[00:14.920 --> 00:19.720]  Touching one of the first commercially available modular robotic systems in the world, actually.
[00:20.620 --> 00:22.320]  For adults, it's the only one.
[00:23.000 --> 00:24.300]  There are a few other ones.
[00:24.420 --> 00:27.060]  One for kids, that is just lovely to see them play with.
[00:27.260 --> 00:29.860]  So, I don't have those with me today.
[00:29.860 --> 00:32.300]  I guess we're a bit too old.
[00:32.580 --> 00:37.560]  And there is another system coming, but it isn't ready yet for the general public.
[00:38.740 --> 00:41.260]  So, my name is Per Sjöborg, and I'm from Sweden.
[00:41.260 --> 00:45.920]  And I'm going to talk to you about the title of my talk you see there.
[00:46.200 --> 00:49.680]  And if you pronounce my name like the fruit pear, you're spot on.
[00:49.840 --> 00:52.820]  And that might be good if you want to talk to me later during HOPE.
[00:53.000 --> 00:53.920]  And I hope you want to do that.
[00:53.980 --> 00:56.160]  We could also test the robots out.
[00:56.300 --> 01:00.580]  I've got everything we need to play with them just for the whole two days that are remaining.
[01:02.860 --> 01:05.940]  I'm going to talk to you about self-reconfiguring modular robotics.
[01:05.940 --> 01:10.920]  And the name isn't very good.
[01:11.080 --> 01:16.820]  It's called other things at other times like programmable matter or utility fog you might have heard of.
[01:17.220 --> 01:22.260]  But I think that self-reconfiguring modular robotics actually has one thing going for it.
[01:22.380 --> 01:25.720]  And it describes pretty well what it's all about.
[01:26.420 --> 01:29.080]  So, let's start with modular robotics.
[01:29.080 --> 01:34.360]  In modular robotics, we build what we want from standardized bricks, modules.
[01:35.180 --> 01:37.960]  And the simpler the modular is, the better.
[01:38.300 --> 01:40.200]  Because it's more generally applicable.
[01:40.880 --> 01:42.780]  It can be used in more situations.
[01:43.940 --> 01:48.500]  So, general versus custom, we go the general way and then we assemble what we need from those.
[01:49.260 --> 01:53.140]  And this we encompass in the ensemble axiom.
[01:53.340 --> 02:00.340]  A modular only includes enough functionality to contribute to the functionality desired in the ensemble.
[02:01.200 --> 02:07.080]  That's why, you see here, that has three degrees of freedom and it's a very, very capable module.
[02:07.260 --> 02:08.320]  Most only have one.
[02:09.140 --> 02:12.400]  Because that's the smallest number you can have.
[02:14.880 --> 02:22.600]  And here you see the blue LEGO pieces, of course, the one that fits the ensemble axiom best.
[02:22.880 --> 02:25.420]  Because you could build the other two from the one.
[02:25.720 --> 02:29.480]  And it shows actually quite clearly that that's actually what they've done.
[02:29.880 --> 02:33.020]  They've built all the three LEGO pieces from LEGO pieces.
[02:33.320 --> 02:40.120]  And you could, of course, use small one-stud LEGO pieces to build all of them if you wanted to.
[02:42.500 --> 02:44.760]  Then we come to self-reconfiguring.
[02:45.500 --> 02:49.220]  And this is just to say that instead of you building and assembling them...
[02:49.220 --> 02:53.360]  And I should point out that what you're holding in your hand is not self-reconfiguring.
[02:53.500 --> 02:54.680]  That's a modular robot.
[02:55.000 --> 02:59.420]  The connector for a self-reconfiguring system is very hard.
[02:59.600 --> 03:01.620]  So, we haven't really quite cracked that yet.
[03:03.000 --> 03:08.300]  But self-reconfiguring automates this process of assembling the modules.
[03:11.060 --> 03:18.620]  It's also worth mentioning that not only can they rearrange themselves among each other, but they can also change their properties.
[03:18.880 --> 03:20.180]  This is what that one can do.
[03:20.520 --> 03:21.520]  They can be longer.
[03:21.640 --> 03:23.640]  They can hold on to each other in different angles.
[03:23.780 --> 03:26.280]  They can change their color or stuff like that.
[03:26.680 --> 03:30.320]  This gives us a very, very wide scope of change.
[03:30.320 --> 03:33.140]  We can adjust ourselves a fraction of an inch.
[03:33.300 --> 03:37.320]  Or we can turn ourselves from a sofa to a table.
[03:37.960 --> 03:39.300]  Depending on what we want to do.
[03:40.820 --> 03:44.740]  So, the automation and modularity put together.
[03:44.740 --> 03:49.980]  I mean, nobody can deny the significance of automation or modularity.
[03:50.140 --> 03:53.560]  It's basically the foundation for our society today, huh?
[03:54.960 --> 03:57.960]  Put together though, we get something really exciting.
[03:58.200 --> 04:00.880]  We get a universal physical machine.
[04:03.840 --> 04:09.300]  And to draw an analogy there is to say that the computer...
[04:09.300 --> 04:11.160]  Well, we just call it the computer.
[04:11.460 --> 04:15.140]  But it is actually a general computing machine.
[04:15.440 --> 04:22.080]  Capable of simulating any dedicated computing machine.
[04:22.120 --> 04:25.180]  The ones we had before we got what we call the computer.
[04:25.180 --> 04:28.320]  And even further back than that, they were actually people.
[04:28.540 --> 04:30.640]  It was a job title to be a computer.
[04:30.980 --> 04:33.060]  You did computation by hand.
[04:34.120 --> 04:41.700]  So, what would we get if we could create this amazing working self-reconfiguring modular robotic system?
[04:44.080 --> 04:52.940]  And to talk about that, I'd like to ask you a question, which I paraphrased from Douglas Engelbart.
[04:52.940 --> 04:58.020]  And it's from his presentation, usually referred to as the mother of all demos.
[04:58.420 --> 05:02.040]  If you haven't seen that one online, you should definitely check it out.
[05:02.180 --> 05:08.060]  He presents something that is a mix of Twitter and Word and Windows and all of this put together.
[05:08.200 --> 05:09.480]  And it's like 40 years old.
[05:09.600 --> 05:11.560]  And he shows us how it could work, you know?
[05:12.900 --> 05:24.660]  So, he says if you had a system that was low cost, environmentally friendly, virtually unbreakable, could form many physical structures, perform many tasks, what value could you derive from that?
[05:25.340 --> 05:29.840]  And that's what I hope to talk about today.
[05:30.900 --> 05:35.740]  I chose the title, No Natural Resources Were Hurt Assembling the Sofa.
[05:36.120 --> 05:44.760]  Because one, it's naturally a paraphrase of the line you sometimes or most often you see in the end of movies.
[05:46.780 --> 05:49.860]  No animals were harmed in the making of this motion picture.
[05:51.360 --> 06:01.720]  So, I chose this title because one of the things that self-reconfiguring modular robotics would allow us to do is to drastically change our relationship to raw materials.
[06:01.740 --> 06:04.620]  The stuff that the world around us is made of.
[06:09.600 --> 06:19.100]  Because when assembling one thing in a self-reconfiguring modular robotic system, as we do in a computer, we use only energy.
[06:19.400 --> 06:24.660]  I mean, when we load another software after we've used another software, we use only energy.
[06:24.780 --> 06:30.080]  The computer stays the same, and it's the same thing with the modules in a self-reconfiguring modular robotic system.
[06:30.080 --> 06:33.220]  The modules stays the same, and we use them for different things.
[06:33.360 --> 06:36.140]  We add a little bit of energy that allows them to self-reconfigure.
[06:36.380 --> 06:40.720]  And then we've got a table, a sofa, a chair, a bed, whatever we need.
[06:45.320 --> 06:47.420]  And, of course, we can do that time and time again.
[06:50.780 --> 06:57.720]  And this could drastically slow down the rate with which raw materials flow through our society the way it does today.
