WEBVTT - Borg Chef, Part II 0:00:00.160 --> 0:00:07.160 Brought to you by Toyota. Let's go places. Welcome to 0:00:07.360 --> 0:00:15.079 Forward Thinking. Hey there, and welcome to Forward Thinking, the 0:00:15.240 --> 0:00:19.320 podcast that looks at the future and says, yummy, yummy, yummy. 0:00:19.360 --> 0:00:22.720 I got love in my tummy. I'm Jonathan Strickland, I'm 0:00:22.760 --> 0:00:25.400 Lauren Bam, and I'm Joe McCormick. I thought you were 0:00:25.400 --> 0:00:27.840 going to do there's a party in my tummy. No, 0:00:28.240 --> 0:00:31.120 I don't know that one. Nobody knows this. There's a 0:00:31.160 --> 0:00:34.239 party in my tummy. Yummy, yummy, yummy, yummy. No, No, 0:00:34.400 --> 0:00:37.440 I I have no no clue what you're going on about. 0:00:37.600 --> 0:00:43.080 We must move on. Today. We're gonna talk about robots 0:00:43.320 --> 0:00:46.600 and cooking more because we did this before, right, We 0:00:46.640 --> 0:00:48.879 did this back in August two fifteen. We did an 0:00:48.920 --> 0:00:52.240 episode called The Borg Chef, and in that we talked 0:00:52.240 --> 0:00:54.800 about AI, we talked about robots, we talked about a 0:00:54.800 --> 0:00:57.800 couple of different approaches people were thinking about. But there's 0:00:57.840 --> 0:01:01.400 been more since then that I thought would be fun 0:01:01.440 --> 0:01:03.640 to kind of revisit and talk about, maybe even get 0:01:03.640 --> 0:01:08.720 a little debate going about the value of a robot 0:01:08.760 --> 0:01:11.440 in the place of a chef or um, you know, 0:01:12.120 --> 0:01:16.160 is the is there enough value to uh to to 0:01:16.400 --> 0:01:20.800 combat the cost of these um doesn't make sense. So 0:01:21.000 --> 0:01:24.560 let's talk a little bit, uh backtrack a bit. Well, 0:01:24.600 --> 0:01:28.040 we've we mentioned in the previous episode that kitchen automation 0:01:28.080 --> 0:01:30.840 is one of those things that's been going on for decades. Uh, 0:01:31.040 --> 0:01:35.760 from everything from pretty simple stuff dishwashers toasters, uh, you know, 0:01:36.000 --> 0:01:40.160 electric can openers up to digital ovens where you can 0:01:40.160 --> 0:01:42.880 program things so that they preheed at a certain time, 0:01:43.160 --> 0:01:45.080 that kind of stuff. And then there are all those 0:01:45.720 --> 0:01:48.760 very specific little appliances you know, the bread and machine 0:01:49.120 --> 0:01:51.120 stuff like. Yeah, a lot of stuff where we have 0:01:51.200 --> 0:01:57.680 tried very hard to make it easier, less time consuming, 0:01:58.160 --> 0:02:01.520 less labor intensive. Right, anything we can do to make 0:02:01.680 --> 0:02:06.760 make the process of cooking more accessible and and less 0:02:06.800 --> 0:02:10.400 of a requirement on our energy and time. UM. That's 0:02:10.440 --> 0:02:14.680 been the case for for decades, but we are really 0:02:14.720 --> 0:02:19.760 seeing a push for that to go into overdrive, particularly 0:02:19.800 --> 0:02:25.040 with home automation and robotics being the two big areas. Also, 0:02:25.200 --> 0:02:27.840 I would argue apps play a big part in that too, 0:02:27.919 --> 0:02:31.239 being incorporated into every device you can imagine at this point, 0:02:31.320 --> 0:02:33.920 like refrigerators and stoves and that kind of stuff. I 0:02:33.960 --> 0:02:36.280 also think some of this demand is simply based on 0:02:36.320 --> 0:02:39.880 people not understanding how bad robots are it cooking. Yeah, 0:02:40.280 --> 0:02:43.560 I can see that. Well. I mean, I think part 0:02:43.600 --> 0:02:45.640 of it is just, you know, all it takes is 0:02:45.680 --> 0:02:49.760 that one day where you've had a really hard day 0:02:49.800 --> 0:02:53.440 at work and it took forever to get home because 0:02:53.480 --> 0:02:56.680 traffic is terrible, and you walk through the door and 0:02:56.720 --> 0:02:59.960 you're hungry, but you have no energy left, and you're thinking, 0:03:00.120 --> 0:03:01.560 what am I going to do? I don't want to 0:03:01.560 --> 0:03:04.440 go out, I don't want to order food. If only 0:03:04.480 --> 0:03:06.560 I had a robo chef, I could just push a 0:03:06.600 --> 0:03:10.920 button and say, you know, make me fedacchini alfredo or 0:03:11.000 --> 0:03:12.680 something along those lines, and it would just do it 0:03:12.760 --> 0:03:15.400 for you. Wouldn't that be lovely? Wouldn't that be awesome? 