> So basically the thing that you said wasn't being done by the existing human robots. Some companies are doing it.

[00:22:15]>> Yeah. And now and in fairness there were lots of people in academia researching but now this is given a bit of oh well maybe you know this VC says maybe I should fund one of those companies because maybe there'll be something there whereas before it was this remote part of academia that no one was excited about.

[00:22:31]>> But and I think you said it's hard right? It's hard to do this right. I mean, if if it were easy to get this sort of dexterity with the force done, it would have been done already, right?

[00:22:41]>> Well, here's a here's a case where, you know, pe people have been lured into large language models, I think, by the totally surprising facility for language that they have.

[00:22:55]Completely surprising, totally unexpected that it works so well. that we accommodate mistakes they make. If you got a robot, it's not competing or not talking to a person. It's talking to physics. And physics doesn't accommodate. Physics, oh, it's just a robot. I I'll be different. No, it's got to it's there's no room. There's not not any really room for >> error. Everyone wants things to move fast in in in manufacturing or >> wherever. And and even, you know, most of the demos you see, you know, unpacking a a dishwasher, it's the dishwasher has four dishes in it, >> right?

[00:23:37]>> Yeah.

[00:23:37]>> And it's slow and it's and it's you if you if you know what to watch for in the video, do they put it down till the surface touches? No. They drop it just before because surface interaction implies force, which they can't accommodate. So there's a whole >> there some broken dishes if they really >> there's a genre of humanoid robot videos that if you know what to look for, you see the tricks they they played. And I don't I'm not necessarily saying they're they're they're trying to misinform people sort of they sort of believe it themselves. Oh, look, I can get this to work. I can get this to work. But it's not the same as working in in the way people work, >> right? And in a real life situation where you really were emptying a full dishwasher or something like that. Yeah.

[00:24:19]Yeah. I mean, I I'm not holding my breath for having a a robot assistant uh in my home doing making my coffee and emptying my dishwasher in the next couple of years. I mean, yeah, I actually don't mind doing those things myself.

[00:24:30]>> Well, the other thing to remember is, you know, we have traditionally built machines with special forms for special stuff. And, you know, you could have a humanoid go down to the the the stream with the clothes and put them in the water. You know, the the washing machine works pretty damn well. And so you could think about other ways of connecting to it.

[00:24:50]>> Yeah. Yeah. Interesting. Well, it'll be interesting to see how these other companies you're advising progress. So, one of the, you know, one of the things that this, we did this list of 250 greatest innovators tied to America's 250th birthday. And I'm curious, you know, given that you've been in the United States for quite a long time now, what you think the American system has done to enable innovation in the time that you've been here and what the US needs to do going forward to make sure that we have another 250 years of great innovation.

[00:25:20]>> Yeah. The second question is harder.

[00:25:22]>> I can I can look backwards and tell you what has been working and was still working. Um and that is uh and it really started or got really got going after the second world war because during the second world war lots of universities had jumped in to work on all sorts of things uh airborne radar um nuclear weapons etc. And um there was a real um decision uh made at the highest levels to have funding from the US government to research universities. Um and there was always a complaint but it's not fair. These other places don't get it.

[00:26:00]But there was >> like the R1 universities get the funding.

[00:26:03]>> They have the center of excellence and they really push that. And so there's been a tremendous amount of stuff produced over a long amount of time and it takes many many paths that fail. You got to get them you know you have to be prepared for that because you don't know which ones will succeed >> and you learn from failure right.

[00:26:23]>> Yeah. Deep deep learning for instance which is what has enabled the current uh large language models really came of its own in 2012 but it's based on technology that was developed here at Stanford in 1960 by Bernie Widro who passed away last year and then there've been six generations of it at the end of each everyone said ah it doesn't work after all doesn't work after all doesn't work after all and that's what finally worked so you never know continuing iterations so unfortunately what's happened right now is that um there's this big thing, oh these things are working really well.

[00:27:00]We'll just push that way. And we've seen the government there's been a pull down of funding for lots of things to push on AI which you know I'm an AI person so maybe I should be happy about that but to push on AI and quantum >> and other things are getting left behind. There's going to be lots of good ideas in even computer science which is not AI right >> or there were techniques in AI which had were getting close to somewhere which now won't be able to get funding. So the world has changed and it's been pretty drastic and quick in the last uh year and a half.

[00:27:36]>> Mhm.

[00:27:37]>> Um I've always told people that the world keeps changing and that we have to adapt. Um, but I'm not sure that VC money is going to have the patience um for long term because I'm sure that lots of things that are hot in VC world now will fail dramatically >> cuz they haven't even been lab demonstrated, if they're not lab demonstrated, the chance, you know, maybe one in 10 they might get >> a little further, one in a 100 they might get quite a bit further and one in a thousand they'll get along with. And there's not enough VC money to make all those bets, >> right? And the and I think that what VCs are funding is got to be further along than what's done at the basic research level, right? Which is which is what a lot of universities are doing.

[00:28:23]>> And that's not what's happening now. And I'm not I'm not I'm not putting down the VCs. They've they've got a model.

[00:28:28]They're trying to do that. Um uh but I'm I think the system's got a little broken.

[00:28:33]>> Mhm.

[00:28:34]>> More than a little broken. Now, should we go back to exactly what we had before? I don't know. We've never done that before. But we keep changing.

[00:28:41]>> Yeah.

[00:28:42]>> But so I think we're at a the the the eggs are broken. We can't put those eggs back together. Should we just get some new eggs? I'm I'm not sure.

[00:28:51]>> And we'll sort of see how it goes. Well, it'll be interesting to see how as time passes, you know, universities deal with the government for getting more research funding or not, you know. I mean, yeah, some I mean, I've definitely talked with people who feel like there was too much money going to some of these universities for things that weren't important. But I think there's the >> the trouble is you never know what's when the non-important thing is going to be important.

[00:29:14]>> Exactly.

[00:29:15]>> And you know we had just here in this garage we had vacuum tubes which are used for radios and then Huelet and Packard went off and made audio oscillators and and then test equipment for not radio stuff >> that we used for all kinds of things >> and that turned into modern electronics and without modern electronics we wouldn't have silicon electronics we wouldn't have computers. we wouldn't have. So those guys were doing stuff which was sort of why not just build more radios, you know? No, they said we're going to build something different.

[00:29:42]>> You never know where it's going to go.

[00:29:43]Yeah. Fascinating.

[00:29:45]>> Well, Ronnie, thank you so much for your time. It was so great to talk with you.

[00:29:47]I appreciate your you being here.