[06:57.720 --> 07:02.780]  It goes from natural resources to products and then to waste, basically landfill.
[07:03.060 --> 07:08.900]  We try to recycle a bit there, but it's quite hard because we get this huge mix of everything we throw in the bin.
[07:09.040 --> 07:12.940]  It's hard to know what's actually in there and to recycle it the best way.
[07:15.720 --> 07:18.960]  So you use only what you need right now.
[07:18.960 --> 07:21.420]  Because whatever you...
[07:21.960 --> 07:30.540]  Since you could always assemble what you need at any given time, you only have what you need right now.
[07:30.820 --> 07:35.120]  This could seem simple, but what you don't need is not there.
[07:35.360 --> 07:37.320]  Because when you needed it, it would assemble.
[07:38.760 --> 07:44.460]  And this is significant when you think about how little we actually use most of the stuff we have.
[07:45.120 --> 07:51.320]  It's been reported that the life... in its life, the average power drill is used 14 minutes.
[07:52.040 --> 07:56.120]  The rest of the time, 15, 20 years, it just sits there on the shelf.
[07:58.040 --> 08:04.980]  So that's... the fact that we have what we need, but don't have what we don't need, is very significant.
[08:06.820 --> 08:08.960]  But it doesn't actually end there.
[08:09.660 --> 08:18.020]  There are a number of other features that changes our relationship to raw material that is instrict as a part of self-reconfiguring modular robotics.
[08:21.040 --> 08:22.880]  I'm going to talk about a few of them.
[08:22.880 --> 08:26.860]  If you want to talk about this more, I'm here all of the HOPE conference.
[08:27.460 --> 08:29.140]  But I'm going to mention a few of them.
[08:29.140 --> 08:35.240]  One of the more important things is that we could use the module for its entire service life.
[08:35.580 --> 08:37.640]  Until it simply isn't usable anymore.
[08:38.280 --> 08:42.900]  And very few things... I guess that power drill is thrown away long before it's worn out.
[08:43.060 --> 08:46.000]  I have a hard time seeing you wearing it out in 14 minutes.
[08:48.320 --> 08:50.720]  So you can use it for the full service life.
[08:50.720 --> 08:58.980]  And even if it's not up to 100% specification, there's always many situations where its capabilities are more than enough to be used.
[09:01.880 --> 09:08.280]  And when the module breaks, I mean, they're going to break, they're going to be so old we don't want to use them anymore.
[09:08.740 --> 09:17.200]  We can actually, since we do a few types of modules, we could actually bring them back into the factory and disassemble them and reuse the components within them.
[09:17.200 --> 09:23.380]  So say that something is broken, it's very likely that most of the parts in whatever is broken are perfectly alright.
[09:23.600 --> 09:25.640]  It's just one or two parts that are broken.
[09:26.060 --> 09:33.800]  Here, since we have few types, but many of them, we can bring them back into the factory, disassemble the module, and use the parts that are okay.
[09:34.420 --> 09:34.900]  Again.
[09:35.080 --> 09:41.060]  And of course, we could do this again and again and again until that particular part is not usable anymore.
[09:46.190 --> 09:52.550]  And even if it's... when we design the modules, where they move, for instance, or how they're used, they're worn.
[09:52.770 --> 09:59.970]  But since we know kind of where they are worn and how we can handle that in another way.
[10:00.050 --> 10:06.590]  Say we disassemble the module, we get a part out, it isn't really up to spec, but it's not too far out to spec.
[10:06.590 --> 10:13.610]  So we can rework the module, maybe insert a washer where it moves, a little bit thicker washer than we used the first time.
[10:13.790 --> 10:16.330]  And then we could actually reuse the module again.
[10:16.690 --> 10:25.390]  Maybe in a module that can only... the part in a module that is only... has a lower specification, but it's still usable for some, most of the time.
[10:31.500 --> 10:34.660]  And sooner or later, the part is not going to be useful anymore.
[10:34.940 --> 10:38.020]  It's going to be out of spec, it's going to be old, we don't want it anymore.
[10:38.020 --> 10:47.480]  But since we know exactly what the material is made out of is, we could actually recycle the raw material far more efficient than we do today.
[10:47.480 --> 10:53.080]  Because we know exactly what we have, we know exactly how to treat it, to recycle it the most efficient way.
[10:53.740 --> 10:57.760]  This is actually one of the bigger problems in recycling today, know exactly what you've got.
[10:58.380 --> 11:03.440]  That and two substances mixed together, that's the two hard issues.
[11:05.860 --> 11:16.480]  But to come back to the beginning where we actually make the module, since the surface of whatever we assemble, and the internal structure are two separate parts.
[11:17.360 --> 11:20.880]  I guess we wouldn't build a bed out of the module I've just passed around.
[11:20.980 --> 11:23.720]  It's kind of hard and would probably be very uncomfortable.
[11:29.640 --> 11:45.000]  The fact that the surface and the internals are two separate things, which means we could actually use recycled raw materials to make the 90% of the internal structure, and just use new materials for the surface and what you see and interact with.
[11:45.000 --> 11:53.720]  So not only can we recycle much more efficiently, we can also use recycled material that is available in a way that is not possible today.
[11:56.680 --> 11:58.020]  I'm a little bit behind.
[12:07.700 --> 12:10.160]  So, sad to say, this is not all for free.
[12:10.440 --> 12:13.100]  This is... modular robotics has many issues.
[12:13.300 --> 12:19.160]  And some of them are not solved, some of them are kind of solved, and others we're working on right now.
[12:19.160 --> 12:22.680]  Overhead of modularity is one of the challenges we face.
[12:23.400 --> 12:32.420]  Because a dedicated custom component can always be perfectly adapted to the situation it's supposed to work in.
[12:32.880 --> 12:38.320]  Practically, that's rarely so, because we cannot have one tool for one small niche task.
[12:38.620 --> 12:40.860]  Then we need an infinite number of tools.
[12:41.140 --> 12:51.760]  So, although the custom argument there is valid and interesting, in practice, it's not as significant as one might think.
[13:00.000 --> 13:14.020]  And also, in a self-reconfiguring modular robotics system, we could adapt the amount of overhead we incorporate in the system, depending on what we want for the system.
[13:14.180 --> 13:17.960]  So we could, for instance, say that we want the system to be very responsive.
[13:17.960 --> 13:31.500]  We want it to be able to transform fast, and then we have to accept that there is a higher overhead than for a system that doesn't have to be able to transform as rapid.
[13:32.060 --> 13:38.520]  For instance, this is... we could actually turn it from zero up to some form of max value.
[13:39.160 --> 13:44.640]  We could build a brick wall with a system like this, and then we can say, the wall is not going to change for a very long time.
[13:45.280 --> 13:47.840]  So let's just remove the robot that built the wall.
[13:48.180 --> 13:52.320]  So it can... at that time, it's not longer a self-reconfiguring system.
[13:52.460 --> 13:53.220]  It's just there.
[13:54.040 --> 13:57.100]  And in another solution, we might want to change it faster.
[13:57.600 --> 14:06.860]  And then we say, okay, there is a certain overhead with this, but we're going to keep a number of these reconfiguring units in the system, but it can also then change.
[14:07.160 --> 14:19.780]  So when it comes to the argument of overhead of modularity, I usually say that it is the lowest form, the lowest level of overhead that should be compared to a custom-made part.
[14:19.940 --> 14:25.920]  Because anything more than the lowest level is a feature, not a cost.
[14:26.440 --> 14:27.180]  It's a bug.
[14:27.460 --> 14:28.840]  It's a feature, not a bug.
[14:31.860 --> 14:41.900]  So that argument can be quite well-handed, especially if you then consider the fact that there is an overhead of unmodularity or non-modularity too.
[14:42.180 --> 14:49.360]  Because, of course, you have to make that custom component, and that's much harder than making just more of the same thing you already know how to do.