0:03:16.040 --> 0:03:18.800 And the problem is that, yeah, that would be awesome, 0:03:19.200 --> 0:03:21.519 But we're not there yet. We're not at a point 0:03:21.600 --> 0:03:25.520 where where robo chef can jump in and take over. Um. 0:03:25.560 --> 0:03:28.880 But well not if you care about it being cost 0:03:28.919 --> 0:03:33.600 effective and good. Right, if you wanted to be expensive 0:03:33.600 --> 0:03:37.880 and bad, we got you covered. But uh yeah. And 0:03:38.240 --> 0:03:39.640 one of the reasons why I wanted to bring this 0:03:39.720 --> 0:03:42.640 up as I had read a book review in Popular 0:03:42.680 --> 0:03:45.560 Science as a book review for the book Unnaturally Delicious, 0:03:45.560 --> 0:03:48.840 written by Jason Lusk, and Lusk is a professor of 0:03:48.880 --> 0:03:54.400 agricultural economics at Oklahoma State University. Now, this this book 0:03:54.400 --> 0:03:57.320 actually covers a ton of different stuff, not just robotics 0:03:57.360 --> 0:04:00.160 in the kitchen. That was some of the apparent only 0:04:00.160 --> 0:04:02.160 the early chapters had some of that. A lot of 0:04:02.200 --> 0:04:05.120 it had things to do with like genetically modified organisms 0:04:05.120 --> 0:04:09.360 that kind of stuff. Um and uh, and I don't 0:04:09.400 --> 0:04:13.520 actually agree with all of Lusk's points on everything that 0:04:13.520 --> 0:04:16.080 that he argued, but the robot stuff was kind of fun. 0:04:16.160 --> 0:04:19.120 He talks about going to a restaurant with his wife 0:04:19.120 --> 0:04:21.640 and they were it was actually one of Bobby Flay's restaurants, 0:04:22.240 --> 0:04:25.479 and uh, ordering an appetizer, and they loved the appetizers 0:04:25.560 --> 0:04:30.479 so much that they bought the cookbook for the restaurant 0:04:30.320 --> 0:04:33.640 at which included the recipe for the the appetizer they loved. 0:04:34.000 --> 0:04:36.040 Then they went back home and they tried to make 0:04:36.040 --> 0:04:39.159 the appetizer. And first off, he said, the amount of 0:04:39.200 --> 0:04:42.080 time that it suggested it would take to make the 0:04:42.120 --> 0:04:45.960 appetizer was orders of magnitude off, Like, it was way 0:04:46.040 --> 0:04:49.920 less time than what actually took. Sure, I'll save my 0:04:50.000 --> 0:04:52.720 comments for the end of the story, okay uh, and 0:04:52.760 --> 0:04:56.240 then he says, well, then we go through. We follow 0:04:56.279 --> 0:04:59.560 the instructions, were done with everything, but I just wasn't 0:04:59.560 --> 0:05:03.600 the same. It wasn't as good as when we went 0:05:03.640 --> 0:05:07.120 into the restaurant and had it there. It was something 0:05:07.160 --> 0:05:09.560 was missing, and they weren't sure if they messed up 0:05:09.560 --> 0:05:12.400 something with the recipe, if maybe they left something cooking 0:05:12.440 --> 0:05:15.360 too long or didn't cook it long enough. There was 0:05:15.400 --> 0:05:18.200 no way for them to really know short of retracing 0:05:18.240 --> 0:05:21.360 every step they had taken up through preparing that appetizer. 0:05:22.120 --> 0:05:23.560 So he said, what if we were able to have 0:05:23.640 --> 0:05:29.599 a robot UH that could prepare a dish exactly the 0:05:29.640 --> 0:05:33.120 way a chef prepares it. So the robot uses machine 0:05:33.200 --> 0:05:36.840 learning to study the way a chef prepares a dish 0:05:37.160 --> 0:05:41.440 and can, without fail, replicate that same sequence to the 0:05:41.440 --> 0:05:44.720 point where it's using the the perfect temperature, it's using 0:05:44.720 --> 0:05:48.560 the perfect motion to uh to fold ingredients together or 0:05:48.600 --> 0:05:51.279 stir them or whisk them or whatever, so that you 0:05:51.279 --> 0:05:56.040 would have consistent results every single time. That was