[14:52.840 --> 15:03.060]  To make this system where we use the modules we need at a specific time work the best way it can, we need to pool our modules.
[15:03.320 --> 15:09.040]  And we need to draw modules from a pool when we need them and return them to the pool when we don't.
[15:09.160 --> 15:10.620]  Basically like car sharing.
[15:11.220 --> 15:17.360]  This, though, is both an opportunity and a problem, because the pool will be a common thing.
[15:18.040 --> 15:25.740]  And there's many issues around privacy, security, access to modules, and what you can actually do with them.
[15:25.880 --> 15:30.540]  I mean, if you want to do something that is likely to destroy the modules, are you allowed to?
[15:30.900 --> 15:32.280]  And stuff like that.
[15:32.440 --> 15:33.820]  So pooling is an issue.
[15:34.160 --> 15:37.600]  It's an opportunity, but it's also an issue.
[15:41.950 --> 15:45.830]  Self-reconfiguring modular robotics is going to address many things.
[15:45.830 --> 15:52.750]  Basically, in the end, we see that everything physical around you that you can touch will be built by these systems.
[15:54.190 --> 15:55.970]  That's far off in the future.
[15:56.250 --> 16:01.870]  But even if we just do 10% of what you see around you, which is certainly possible, especially...
[16:02.870 --> 16:04.310]  I can mention architecture.
[16:04.330 --> 16:07.850]  We're making great headway in architecture and house building.
[16:10.050 --> 16:13.130]  This means this is going to upset a lot of people.
[16:13.430 --> 16:20.870]  I mean, when so much changes, and there are ways of arguing for the fact that it's going to be a very rapid change.
[16:22.230 --> 16:33.450]  The only solution I have there is to be open, to allow everybody access to the systems, to allow education about these systems for everyone, free.
[16:34.550 --> 16:43.770]  And, of course, as level playing field as possible, so that everyone can participate and do stuff with these new technologies.
[16:47.610 --> 16:50.550]  So what does this mean for the hacker community?
[16:50.690 --> 16:51.330]  We're here at HOPE.
[16:51.530 --> 16:59.090]  I mean, most of us like to do stuff and to tinker with things and to just not consume, but also kind of produce.
[17:04.940 --> 17:11.920]  There are many things that I think that the hacker community would love with modular robotics.
[17:13.380 --> 17:20.740]  Modules are reasonably cheap as compared to custom parts that are made in small series or stuff like that.
[17:20.940 --> 17:24.460]  These can be made in huge series and will then drop in price.
[17:24.660 --> 17:27.460]  And, of course, you only pay for them when you actually use them.
[17:27.460 --> 17:36.000]  So you don't have to buy them in the same way you do with general components that it's hard to know if somebody wants them when you're done with them.
[17:41.830 --> 17:55.250]  And it's also so that since they are so general, if you wanted to develop work with the software part, which is probably 90% of what we're talking about, the hardware of modular robotics is very hard, but you only have to do it once.
[17:55.510 --> 17:59.390]  When somebody builds a connector that works okay, we've done that.
[17:59.390 --> 18:00.950]  I mean, just make more of it.
[18:01.150 --> 18:04.050]  But the software part is going to be a huge work.
[18:04.670 --> 18:16.590]  And then you can start with developing a user simulator, you debug, you've got your system working okay, but you have to test it on the physical modules to see if it works in real life.
[18:16.770 --> 18:19.630]  And I can tell you there's a very big step between those two.
[18:19.850 --> 18:26.950]  A lot of modular robotic system works great in simulation, very few works at all in reality.
[18:27.970 --> 18:37.610]  But since they're all the same, it would be very easy for somebody, and I'm actually working with some people to try to do that, to provide the modules for testing purposes virtually, remotely.
[18:37.870 --> 18:50.030]  So you do your simulation work, you do your development, you go onto this website, you can test them against the modules and see whether they work, get feedback from what doesn't work, and do it that way.
[18:50.170 --> 19:01.810]  And that means that you'd only pay for the little bit of time, which is going to be ridiculously cheap, just like web hosting today, or all the things we take for granted in the web world.
[19:04.950 --> 19:08.330]  As I said, experimentation costs next to nothing.
[19:08.550 --> 19:11.370]  The modules are cheap, and you just get them when you need them.
[19:12.630 --> 19:28.630]  Another factor is that when you want to start doing something with modular robotics, here you get the advantage of it not being custom components, because it's going to be leveraged in, as I see, three ways.
[19:28.870 --> 19:36.290]  The first way, of course, is that you have this ecosystem of all the modules that somebody ever did, and you can just use those.
[19:36.510 --> 19:45.370]  And it's easy to incorporate what somebody did, because the module, the connector works as a black box interface, basically.
[19:45.690 --> 19:50.270]  If somebody's module works, it's just take it and run with it.
[19:57.010 --> 20:01.170]  The second thing that I find to be a big leverage is
[20:05.960 --> 20:12.540]  that somebody's probably had the same problem, or wanted to work with something similar to what you wanted to.
[20:12.790 --> 20:23.560]  And since the module parts are general, you could probably more easily build on what they did than if they build it out of custom components from scratch every time.
[20:23.710 --> 20:30.030]  Because there's going to be these small, nagging, real-world issues that makes it impossible for you to use what they did.
[20:30.320 --> 20:37.230]  And of course, when you're done and you've got something working, your module will go into the pool, and everybody could just use that.
[20:37.710 --> 20:40.390]  And they could use it as it is, or they could modify it.
[20:41.370 --> 20:46.710]  And of course, you could just share this with the world and win great karma, or you could sell it if you wanted.
[20:51.520 --> 20:56.000]  To end this talk, I would like to come back to this question again.
[20:56.200 --> 21:09.680]  But this time, I'm going to just say that if you had this system that is low-cost, environmentally friendly, virtually unbreakable, the modules replace themselves if they break.
[21:10.080 --> 21:11.460]  They're self-reconfiguring.
[21:12.100 --> 21:14.880]  Could form any physical structures, perform any tasks.
[21:15.140 --> 21:16.800]  What value could you derive from that?
[21:17.700 --> 21:24.360]  I hope that I've put some ideas in your head and that you think that it could be cool to try this and that out.
[21:24.620 --> 21:25.780]  Because it's very easy.
[21:26.000 --> 21:30.260]  We could try out many, many, many different things here just during the two days at home.
[21:34.720 --> 21:36.280]  Because it's basically free.
[21:36.980 --> 21:41.500]  When we leave here, the modules will be exactly the same as when I arrived.
[21:41.880 --> 21:43.980]  So there's no cost associated with that.
[21:44.340 --> 21:46.180]  Except a little rent, I guess.
[21:47.340 --> 21:49.900]  Somewhere, somewhere, somebody had to pay for them.
[21:52.820 --> 21:57.980]  I would really like for you to help me and others bringing this out into the real world.
[21:57.980 --> 21:59.900]  It's just about to come there.
[22:00.080 --> 22:02.960]  This one you see is made with rapid prototyping.
[22:03.140 --> 22:04.600]  But it is commercially available.
[22:04.900 --> 22:07.340]  The one for kids is also commercially available.
[22:08.040 --> 22:13.280]  And more solid systems where you could build real structures.
[22:13.280 --> 22:15.120]  These snap together and that's great.
[22:15.340 --> 22:21.260]  But there is systems coming out that you could build these chairs out of easily on this table easily.
[22:21.960 --> 22:23.280]  And they're still active.
[22:23.420 --> 22:26.860]  They still have motors or sensors and an operating system.
[22:26.860 --> 22:28.380]  So you could also program them.
[22:28.600 --> 22:32.780]  I'm looking forward to building a go-cart out of those when I get them.
[22:33.460 --> 22:36.120]  Which is sadly enough probably towards the winter.
[22:36.280 --> 22:38.380]  Which means I'll have to build a Ski-Doo instead.
[22:38.520 --> 22:39.500]  Because I live in Sweden.
[22:40.560 --> 22:42.600]  So that will be a modular Ski-Doo.
[22:44.460 --> 22:48.620]  There are a few open-source projects that you can contribute to.
[22:48.620 --> 22:50.260]  that you can use.
[22:50.860 --> 22:53.340]  None of them are self-reconfiguring, I'm sorry to say.
[22:53.540 --> 22:54.160]  They are modular.
[22:55.240 --> 23:03.480]  I'm working on a project that would probably, hopefully, make it possible to create an open-source.
[23:03.980 --> 23:05.140]  Free for everyone.
[23:05.420 --> 23:07.300]  Free as in speech and beer.
[23:07.540 --> 23:09.800]  I guess, except for the modules.
[23:10.340 --> 23:12.780]  But as I said, they can be very cheap.
[23:13.800 --> 23:16.720]  But more important than free maybe is open.
[23:17.180 --> 23:19.760]  You will have a bill of rights with access to them.
[23:19.920 --> 23:22.000]  And you will have a set of responsibilities.
[23:22.620 --> 23:28.860]  And the community as a whole would work with setting those rights and obligations.
[23:29.820 --> 23:36.080]  I'll be putting up some links to these open-source projects on my blog tomorrow.
[23:36.520 --> 23:42.240]  And also a number of videos where you can see the self-reconfiguring systems in action.
[23:42.440 --> 23:45.400]  Because they actually do work quite nicely in the lab.
[23:45.560 --> 23:47.780]  Which, of course, is like saying they don't work at all.
[23:48.020 --> 23:53.840]  But it's still cool to see them transform from a snake to a dog, basically.
[23:53.840 --> 23:58.140]  You can see them coming, walking along as a four-legged creature.
[23:58.460 --> 24:01.460]  Detecting that the slope is too high for them to do that.
[24:01.620 --> 24:02.780]  And then they become a snake.
[24:02.960 --> 24:05.420]  And then they do the incline as a snake.
[24:05.660 --> 24:08.160]  And then they realize that now there's flat ground.
[24:08.260 --> 24:10.120]  And they go over to walking like a dog again.
[24:10.320 --> 24:12.360]  So there's a lot of cool videos out there.
[24:13.280 --> 24:19.500]  I'm also going to put up some resources where you can find more information.
[24:19.500 --> 24:21.600]  It's very heavily research-based.
[24:21.600 --> 24:25.300]  Because most of this development is done in the university environment right now.
[24:25.660 --> 24:28.740]  But I think it is ready to come out into the...
[24:28.740 --> 24:31.280]  It's garage ready is what I'd like to say.
[24:31.500 --> 24:33.920]  You can actually tinker with this in your garage.
[24:34.060 --> 24:36.020]  And you could do it tomorrow.
[24:36.500 --> 24:39.580]  And then contribute to the pool.
[24:39.780 --> 24:41.440]  And everything just grows exponentially.
[24:43.380 --> 24:44.640]  So, any questions?
[24:45.380 --> 24:46.340]  Have I said... Yeah?
[25:00.290 --> 25:04.150]  Actually, scaling up from the chair is what we think is...
[25:04.150 --> 25:05.210]  It's not easy.
[25:05.350 --> 25:11.290]  The requirements list for the connector, which of course is what will carry the load, is like 50 different requirements.
[25:12.770 --> 25:14.570]  It wouldn't use the same...
[25:14.570 --> 25:18.470]  Because satisfying all those 50 in one connector is just ridiculously impossible.
[25:19.010 --> 25:24.230]  But say that we wouldn't use the same connector, but we could use one connector for the chair.
[25:24.490 --> 25:26.050]  And then one connector for the house.
[25:27.370 --> 25:30.610]  And yes, there are alternatives for both.
[25:30.610 --> 25:34.150]  To make a really, really small connector is actually harder.
[25:35.130 --> 25:40.450]  One that could become sub-millimeter and we could do not only a working chair, but a beautiful chair.
[25:40.650 --> 25:43.790]  The one that we can look at and see, okay, it looks like a chair.
[25:47.010 --> 25:48.550]  Yeah, not one...
[25:48.550 --> 25:50.530]  That's the holy grail, basically.
[25:50.730 --> 26:01.430]  To make one connector that not only scales from the really small to the really big, but that the scales are also compatible so that you could attach something that was 10% smaller to something.
[26:01.610 --> 26:04.130]  And then something that 10% smaller than that.
[26:04.290 --> 26:09.050]  And then suddenly you go all the way down to Mr. Feynman's levels.
[26:10.950 --> 26:13.030]  There is plenty of room at the bottom, I guess.
[26:13.630 --> 26:15.310]  But yeah, that's a very good question.
[26:15.430 --> 26:21.870]  But yes, theoretically and practically, going up from the chair is not an issue.
[26:25.100 --> 26:25.780]  Any other?
[26:27.400 --> 26:55.300]  tourists from China?
[27:01.020 --> 27:06.580]  Did you see sorts of things like that?
[27:16.300 --> 27:20.660]  Yeah, what we call that is basically what we're doing today.
[27:21.360 --> 27:26.260]  If we would want to do robotic architecture, we wouldn't start building the whole house.
[27:26.440 --> 27:33.040]  What we would do is, for instance, these walls that are behind me is actually a modular system.
[27:33.220 --> 27:37.920]  I mean, now we can have one big room or we can have two small rooms or we can adapt our environment.
[27:38.140 --> 27:49.600]  So doing what you say is that the coexisting of regular technology with this is absolutely... that's certainly something that would be where we would start.
[27:51.980 --> 27:58.380]  Use... basically what you build a house like this, you build the floors and you build the walls, but you leave everything open.
[27:58.860 --> 28:05.160]  I have an example there that is, if you did that to your apartment, it would be more significant than you think.
[28:05.320 --> 28:08.660]  Say that you had all the space just in one open space.
[28:09.920 --> 28:20.520]  And then you had a... then a reasonably simple modular robotic self-configuring modular robotic system, able to put up walls wherever you... you just point, I want a wall there and it builds a wall there.
[28:20.940 --> 28:29.380]  Say that you had this apartment, and since you could use the space for whatever you want, you don't have to buy as much.
[28:29.740 --> 28:36.020]  So you'd actually afford to buy an apartment in that good neighborhood that you always wanted, but you couldn't because it's very expensive.
[28:37.000 --> 28:41.680]  But since you now can adapt your space to whatever you need, there's no problem.
[28:41.680 --> 28:45.160]  Say that you buy... you're going to live there with your significant other.
[28:45.320 --> 28:50.200]  You buy, I guess in square feet, I don't know, 40 square meters maybe.
[28:50.560 --> 28:52.260]  And you want to have a dinner party.
[28:52.400 --> 28:53.600]  This is the example I'm running.
[28:53.920 --> 28:55.620]  So you're going to have a dinner party.
[28:55.740 --> 28:57.440]  You need a big kitchen to prepare the food.
[28:57.560 --> 28:58.760]  So you have a big kitchen.
[28:58.940 --> 29:02.180]  You prepare the food and after that, the food is done.
[29:02.320 --> 29:05.780]  It's in what is... it's a reasonably big fridge and freezer.
[29:06.140 --> 29:11.100]  You don't usually have that size fridge freezer because you're having guests and you need a lot of food.
[29:11.100 --> 29:15.460]  You need two bathrooms because you're going to take a shower.
[29:16.440 --> 29:18.280]  Your guests start to arrive.
[29:18.280 --> 29:22.840]  You need one big space for socializing.
[29:23.040 --> 29:26.500]  And you want just two toilets, no showers, just two toilets.
[29:26.760 --> 29:29.660]  Or even four because you don't want lines to form.
[29:30.240 --> 29:33.560]  Then you need a table so that everybody can sit down.
[29:33.800 --> 29:35.020]  You get a table.
[29:36.260 --> 29:41.840]  And when the dinner is over, a couple of your friends want to spend the night because you're going to go shopping or whatever tomorrow.