that 0:05:56.120 --> 0:05:58.640 was the dream. So you'd have a not just a 0:05:58.760 --> 0:06:05.120 robot chef, but SpeI typically a robot Bobby flay. Yeah, uh, 0:06:05.320 --> 0:06:10.280 that sounds terrifying. But but okay, my my other whole, my, 0:06:10.640 --> 0:06:12.800 I just wanted to put a tiny caveat into this 0:06:12.839 --> 0:06:16.560 story here that you know, it's it's assuming, it's assuming 0:06:16.600 --> 0:06:20.400 a lot of things about a recipe to to say 0:06:20.400 --> 0:06:23.400 that a robot chef would be able to duplicate a 0:06:23.440 --> 0:06:25.600 recipe perfectly. And you know, if if if it's if 0:06:25.600 --> 0:06:28.200 it's using machine learning to to watch a chef and 0:06:28.240 --> 0:06:31.840 repeat steps back, then then that's a thing. But if 0:06:31.880 --> 0:06:34.760 it's it's reading off of a recipe, then there's there's 0:06:34.800 --> 0:06:38.200 many many things that can go wrong, mostly the human 0:06:38.320 --> 0:06:41.839 error in writing recipes, which I suspect is what was 0:06:41.920 --> 0:06:44.320 the problem. It could very easily have been that someone 0:06:44.360 --> 0:06:47.840 had put in the wrong measurement for something or left 0:06:47.880 --> 0:06:50.800 an ingredient out entirely sure, but I mean you also 0:06:50.839 --> 0:06:53.039 have to assume that the atmosphere conditions in your house 0:06:53.040 --> 0:06:55.000 are comparable to those of the chef's kitchen, that your 0:06:55.040 --> 0:06:58.160 appliance as an equipment are precise and nonreactive. Is in 0:06:58.680 --> 0:07:02.000 you're not having a reaction between like aluminum and lemon juice, 0:07:02.000 --> 0:07:04.159 which is a thing that happens all I'm trying to say, 0:07:04.160 --> 0:07:09.360 here is that cooking and especially baking our our edible chemistry, 0:07:09.480 --> 0:07:13.239 and there are so many things that can go awry. Sure, 0:07:13.400 --> 0:07:15.760 I mean, the first thing that occurred to me when 0:07:15.840 --> 0:07:17.760 you said that they went home and they made it 0:07:17.800 --> 0:07:19.960 and it wasn't the same was, well, did they get 0:07:19.960 --> 0:07:22.760 their ingredients from the same place? Because I imagine the 0:07:22.880 --> 0:07:25.360 ingredients they bought were probably not as good as the 0:07:25.520 --> 0:07:29.160 ingredients that were bought by this restaurant. It's quite possible 0:07:29.200 --> 0:07:33.240 did they weigh them or measure them? Yeah? So if 0:07:33.280 --> 0:07:35.520 if your weight isn't a measurement, you guys know, if 0:07:35.520 --> 0:07:38.360 you're able to if you're right, if they buy volume 0:07:38.480 --> 0:07:41.840 or weight? Uh. If if in fact, they were following 0:07:41.840 --> 0:07:44.520 it as closely as they possibly could with the ingredients 0:07:44.600 --> 0:07:47.520 that were as close to the same quality as the restaurant, 0:07:47.800 --> 0:07:50.080 there still could be other factors at play, like Lauren 0:07:50.160 --> 0:07:53.000 is saying, that could affect this. Now, if you had 0:07:53.000 --> 0:07:56.000 a robot, it may very well be that the robot 0:07:56.040 --> 0:07:58.760 making the recipe would also come up with something that 0:07:58.880 --> 0:08:01.040 does not taste as good is what they had at 0:08:01.080 --> 0:08:04.240 the restaurant. I would argue, however, the robot would be 0:08:04.240 --> 0:08:06.960 more likely to consistently make it not as good as 0:08:07.000 --> 0:08:11.360 the restaurant, because that's what robots do, right. Robots are 0:08:11.440 --> 0:08:15.840 really good at repeating steps that they have been programmed 0:08:15.840 --> 0:08:18.200 to do, or, in the case of machine learning, taught 0:08:18.280 --> 0:08:20.960 how to do, assuming the starting conditions are the same. 0:08:21.440 --> 0:08:24.200 Another thing that's difficult with cooking, Yes, obviously if you 0:08:24.320 --> 0:08:26.680 are one of the things we pointed out in our 0:08:26.680 --> 0:08:30.200 previous episode. One of the specific robots we talked about 0:08:30.240 --> 0:08:33.680 was Moley. That's the kitchen that has the two arms 