[29:42.900 --> 29:45.020]  And then, of course, you need three bedrooms.
[29:45.360 --> 29:49.200]  And within this space, you could do all that space-wise.
[29:49.560 --> 29:55.100]  But it would be very expensive to have all of these just standing, waiting for you to use them.
[29:55.100 --> 30:01.580]  So if we could use our whole space all of the time, it would be very efficient.
[30:02.160 --> 30:05.000]  But as I said, yeah, that's definitely what we're going to do.
[30:05.080 --> 30:06.360]  We're going to allow them to co-exist.
[30:08.100 --> 30:09.080]  Any other questions?
[30:20.660 --> 30:22.940]  Yes, the systems are self-healing.
[30:32.540 --> 30:38.020]  The modules themselves would probably not be self-healing, but the solution would be the whole.
[30:38.900 --> 30:41.720]  If your chair contained a broken module.
[30:42.000 --> 30:48.280]  For instance, the chair could have a feature that detected the length of the person sitting on the chair and adapting to it.
[30:48.440 --> 30:52.040]  So that even if you're four foot two, you sit comfortably.
[30:52.240 --> 30:57.400]  If you're six foot four, you sit comfortably because the chair detects how long you are and adapts to you.
[30:59.860 --> 31:09.700]  And, of course, if it then had a broken module, that module could be automatically replaced and then, I guess, also automatically returned to the factory.
[31:09.960 --> 31:15.760]  And there you have to take a decision whether to reuse the parts or recycle the material in the parts.
[31:15.760 --> 31:21.200]  But the modules themselves would probably not be self-healing on that level.
[31:22.000 --> 31:23.460]  This is actually also... yeah?
[31:26.020 --> 31:26.880]  No, no.
[31:27.000 --> 31:30.900]  That's also the difference between the self-reproducing machine.
[31:31.920 --> 31:44.520]  This could become a self-reproducing system because there's nothing stopping you from building a mining truck or any kind of system needed to make the modules themselves.
[31:44.520 --> 31:49.660]  So, it's not a self-reproducing machine, but it's a self-reproducing system.
[31:50.500 --> 31:50.900]  Please?
[31:52.160 --> 31:53.460]  I'll start by the name.
[31:53.660 --> 31:54.600]  I'm a bit of a skeptic.
[31:54.800 --> 31:55.080]  Yeah?
[31:56.160 --> 32:00.380]  Architects are very skeptical about this because they tried before and it's very hard to get it to work.
[32:00.620 --> 32:00.820]  Sure.
[32:04.150 --> 32:06.930]  In electrical... not a lot of sleep, but a lot of weight.
[32:07.290 --> 32:12.330]  And I can see a lot that there's work routine in what you're saying.
[32:12.530 --> 32:12.750]  Yeah?
[32:12.750 --> 32:14.870]  I actually have a lot of competition to...
[32:15.590 --> 32:16.650]  I'm here for two days.
[32:16.650 --> 32:24.230]  Just right away, on the surface, there are two things that really stand up to me.
[32:24.650 --> 32:26.290]  And one is that the system...
[32:27.670 --> 32:34.870]  This idea is very related to the idea of 3D printers and repetition and all that sort of thing.
[32:35.090 --> 32:44.530]  And some of the problems there are the design is expensive in terms of these virtual resources and are patentable eventually as well.
[32:46.870 --> 32:50.390]  And these objects are... take a long time to be familiar right now.
[32:50.770 --> 32:54.030]  There are lots of proposal and feasible issues and reputation.
[32:54.470 --> 33:00.610]  Then on the other side, the other... a lot of the characteristics that you're talking about already exist in nature.
[33:00.610 --> 33:13.570]  So this... you know, the cell as a monocle... composting as a... as a self-fueling and self-adventing system is kind of a philosophical basis perhaps of what you're discussing.
[33:14.370 --> 33:20.290]  And it, you know, quote-unquote, doesn't have value because we choose in our economics not to value it that way.
[33:20.290 --> 33:27.230]  So, I guess I can see some gaps in making this...
[33:31.320 --> 33:32.420]  Yeah, yeah.
[33:32.640 --> 33:38.420]  How... how much... how much does that go into the design or the... the design phase of all this stuff?
[33:39.100 --> 33:42.840]  Well, I'd like to say that... that we learn an enormous amount from...
[33:42.840 --> 33:46.260]  because as you say, that we are self-reconfiguring modular robot.
[33:46.420 --> 33:49.640]  Every one of you that sits here is a self-reconfiguring modular robot.
[33:50.260 --> 33:52.780]  You're... oh, well... robot is another thing.
[33:52.920 --> 33:54.260]  But you're self-reconfiguring anyway.
[33:54.660 --> 33:57.240]  And you're definitely self-healing, which is very nice.
[33:57.340 --> 33:58.980]  Otherwise, all of you would be dead.
[33:59.260 --> 33:59.860]  And me too.
[34:00.020 --> 34:02.840]  The smallest cup would kill us immediately if we weren't self-healing.
[34:03.940 --> 34:07.720]  So what I'd say is that... for the... I think it was the first question was that...
[34:08.600 --> 34:11.140]  Yeah, we learn an enormous amount from... from nature.
[34:12.000 --> 34:16.540]  How to control these large groups of... of modules is very hard.
[34:16.680 --> 34:19.620]  It's not... you can't just write line code saying do this to...
[34:19.620 --> 34:19.900]  do that.
[34:20.040 --> 34:21.800]  It's... becomes impossible immediately.
[34:22.260 --> 34:24.600]  They use pheromone control.
[34:25.120 --> 34:29.840]  Basically like ants control the... the... the... the hive or the... the... the... the ant hive.
[34:30.980 --> 34:35.460]  They also use... actually one of the... the architecture projects.
[34:35.520 --> 34:39.900]  They build houses with the same... or what we think is the same algorithms as...
[34:40.620 --> 34:41.880]  termites build their nests.
[34:42.100 --> 34:45.380]  Termites are not very smart creatures, but they build...
[34:45.380 --> 34:46.700]  fantastic nests.
[34:46.920 --> 34:48.500]  They maintain them very nicely.
[34:48.500 --> 34:51.900]  and they can grow the nest while the nest is in use.
[34:52.380 --> 34:57.160]  So say that... that'd be like moving into a one-bedroom apartment and living there...
[34:57.160 --> 35:00.300]  while the construction worker made it a two-bedroom apartment.
[35:01.140 --> 35:03.220]  And of course fix the roof or whatever.
[35:03.460 --> 35:03.500]  Yeah?
[35:03.900 --> 35:05.340]  And wait, you had a second question.
[35:05.760 --> 35:07.860]  And... could you... could you... do that again?
[35:09.660 --> 35:12.950]  I don't have a specific question, but that's the... the first part of...
[35:13.540 --> 35:16.180]  Yeah, you talked about it being related to 3D printing.
[35:16.380 --> 35:16.540]  Yeah.
[35:16.660 --> 35:23.000]  You can do the module street... you... yeah, you can... you can do it 3D printed, but it's not necessarily...
[35:23.000 --> 35:27.660]  I think it is actually... maybe I shouldn't say this here, but it's actually competing with 3D printing.
[35:27.660 --> 35:33.840]  Because 3D printing is... I... I know that they're trying to make it a... a full circle thing where... where they...
[35:33.840 --> 35:38.060]  where they print and then could... granulate the plastic again and redo it again.
[35:38.260 --> 35:40.940]  But... we kind of skipped that step.
[35:41.860 --> 35:48.440]  They can do very nice orbit with... like... very nice position, but... but they have to go through the printer.
[35:48.440 --> 35:53.800]  We can... just reassemble the modules... without going through the printer.
[36:02.040 --> 36:08.720]  But... but... the idea that... to replicate an object... you need... you need design... and that design is very... heavy in intellectual resources.
[36:09.000 --> 36:09.400]  Yeah, yeah.
[36:09.780 --> 36:10.480]  So... we can...