0:08:33.720 --> 0:08:36.840 coming down from like the kitchen hood, the oven hood. Oh, 0:08:36.880 --> 0:08:41.400 it's beautiful. And you pointed out, Joe in that episode 0:08:41.480 --> 0:08:45.240 that something as simple as replacing the extra virgin olive 0:08:45.240 --> 0:08:48.400 oil with arsenic would create an entirely different dinner party 0:08:48.480 --> 0:08:50.760 than the one you had intended, right, because it doesn't 0:08:50.760 --> 0:08:54.400 know what the ingredients it's using are. It's it's spatial programming, 0:08:54.600 --> 0:08:58.599 so it says, okay, X ingredient goes here, x y 0:08:58.800 --> 0:09:02.000 ingredient goes here, So it doesn't actually know that this 0:09:02.040 --> 0:09:06.040 dish contains olive oil, balslamic vinegar or whatever. It knows 0:09:06.120 --> 0:09:10.680 that it contains slot number one and slot number four. Yeah, 0:09:10.720 --> 0:09:15.280 almost every single kitchen robot i've ever seen uses different 0:09:15.320 --> 0:09:20.480 shaped containers to represent an order of operations, right like, 0:09:21.040 --> 0:09:23.640 at this stage of cooking, add whatever is in the 0:09:24.000 --> 0:09:28.280 square yellow container. At this next step, add whatever in 0:09:28.320 --> 0:09:32.040 the cylindrical red container, and that way, the robot knows 0:09:32.480 --> 0:09:34.720 this is the order I follow. But the robot does 0:09:34.760 --> 0:09:39.440 not know what is actually in those containers, what it means, 0:09:39.920 --> 0:09:42.080 or why it's ordered putting them in this order. It 0:09:42.160 --> 0:09:44.880 just learned how to do it in this way, So 0:09:44.920 --> 0:09:48.000 it's essentially aping what someone else is doing. It's not 0:09:48.720 --> 0:09:51.720 it's not learning in the sense of you understanding what 0:09:51.840 --> 0:09:55.840 the process is. So when we humans learn, it's very different, 0:09:55.960 --> 0:09:59.600 right We We first might learn just through route, right, 0:09:59.720 --> 0:10:02.200 but then we start to understand why it's working the 0:10:02.200 --> 0:10:04.800 way it's working. That's the level of understanding that goes 0:10:04.880 --> 0:10:08.480 much deeper than just parroting what someone else has already done. 0:10:09.280 --> 0:10:12.960 So that's already a limitation, but again, at least you 0:10:12.960 --> 0:10:17.000 would get consistency, which is something I can never attest 0:10:17.240 --> 0:10:19.960 like I can never I can never be sure that 0:10:20.400 --> 0:10:22.720 something I'm cooking one time is going to come out 0:10:23.440 --> 0:10:26.240 even remotely similar another time, because most of the stuff 0:10:26.280 --> 0:10:30.880 I cook it's by feel. I'm not doing following specific recipes. Though, 0:10:31.000 --> 0:10:33.120 for those of you listening at home, you want to 0:10:33.120 --> 0:10:37.120 know what's the easiest way to make my cooking more consistent? 0:10:37.200 --> 0:10:40.439 There is one way I can recommend, use a digital thermometer. 0:10:40.800 --> 0:10:43.319 Use a digital thermometer on everything, and you get way 0:10:43.360 --> 0:10:46.760 more consistent. That's true. Yeah, and uh, your oven is 0:10:46.800 --> 0:10:49.720 probably lying to you. The other thing I would point 0:10:49.720 --> 0:10:54.120 out is that while Lusk says, wouldn't this be wonderful? 0:10:54.160 --> 0:10:56.080 You could have a robot chef make you the perfect 0:10:56.080 --> 0:10:58.880 dish every time, part of me thinks, well, that also 0:10:58.920 --> 0:11:04.520 takes out the opportunity to improvise and to tweak and 0:11:04.600 --> 0:11:08.520 to experiment, where you might end up finding that by 0:11:08.559 --> 0:11:12.560 departing from this approach that you love, I mean you 0:11:12.600 --> 0:11:16.120 love the finished dish, you create something even better. Right, 0:11:16.360 --> 0:11:19.199 you take innovation out. Now you would have to incorporate 0:11:19.240 --> 0:11:22.160 something like IBM chef Watson to bring it back in. 0:11:22.679 --> 0:11:25.880 But then who knows what you would end up with? Yeah, well, 0:11:25.920 --> 0:11:28.319 I mean this comes up to a distinction I would make. Now, 0:11:28.360 --> 0:11:31.640 some culinary people might uh not like me using the 0:11:31.760 --> 0:11:33.920 terms these this way, but here's what I would say 0:11:33.960 --> 0:11:36.160 that the difference between a cook and a chef. A 0:11:36.240 --> 0:11:39.199 cook knows what to do in the kitchen. A chef 0:11:39.240 --> 0:11:42.200 knows why you do what you do. I would agree 0:11:42.200 --> 0:11:45.680 with that that definition understand so he or she would 0:11:45.760 --> 0:11:50.080 understand all of the science behind why meat should be 0:11:50.080 --> 0:11:53.240 cooked to a certain temperature or on a certain type 0:11:53.240 --> 0:11:57.080 of appliance, or with why a certain flavor goes with 0:11:57.120 --> 0:11:59.720 another flavor right or or at the very least, A 0:11:59.800 --> 0:12:03.880 chef has has room to play due to their profession. 0:12:04.320 --> 0:12:08.160 Uh And and a cook is following orders, probably to 0:12:08.160 --> 0:12:11.440 to to get a thing out to a person to 0:12:11.520 --> 0:12:14.559 a certain degree of quality that has been determined by 0:12:14.640 --> 0:12:17.280 the chef. The chef is the person who creates the dish. 0:12:17.600 --> 0:12:20.040 The cook is the person who replicates the dish. I 0:12:20.040 --> 0:12:23.080 would argue, Um, and in some cases we play chef, 0:12:23.200 --> 0:12:25.160 or we may play cook. But you know, there are 0:12:25.240 --> 0:12:28.080 some people who are very much more in one camp 0:12:28.080 --> 0:12:29.640 than the other. I would argue, I'm more of a 0:12:29.640 --> 0:12:33.560 cook definitely. Um. I've tweaked recipes, but that's about as 0:12:33.559 --> 0:12:37.920 far as I've gone. I've never invented one. So the Telegraph, 0:12:38.080 --> 0:12:41.480 by the way, also references the moley. This this armed kitchen, 0:12:42.679 --> 0:12:44.800 as in a kitchen that has arms, not a kitchen 0:12:44.800 --> 0:12:47.960 that has weapons. In an article about ten ways humans 0:12:47.960 --> 0:12:51.479 are still superior to robots, which is not an endorsement 0:12:51.520 --> 0:12:54.560 in the Moley's abilities, I would argue, Uh, that says 0:12:54.600 --> 0:12:56.760 the robot has really only been shown to master the 0:12:56.840 --> 0:12:59.560 art of making crabbisk, which is specifically what we were 0:12:59.559 --> 0:13:04.120 talking about last Yeah, uh and on Yeah, making crab bisk. 0:13:04.200 --> 0:13:06.440 It's great at making crab bisks. So if you think 0:13:06.520 --> 0:13:11.160 crab bisk is worth paying fifty thousand pounds sterling to 0:13:11.200 --> 0:13:15.040 get a robot that can really make crab misk, have 0:13:15.200 --> 0:13:17.080 we got the robot for you? What's what's that like? 0:13:17.080 --> 0:13:21.160 Like something something like that. Yeah, I didn't do the conversion, 0:13:21.160 --> 0:13:23.560 but you have fifty pounds sterling. Is is quite a 0:13:23.600 --> 0:13:26.800 bit of money. That's very expensive bisk. Those crabs were 0:13:26.840 --> 0:13:28.800 living in the lap of luxury before you plucked about 0:13:28.840 --> 0:13:31.640 the ocean, I assume. But just think about the scenario 0:13:31.679 --> 0:13:34.600 you described at the beginning. You come home tired and 0:13:34.800 --> 0:13:37.320 you don't have the energy to make dinner, But gosh 0:13:37.360 --> 0:13:40.280 darn it, you can have crab bisk again like you 0:13:40.320 --> 0:13:43.520 did every other night. Joe, please let's not talk about 0:13:43.559 --> 0:13:47.240 me having crab. We'll get into that a little later. Yes, yes, 0:13:47.320 --> 0:13:50.160 so well uh so so so moley is is not 0:13:50.240 --> 0:13:53.880 necessarily at the top of its game yet. Yeah, it's not. 0:13:54.040 --> 0:13:56.920 It's not out. It's not available for consumer purchases or 0:13:56.960 --> 0:13:59.360 commercial purchase. It's really going to be more a commercial 0:13:59.360 --> 0:14:01.640 thing than a consume more things. But it is. It 0:14:01.720 --> 0:14:05.040 is still targeted for release by That was according to 0:14:05.080 --> 0:14:08.880 