[36:10.480 --> 36:12.340]  I got the perfect answer for you right here.
[36:12.520 --> 36:13.140]  You know what I mean?
[36:13.340 --> 36:13.640]  Yeah.
[36:14.620 --> 36:16.040]  Again, we learn from nature.
[36:17.840 --> 36:24.020]  We... along... we've tried to do... self-organizing... systems for a long time.
[36:24.380 --> 36:27.100]  To... to evolve... solutions to our problems.
[36:27.100 --> 36:30.420]  The way... we evolved from... from... from the environment.
[36:30.700 --> 36:31.620]  Darwinian, basically.
[36:31.900 --> 36:35.980]  But it's very hard to do... when... when our components are... too large.
[36:36.500 --> 36:41.080]  So... there is actually... two professors in Japan... that just published a book on this.
[36:41.200 --> 36:44.660]  It's called Self-Organizing... Self-Organizing Robots.
[36:44.660 --> 36:45.920]  It's a fantastic read.
[36:47.800 --> 36:50.340]  Surprisingly readable... for a book of that nature, actually.
[36:51.480 --> 36:56.080]  And... what... they... talk about... to... to... they talk about a lot of stuff in the book.
[36:56.180 --> 37:03.580]  But what they... talk about is that... instead of saying... I want to draw a chair... it should look exactly like this... behave exactly like that...
[37:03.580 --> 37:05.060]  I define a set of rules.
[37:05.380 --> 37:11.980]  And then I... preferably start in simulation... by evolving something that looks... reasonably close to what I want.
[37:12.540 --> 37:21.360]  And then, of course, I can put it out there... in the real world... and... allow... the components... to evolve... a chair... that fits my needs.
[37:21.360 --> 37:27.400]  This is also very important because... self-frequent... self-healing comes from... the self-organizing.
[37:27.600 --> 37:31.780]  Because it simply detects it's not fulfilling the goal... and then it adapts to that.
[37:32.100 --> 37:38.260]  And... we've seen that in simulation... and we... also to a certain degree in hardware... the connector is still plaguing us.
[37:38.440 --> 37:41.480]  The bad connectors we have are still a... very hard problem.
[37:42.820 --> 37:49.540]  But in simulation... we see that... Gates, for instance... how robots work... evolve very nicely.
[37:50.080 --> 37:55.840]  And many other things just evolve... if you allow... if you give it... the ecosystem to do so...
[37:55.840 --> 37:57.820]  We can evolve the products we do.
[37:58.000 --> 38:00.560]  So, yeah... design is... fantastically important.
[38:01.240 --> 38:08.180]  And with... a self-reconfiguring module... module robotic system... we could do design in a fundamentally different way.
[38:09.100 --> 38:11.300]  But I'd definitely like to talk to you more.
[38:11.480 --> 38:13.200]  So, just... I'm outside when I'm done.
[38:13.280 --> 38:16.200]  And I'm going to be here tomorrow... and Sunday too.
[38:16.420 --> 38:20.520]  And I'm going to have the robots here for you to... to try different setups.
[38:22.700 --> 38:22.820]  Yeah?
[38:23.440 --> 38:26.000]  What about the complexity side?
[38:26.660 --> 38:26.940]  Yeah.
[38:27.920 --> 38:28.420]  Yeah.
[38:29.400 --> 38:29.540]  Yeah.
[38:30.640 --> 38:30.740]  Yeah.
[38:30.900 --> 38:31.880]  Cellular Automaton.
[38:32.080 --> 38:35.040]  This area draws on some... heavy names.
[38:35.040 --> 38:38.020]  And it's... you mentioned the Cellular Automaton.
[38:38.320 --> 38:39.740]  Which is von Neumann's work.
[38:40.000 --> 38:42.780]  And he actually... he is the father of this work.
[38:42.920 --> 38:44.440]  With the Cellular Automaton.
[38:44.780 --> 38:47.520]  The sad thing is he died before he finished that work.
[38:47.660 --> 38:48.680]  He died very young.
[38:49.080 --> 38:50.060]  I think he was 56.
[38:50.800 --> 38:54.640]  There is speculation that he was exposed to radiation during the Manhattan Project.
[38:54.880 --> 38:55.840]  Which he, of course, worked on.
[38:56.500 --> 38:57.180]  But yeah.
[38:57.320 --> 39:00.800]  The Cellular Automaton is the basis for our research.
[39:00.980 --> 39:02.020]  For our development.
[39:02.020 --> 39:03.820]  Then comes...
[39:06.100 --> 39:09.240]  Penrose, the mathematician, did important work.
[39:11.580 --> 39:16.760]  And the modern version of this research was started by a man called Toshiro Fukuda.
[39:16.980 --> 39:21.600]  When he was working on the CNC machine with interchangeable tools.
[39:22.120 --> 39:24.640]  So he... he was working on making the robot...
[39:24.640 --> 39:26.000]  The machine...
[39:26.000 --> 39:26.840]  Collect the tool.
[39:27.040 --> 39:28.220]  A drill of a certain dimension.
[39:28.540 --> 39:29.400]  Use it for a while.
[39:29.480 --> 39:30.580]  Return it to the storage.
[39:31.140 --> 39:33.220]  Pick up another tool and do something with it.
[39:33.320 --> 39:36.180]  And he said, instead of you doing it just for the tool...
[39:36.180 --> 39:38.040]  Why can't I do it for the whole machine?
[39:38.540 --> 39:42.060]  I mean, sometimes I need a strong but reasonably slow engine...
[39:42.060 --> 39:42.400]  Motor.
[39:42.840 --> 39:47.040]  And other times I need a very fast, but not necessarily that strong motor.
[39:47.660 --> 39:52.380]  And if I could put the strong motor back in the pool when I don't need it, somebody else might need it.
[39:52.480 --> 39:54.040]  So he started the work...
[39:55.080 --> 39:56.900]  And this was late 80s.
[39:57.700 --> 40:04.160]  And up till I would say 2000s, progress was very academic, very heavy on math.
[40:04.160 --> 40:09.100]  I tried to read the cellular autonomous papers and they're just impossible to understand for me.
[40:09.800 --> 40:11.220]  But I think...
[40:11.220 --> 40:13.960]  I feel now there are commercial projects.
[40:14.840 --> 40:19.440]  The basic fundamental algorithms and stuff are reasonably good.
[40:19.900 --> 40:23.760]  So I think that five years from now, this is garage ready.
[40:24.020 --> 40:27.020]  Five years from now, everybody can tinker with this in their home.
[40:27.200 --> 40:29.160]  And you could start already now.
[40:29.160 --> 40:35.020]  Because it's actually much easier to build a module and then build a chair out of modules, than it is to build a chair from start.
[40:35.920 --> 40:37.980]  Especially if somebody else did the module already.
[40:41.270 --> 40:41.990]  Do you think...
[40:41.990 --> 40:42.610]  Do you see...
[40:42.610 --> 40:45.110]  Do you see a central contract with a really impressive model?
[40:45.450 --> 40:47.770]  And this process is very commercialized.
[40:48.070 --> 40:50.510]  Because what a version of what we do is we look...
[40:50.510 --> 41:06.730]  And that means to hear a lot that we can sell consumers for license, for a type of cell that they've got to use themselves, rather than being a critical thing that we live in a lot of companies, and other companies that basically push through the social concept.
[41:06.910 --> 41:12.910]  Do you think, therefore, this data is possible to be agile, Yes, yeah.
[41:13.210 --> 41:17.950]  And not only... I mean, there's actually worse guys than the commercial guys because you can deal with those.
[41:17.970 --> 41:19.650]  If you give them money, they're nice to you.
[41:19.810 --> 41:29.130]  But if the government does that to you and your government isn't nice, I don't want to live in a world where a bad government controls these things.
[41:29.250 --> 41:31.330]  So, yeah, there's definitely privacy issues.
[41:31.550 --> 41:36.510]  There's so many issues around safety and security and your control over your environment.
[41:36.550 --> 41:37.590]  Yes, absolutely.
[41:38.990 --> 41:51.170]  I don't see this as a conflict because even though we're going for a high resolution heterogeneous system, that's way, way off in the future.
[41:51.910 --> 42:01.050]  So, I think that actually I'm working on a project where we're going to set up an environment where you could send your designs for a module in and we make it for you.
[42:01.050 --> 42:06.390]  So, you don't have to be an expert in machining or 3D printing or whatever.
[42:06.610 --> 42:10.270]  You just have to be somebody that wants to contribute and we take care of it.
[42:10.450 --> 42:12.410]  Because that's a challenge in modular robotics.
[42:12.410 --> 42:15.570]  When you do a prototype, generally, you do one or two.
[42:15.950 --> 42:20.150]  But to do anything significant, you have to have 10 or 15 or even 1,000.
[42:20.590 --> 42:27.530]  And when you do 1,000 of something, you do too many of them to be done by hand.
[42:28.050 --> 42:30.930]  And you do too few of them to do with them in industry.
[42:31.210 --> 42:32.890]  So, that's a big challenge.
[42:33.090 --> 42:36.470]  But yeah, there's definitely many issues around privacy, security.
[42:38.030 --> 42:40.550]  And I don't think that the commercial guys are the big problem.
[42:40.730 --> 42:41.570]  We can deal with those.
[42:41.730 --> 42:44.610]  But governments are certainly a problem.
[42:46.710 --> 42:54.430]  But I think there's an opportunity to do it better this time than we did maybe with computers.
[42:54.430 --> 42:57.470]  Because we all are out here and we're doing stuff.
[42:57.750 --> 43:00.750]  It's not done by large institutions and the government.
[43:01.010 --> 43:02.170]  We could do this together.
[43:02.550 --> 43:05.010]  And then, of course, we will also own it.
[43:05.870 --> 43:06.850]  So, yeah.
[43:08.170 --> 43:08.650]  Yeah?
[43:19.300 --> 43:20.220]  Yeah, yeah, yeah.
[43:20.400 --> 43:24.000]  I had the pleasure of interviewing the guy for my podcast who worked on those.
[43:24.160 --> 43:25.020]  They are just amazing.
[43:25.900 --> 43:28.580]  Should I introduce permanent magnet and then you can go on?
[43:29.040 --> 43:34.760]  We know about the regular magnet and we know about the electromagnetic effect.
[43:34.980 --> 43:37.240]  But if you turn power on, it becomes magnetic.
[43:37.880 --> 43:39.980]  And when you turn the power off, it's not magnetic.
[43:39.980 --> 43:44.200]  That's a terrible thing of doing it because it gets hot, it takes a lot of power, all of those bad things.
[43:44.380 --> 43:47.120]  And the permanent magnet we can't, of course, use because we can't turn it off.
[43:47.340 --> 43:49.300]  But he mentions electropermanent magnets.
[43:49.460 --> 43:55.960]  And they are a form of magnet that when you give them a pulse of energy, they toggle the state between magnetic and not.
[43:56.960 --> 44:00.360]  So a short burst of power, it becomes magnetic.
[44:00.600 --> 44:02.860]  Short burst of power, it's not magnetic anymore.
[44:03.160 --> 44:04.660]  It's like magic, I guess.
[44:04.660 --> 44:10.940]  And yeah, so now everybody knows what a permanent electromagnet is, which I guess some of you already know.
[44:11.080 --> 44:12.040]  So, on with the question.
[44:44.710 --> 44:48.250]  Yeah, if you remove the micro from the name, it's already here.
[44:48.250 --> 44:53.450]  Yeah, and MIT in Boston, they've got... he works on the electromagnet.
[44:53.590 --> 44:54.810]  That's where they use it.
[44:54.970 --> 44:56.070]  It's called the robot pebble.
[44:56.330 --> 44:57.890]  It's 11 millimeters cube.
[44:58.730 --> 45:03.390]  And they use stochastic self-assembly.
[45:03.390 --> 45:07.290]  So what you do is you load every module with a schematic.
[45:07.670 --> 45:10.790]  It's not... it's complicated, but let's not go into that right now.
[45:11.050 --> 45:12.870]  You put them in a bag, you shake them.
[45:13.030 --> 45:17.950]  And when the modules hit and know, okay, this is correct hit, this is not correct hit, they assemble.
[45:18.890 --> 45:21.050]  So then you can pull your object out of the bag.
[45:21.330 --> 45:24.250]  You could also... so you just assemble them into one big block.
[45:24.610 --> 45:29.070]  And it says, every module decides whether I'm a part of the solution or not.
[45:29.250 --> 45:32.090]  If I'm not part of the solution, I turn my magnets off and I fall off.
[45:32.090 --> 45:37.510]  So from a cube, you get a dog by loading the shape of the dog into the cube.
[45:37.610 --> 45:40.110]  And everybody else just says, well, I go back to the pool.
[45:40.490 --> 45:41.270]  Call me when you need me.
[45:41.710 --> 45:43.590]  So yeah, certainly the electromagnet...
[45:43.590 --> 45:46.210]  And they are very nice because they scale with size.
[45:46.330 --> 45:47.770]  We talked about size with you before.
[45:48.050 --> 45:50.990]  And they get better and better the smaller they are.
[45:51.110 --> 45:51.970]  Yeah, and as I get
[45:55.940 --> 46:01.640]  the balance of the magnet works in micro-scale, do you think that that's impossible to be a computer or do you think that's...
[46:03.140 --> 46:07.540]  I should actually... I'll talk to the people that know and put it on my blog.
[46:07.720 --> 46:12.320]  But yeah, what I understand is that, yes, it's not that far off.
[46:12.800 --> 46:16.680]  They did an 11 millimeter cube because that's what they could do in the lab.
[46:16.680 --> 46:18.270]  Again, they had to do a lot of them.
[46:18.270 --> 46:19.600]  There is another project.
[46:20.120 --> 46:21.540]  Do you know what a bristlebot is?
[46:21.880 --> 46:25.080]  The one you do on the tip of a...
[46:27.260 --> 46:29.300]  I like hackers, you figure that out by yourself.
[46:29.300 --> 46:31.580]  I don't have to give you a manual or tell you how to do it.
[46:32.000 --> 46:33.790]  That's the test program it runs now.
[46:33.900 --> 46:36.000]  It shows you that all the degrees of freedom work.
[46:37.360 --> 46:40.840]  Don't get your fingers pinched because it's reasonably strong.
[46:43.340 --> 46:44.860]  Yeah, do you know the bristlebot?
[46:45.000 --> 46:47.560]  You do it from a head of a toothbrush.
[46:47.940 --> 46:51.250]  You put a little vibrating motor on there and it runs around on your table.
[46:52.080 --> 46:55.560]  A guy decided to do 1,024 robots.
[46:55.980 --> 47:00.160]  So one single PhD student should build a kilowatt of robots.
[47:01.120 --> 47:02.940]  That's a big project, right?
[47:03.200 --> 47:05.360]  So he had to look at all of these parts.
[47:05.580 --> 47:09.750]  And his previous work was the algorithms, the mathematics behind it.
[47:09.750 --> 47:12.520]  But he decided, I'm going to build 1,024 robots.
[47:12.660 --> 47:13.620]  And guess what he did?
[47:13.840 --> 47:15.160]  They're 15 bucks a pop.
[47:15.300 --> 47:16.860]  So he spent 15,000.
[47:17.060 --> 47:18.940]  But he actually had to build them by hand.
[47:19.140 --> 47:22.080]  And I was saying to him that you're very...
[47:22.080 --> 47:25.360]  He was like saying, oh, they're so hard to build 1,000 of them.
[47:25.440 --> 47:27.420]  And I said, you're the lucky guy.
[47:27.660 --> 47:31.820]  Because the next PhD student, he has to build a megabyte of robots.