the Telegraph article, which was published in December, so not 0:14:08.920 --> 0:14:14.080 that long ago. So unless something is remarkably changed, ye is, 0:14:14.080 --> 0:14:17.199 when the company hopes to release this or unleash it 0:14:17.360 --> 0:14:19.760 upon the world. Uh. And I did want to put 0:14:19.800 --> 0:14:23.160 in here that that this company is not the only 0:14:23.560 --> 0:14:28.520 entity that's in in the robo chef game. Uh. DARPA 0:14:28.600 --> 0:14:32.040 has funded a robotic chef, which is I mean, if 0:14:32.080 --> 0:14:35.440 you've seen DARPA's attempts at making robots capable of opening 0:14:35.520 --> 0:14:37.880 up a door, it's going to fill you with confidence. 0:14:38.720 --> 0:14:42.160 They funded a study out of the University of Maryland. 0:14:42.240 --> 0:14:46.120 It was published in early and researchers in the study 0:14:46.160 --> 0:14:50.400 trained robots to learn how to duplicate actions by watching 0:14:50.440 --> 0:14:54.160 cooking videos on YouTube. Now if they also learned how 0:14:54.200 --> 0:14:57.560 to leave comments that would have been the most destructive 0:14:57.680 --> 0:15:01.360 robot ever made, like forget for It, making war Robots 0:15:01.400 --> 0:15:03.520 a cooking robot that can comment on YouTube, and the 0:15:03.560 --> 0:15:07.800 game was games over. They used these data from eighty 0:15:07.840 --> 0:15:11.359 eight videos involving the manipulation of one or more objects, 0:15:11.600 --> 0:15:17.239 and then UH programmed the robots to train themselves in recognizing, grasping, 0:15:17.320 --> 0:15:21.160 and manipulating a variety of kitchen utensils and objects. Now, 0:15:21.240 --> 0:15:24.040 I hope some of these videos were my drunk kitchen 0:15:24.480 --> 0:15:29.040 with Hannah Heart rightly. Also also Death Metal Vegan kitchen 0:15:29.120 --> 0:15:32.960 experience where they just throw the the the like the 0:15:32.960 --> 0:15:35.360 head of lettuce against the wall and then hit it 0:15:35.400 --> 0:15:38.280 with an ax And I don't know I've ever seen 0:15:38.320 --> 0:15:41.320 him do that, but it is a thing. Yeah, I 0:15:41.480 --> 0:15:45.000 don't believe so, but I didn't I didn't have access 0:15:45.040 --> 0:15:46.960 to the full sample set, so I'm not I'm not 0:15:47.040 --> 0:15:50.600 positive it's a possibility. Well, I'm sorry. Please continue. When 0:15:51.040 --> 0:15:53.960 when the robots were then tested on their ability to 0:15:54.040 --> 0:15:56.920 replicate the actions from the videos that they were from 0:15:57.040 --> 0:16:00.520 from other videos they were seeing for the first time, um, 0:16:00.560 --> 0:16:04.480 the robots correctly recognized sev of the objects they were 0:16:04.520 --> 0:16:09.040 dealing with They performed the correct grasps cent of the time, 0:16:09.480 --> 0:16:12.560 and they guessed what action would need to be applied 0:16:12.760 --> 0:16:15.120 to the to the two up to the object or 0:16:15.120 --> 0:16:20.320 objects at hand of the time. So not not bad, 0:16:20.440 --> 0:16:23.400 not bad robots, well, especially when you consider like the 0:16:23.480 --> 0:16:28.720 wide variety that typical kitchen utensils can take shape and form. 0:16:28.720 --> 0:16:31.360 I mean, even something as simple as a ladle can 0:16:31.400 --> 0:16:34.920 look very different from one designer to another, absolutely, and 0:16:34.960 --> 0:16:38.240 that means more challenges for for object recognition on the 0:16:38.240 --> 0:16:40.560 computer side of things. Oh yeah, and they pulled it 0:16:40.560 --> 0:16:45.360 off thanks to a convolutional neural network, which is the 0:16:45.400 --> 0:16:48.040 same thing that Google's Deep Dream is using. Okay, so 0:16:48.160 --> 0:16:51.200 it it tosses ideas around and it's in its own head, 0:16:51.240 --> 0:16:56.440 its own robo uh computer head um and extrapolates on 0:16:56.480 --> 0:16:59.160 the patterns that it finds there before spitting out a 0:16:59.200 --> 0:17:04.320 response to it's seeing. And so so that plus like 0:17:04.359 --> 0:17:09.960 a really highly specific taxonomy of grasp types and motions 0:17:10.440 --> 0:17:14.159 were what drove this to to pretty decent success. So 0:17:14.200 --> 0:17:16.280 that way it would know things like with the ladle, 0:17:16.320 --> 0:17:19.200 you're dipping as opposed to stirring or whisking that kind 0:17:19.200 --> 0:17:21.520 of thing. Yeah, because obviously a whisking robot with a 0:17:21.600 --> 0:17:24.960 ladle would be hilarious but not terribly effective at making 0:17:25.000 --> 0:17:27.280 whatever it was you were hoping that. I certainly wouldn't 0:17:27.359 --> 0:17:29.840 get your crab bisk. No, you might get you crab 0:17:29.880 --> 0:17:31.719 bisk on the ceiling. That might be about as good 0:17:31.720 --> 0:17:33.920 as you get. I just thought of another great video. 0:17:33.960 --> 0:17:36.399 They should have shown these robots what it should be, 0:17:36.440 --> 0:17:40.440 the Harvey Corman as Julia Child's Alien baking video and 0:17:40.640 --> 0:17:48.000 Star Wars special Whiskster whiske Sirs. I have no idea 0:17:48.080 --> 0:17:50.399 what you guys are talking about, but I'm delighted. I mean, 0:17:50.960 --> 0:17:53.320 I haven't seen it. As we're going to huddle around 0:17:53.359 --> 0:17:55.480 and watch a video when this is over, Okay, I'll 0:17:55.520 --> 0:17:57.399 just let's mark out an hour and a half of 0:17:57.440 --> 0:18:01.880 our lives will get back and and okay. So DARPA 0:18:02.119 --> 0:18:06.800 is probably not interested in robot chefs. I just wanted 0:18:06.840 --> 0:18:08.960 I wanted to put in here um. But but a 0:18:09.000 --> 0:18:13.080 representative of DARPA indicated that this technology could have applications 0:18:13.520 --> 0:18:18.200 and I quote in areas such as military repair and logistics. 0:18:19.920 --> 0:18:24.240 Military repair, I understand logistics that terrifies me, but you know, hey, 0:18:24.280 --> 0:18:27.000 go go DARPA. I'm glad that you're funding robotically. I 0:18:27.040 --> 0:18:29.560 can't win along without breaking a few eggs. I can 0:18:29.560 --> 0:18:32.280 see legit. Okay, So I mean one of the ways 0:18:32.280 --> 0:18:36.720 we see DARPA logistics in robotics is in just carrying 0:18:36.800 --> 0:18:40.000 things like the DARPA four legged you know, the DARPA 0:18:40.080 --> 0:18:43.840 dogs and stuff like that. They move over difficult terrain 0:18:44.400 --> 0:18:48.080 dynamics big dogs exactly. Yeah, So all right, see it well? 0:18:48.119 --> 0:18:50.399 And and I like the idea that with this and 0:18:50.440 --> 0:18:52.879 the Grand Challenge, which you know also I made a 0:18:52.960 --> 0:18:55.560 joke at the beginning about DARPA trying to open a door. 0:18:55.840 --> 0:18:58.959 To be fair, what DARBA does is it funds the program. 0:18:59.680 --> 0:19:02.000 Other people try and make robots to open doors to 0:19:02.119 --> 0:19:06.040 hilarious results. But I like the idea that eventually DARMA 0:19:06.119 --> 0:19:07.639 is going to have a robot that will come to 0:19:07.680 --> 0:19:09.879 your house, open the door, break into your home, and 0:19:09.880 --> 0:19:13.600 then make you dinner. That's the future I want to 0:19:13.640 --> 0:19:16.160 live in. Like it's terrified but also kind of comforting. 0:19:16.359 --> 0:19:19.439 So it does one random house each night. It's like 0:19:19.800 --> 0:19:22.120 tonight might be the night. Right, you hear some pounding 0:19:22.160 --> 0:19:24.639 on the front door, You're like, is it is it 0:19:24.680 --> 0:19:30.280 tonight it's robot claw. We get cram, so canceled that 0:19:30.359 --> 0:19:34.199 pizza order. That's right, right man. I want to live 0:19:34.240 --> 0:19:38.800 in that future so so less ambitious than these robots 0:19:38.880 --> 0:19:42.000 that would be capable of of learning from chefs and 0:19:42.040 --> 0:19:46.919 being able to replicate a chef's motions and processes so 0:19:46.960 --> 0:19:50.960 that you could create amazing dishes. Are some other robots 0:19:51.000 --> 0:19:54.280 that are being incorporated into workplaces. In fact, there's one 0:19:54.320 --> 0:19:56.280 that's been around for a few years. I think the 