[47:31.860 --> 47:33.480]  And that's going to be a pain in the ass.
[47:33.840 --> 47:37.840]  So yeah, scale is very hard, but they're working on it.
[47:39.280 --> 47:49.840]  My view lately has been that we're not going to, as we talked about the chair, we're not going to have the same connector when we build a house as when we build a chair or when we build something really small.
[47:50.100 --> 47:53.380]  We're not going to have the same process of getting the item.
[47:55.000 --> 47:59.750]  But the system as a whole is going to be completely self-reconfiguring.
[47:59.940 --> 48:03.750]  For instance, say that we want to have a really smooth and curved surface of something.
[48:03.750 --> 48:16.860]  We could actually have an assembly unit within the solution that makes components out of really, really small parts and then passes them along at a higher scale at something that could become a bigger thing.
[48:18.140 --> 48:18.520]  Yeah?
[48:20.120 --> 48:31.020]  Could you work on a lot of this, some kind of distribution platform where each one of you is small computer and then if you need more and you just scale it up and you just jump back the same day and then you have...
[48:31.020 --> 48:32.400]  It has to be done that way.
[48:33.200 --> 48:38.640]  I think if we only have like 5 or 10 or 20,000 units, we could probably centrally control them.
[48:39.220 --> 48:46.300]  But when we come down to doing something in a really, really high resolution and we get trillions of units, that's the only way you can do it.
[48:46.440 --> 48:49.940]  There's no way you can centrally control them because timing issues may...
[48:49.940 --> 48:51.620]  I mean, it's simply practically impossible.
[48:51.820 --> 48:55.400]  I meant to do a whole process of just itself.
[48:55.560 --> 49:01.840]  If you could go into your house and use your whole data tray and use more consuming power and just add more to it.
[49:02.180 --> 49:07.060]  Yeah, I guess probably it'd probably be more efficient to use the cloud then, but...
[49:07.060 --> 49:07.920]  Yeah, yeah.
[49:07.920 --> 49:08.040]  Yeah, yeah.
[49:08.300 --> 49:09.900]  But who knows?
[49:10.020 --> 49:21.200]  I mean, to say what you can do with a self-reconfiguring modular robotic system, since it is a universal physical machine, it's basically impossible to say, what can you do with the computer?
[49:21.480 --> 49:24.720]  Well, that depends on how good your programmer you are more than anything else.
[49:24.880 --> 49:26.100]  So, yeah.
[49:26.980 --> 49:27.980]  Any other questions?
[49:30.250 --> 49:34.200]  Have I recruited you to the self-reconfiguring modular robotics project?
[49:34.580 --> 49:34.920]  Yes.
[49:34.920 --> 49:36.960]  Do you like being nice to the environment?
[49:37.860 --> 49:39.900]  And do cool stuff for free?
[49:41.640 --> 49:47.460]  And I mean, if we do this right, I mean, we learned a lot from how computers and the Internet was introduced.
[49:47.580 --> 49:52.020]  If we do it right this time, it could be free as in beer and speech.
[49:52.700 --> 49:58.770]  And the thing that actually costs stuff could be very, very, very, very, very, very cheap.
[49:59.080 --> 50:01.460]  And for us, that's a cool thing to have.
[50:01.600 --> 50:05.340]  But there's a couple of billion people in the world that it would be essential for.
[50:05.560 --> 50:07.290]  So, there's that aspect too.
[50:08.400 --> 50:12.080]  But we have an interesting 20 years ahead of us, I think.
[50:14.320 --> 50:15.100]  So, yeah.
[50:17.100 --> 50:17.660]  Ah, sorry.
[50:17.790 --> 50:18.770]  I didn't see you there in the light.
[50:19.800 --> 50:21.520]  So, this is maybe an unusual question.
[50:21.720 --> 50:21.960]  Nope.
[50:21.960 --> 50:26.060]  So, the current computing technology is very, it's based on discrete systems.
[50:26.270 --> 50:26.680]  It's based right.
[50:26.960 --> 50:29.160]  It's all, it's all, it's hiding the .
[50:29.960 --> 50:30.540]  Mm-hm.
[50:30.800 --> 50:34.620]  And on the other hand, even digitized .
[50:35.200 --> 50:39.720]  Find that they work best when they actually go back to having a continuous behavior.
[50:39.720 --> 50:40.400]  Mm-hm.
[50:40.750 --> 50:43.250]  So, we, we, inside a similar .
[50:44.300 --> 50:48.480]  So, instead of just binary logic, we go with fuzzy logic.
[50:48.790 --> 50:49.140]  Mm-hm.
[50:49.440 --> 50:54.620]  And so, there's this sort of out of nowhere brand of robotics called beambotics, which are .
[50:54.620 --> 50:55.200]  Yeah, yeah, yeah.
[50:55.420 --> 50:55.540]  Yeah.
[50:55.960 --> 50:56.790]  So, what...
[50:56.790 --> 50:57.380]  Mark Tilden.
[50:57.560 --> 50:57.680]  Right.
[50:57.790 --> 50:58.080]  Mark Tilden.
[50:58.270 --> 51:00.400]  I had the pleasure of interviewing him in Hong Kong recently.
[51:00.920 --> 51:00.940]  Ah.
[51:00.940 --> 51:03.900]  So, what is, when you talk about modular robotic systems.
[51:04.140 --> 51:05.790]  Yeah, he's a big mod head, though.
[51:05.940 --> 51:07.640]  He's a big mod head, I can tell you.
[51:07.700 --> 51:07.940]  Absolutely.
[51:08.340 --> 51:08.520]  Yeah.
[51:09.720 --> 51:10.340]  So, you talk about these things.
[51:10.380 --> 51:11.880]  You talk about, like, the screen controller.
[51:12.290 --> 51:12.720]  Mm-hm.
[51:12.820 --> 51:24.040]  Are they internally still running distributed computers, or are they more along the analog and take advantage of the analog systems, and they're just trying to follow, you know, chemical gradients?
[51:24.220 --> 51:26.580]  Are they more that way for a system?
[51:26.800 --> 51:31.500]  I think that we're probably aiming for what you say, the analog systems.
[51:31.750 --> 51:33.750]  They have to be very, very small.
[51:34.020 --> 51:39.540]  And they're, I mean, the CPUs, when you do a thousand of them, and they have to be, as you say, micro scale.
[51:39.770 --> 51:42.460]  We don't have a lot of computer power in each one.
[51:42.640 --> 51:44.100]  So, I'm thinking that.
[51:44.320 --> 51:48.420]  And I think Mark Tilden and the BeamBot certainly advocate for that point of view.
[51:48.540 --> 51:50.380]  And I think that that's valid.
[51:50.560 --> 51:51.920]  I don't think we really know.
[51:52.020 --> 51:55.640]  And I think that, in one way, we can come back to that with different connectors at different scale.
[51:55.640 --> 52:00.120]  Maybe some systems and some parts of systems might want to use one of them.
[52:00.680 --> 52:04.680]  And other system or other parts of a system might want to use both of them.
[52:04.900 --> 52:12.180]  And we could also dynamically adjust where we use each, because we don't build a chair and it ever looks like that.
[52:12.660 --> 52:15.880]  If, yeah, we could just adapt ourselves continuously.
[52:16.300 --> 52:19.000]  It could even, perceivably, do that itself.
[52:20.460 --> 52:29.180]  To evolve a system where 80% is done one way and 20% the other way, it detects that the situation has changed and adapt itself.
[52:35.560 --> 52:35.960]  Anymore?
[52:37.580 --> 52:40.620]  I'm here for two days, so you don't have to ask now if you don't want to.
[52:41.180 --> 52:42.180]  I don't know the time.
[52:42.280 --> 52:43.840]  Maybe the next speaker wants to come in.
[52:44.120 --> 52:44.840]  What's the time?
[52:45.960 --> 52:47.040]  Oh, please, please, please.
[52:48.720 --> 52:49.320]  Thank you.