0:19:56.320 --> 0:19:58.320 last the first video I ever saw that was from 0:19:58.320 --> 0:20:01.600 two thousand twelve h um And it's one that's been 0:20:01.680 --> 0:20:05.360 used in in kitchens in China and also Los Angeles 0:20:05.440 --> 0:20:09.119 now has one apparently. Um So there's a gain in 0:20:09.320 --> 0:20:13.240 a shoe room. Kwan who owns restaurants in Beijing, China, 0:20:13.400 --> 0:20:16.680 and created a robot to cut noodles. And the way 0:20:16.720 --> 0:20:19.600 that it cuts noodles it holds this big essentially like 0:20:19.720 --> 0:20:23.200 almost like a log of dough, and then it has 0:20:23.240 --> 0:20:27.200 a knife in one arm one hand that just kind 0:20:27.200 --> 0:20:31.000 of makes a windshield wiper motion as the other arms 0:20:31.080 --> 0:20:35.160 slowly rotates the log so that it just cuts strips 0:20:35.320 --> 0:20:38.360 off of it, and the strips become these long flat noodles, 0:20:38.760 --> 0:20:40.560 and it can do different style noodles, but that was 0:20:40.600 --> 0:20:43.520 the one that was on display in the video, and 0:20:44.200 --> 0:20:47.360 according to the inventor, it can go about twice as 0:20:47.359 --> 0:20:50.080 fast as your typical human can go, and said that 0:20:50.320 --> 0:20:52.720 in a way, the robot actually became a necessity. It 0:20:52.760 --> 0:20:57.480 wasn't that, uh, he was trying to replace human employees. 0:20:57.520 --> 0:20:59.440 It was that he was having it. He was seeing 0:20:59.520 --> 0:21:03.440 that was coming more difficult to find young people who 0:21:03.600 --> 0:21:07.000 wanted to work in restaurants cutting noodles because the job 0:21:07.040 --> 0:21:11.639 is monotonous, it's repetitive. It's the sort of stuff about 0:21:11.640 --> 0:21:13.680 the interesting part of working in the kitchen, right, And 0:21:13.760 --> 0:21:16.480 also it's it's what we hear when we hear roboticists 0:21:16.480 --> 0:21:19.680 talk about why we want automation to take over jobs 0:21:19.680 --> 0:21:23.680 that are dirty, dull and dangerous. Right. This is possibly 0:21:23.760 --> 0:21:26.960 dangerous because you're using a knife, but more so it's dull. 0:21:27.040 --> 0:21:29.280 And if it's still then you can get give you 0:21:29.359 --> 0:21:33.080 dangerous anyway. Sure, Well, in and repetitive motions, injuries are 0:21:33.119 --> 0:21:35.200 absolutely a thing in that kind of industry. Yeah, you 0:21:35.200 --> 0:21:37.719 could have like essentially the equivalent of tennis elbow, but 0:21:37.840 --> 0:21:40.199 you've got it because you're you're just doing the same 0:21:40.240 --> 0:21:43.640 motion throughout the whole day. The robot can keep doing 0:21:43.640 --> 0:21:46.359 it until it needs maintenance. Right, it's not going to 0:21:46.440 --> 0:21:49.480 feel pain. Sure, they did not program it to do so. 0:21:49.920 --> 0:21:55.439 And of course that automated preparation of certain food stuff 0:21:55.520 --> 0:21:58.680 sizes and portions and stuff like that is nothing new 0:21:58.760 --> 0:22:02.600 inherently because you see out of the industrial level. But now, 0:22:02.640 --> 0:22:04.720 I guess what's interesting here is that this is this 0:22:04.800 --> 0:22:07.440 is at the end of the consumer chain, pretty r 0:22:08.760 --> 0:22:12.440 in a restaurant where you are making homemade noodles within 0:22:12.480 --> 0:22:15.200 the restaurant. So it's not like you've ordered the noodles 0:22:15.320 --> 0:22:17.840 that were produced in some factory that had a huge 0:22:17.920 --> 0:22:21.920 automated assembly line that created the noodles. You're actually making 0:22:21.920 --> 0:22:24.159 the dough in house, and then you are cutting the 0:22:24.160 --> 0:22:27.320 noodles in house with this robot. He also designed it 0:22:27.359 --> 0:22:30.040 to have kind of this humanoid shape and it has 0:22:30.080 --> 0:22:32.119 sort of a retro robot look to it, like it 0:22:32.160 --> 0:22:37.720 makes me think of classic Japanese animation um uh and uh. 0:22:37.880 --> 0:22:41.960 It has this uh, this you know, very stylistic approach 0:22:42.000 --> 0:22:44.240