Scientist Stories: George Whitesides, Soft Robotics

Everybody should have favorites. You should have a favorite member of some sex, right? We have to be dispassionate about it. That's a spouse. You should have a favorite animal, and that's a totem. Mine happens to be rhinoceros because They are extraordinarily stupid animals that solve problems by running through them. And then everyone should have a favorite equation, and I actually have two. One of them is the second law of thermodynamics, but the other is Maxwell's equations. And Maxwell's equations leads me to Ørsted. And I, when I had the chance to come here, regardless of anything else, I would have come because To me, one of the most marvelous experiments that's ever been done is the experiment in which Mr. Ørsted ran a current in this direction and found that a magnet oriented perpendicular to it. I just thought that was a phenomenal experiment, as did everyone else. I still don't understand it, but I think it's a wonderful experiment. So in that spirit, thank you for inviting me, and I'm now going to talk about nothing that has to do with rhinoceri or with Örstedt or anything else. I'm going to talk about robots. Now I want to do this in a slightly unusual way because I'm sure that all of you are accustomed to well-ordered scientific talks. You know, careful thinking and all the rest of it. This is mostly going to be movies. So it's really intended for those of you who are students who no longer read. So, soft robots, what are they? This is fairly easy to explain. My generic question in science is, "Who cares?" And that has to do with robotics. You know, why do you care about robotics? And it's actually a very philosophically interesting question. And then, what you do with them, a little work about applications and how you make them work. Now let me start with this picture. Which is, you say, "What does this have to do with anything?" Or even, "What is it?" Those are two gold coins. And there was a period when competition between nations involved wealth. You know, people would invade one another to get the gold. This is what Spain did with Mexico. And then there was a period in which it was other kinds of resources, and then there were people, and this and that. And now the world has changed, and what you really compete for is knowledge and smart people. That's the basis of what goes on. And the reason for that is not that we have an aesthetic interest in smart people and information, but rather we care about jobs. Because jobs is the basis for economic prosperity in a country. And for those of you who are finishing your studies, something that has to be on your mind, in the back at least, is the question of what comes next. I mean, how do I make a living in a society in which we are, in a sense, our work to a significant extent? What's it going to be? Will there be a job? So it's a very important issue. Now there's a generic problem now with jobs in that In part because of the efficiency of computer systems, there are simply too many people for the number of available jobs. The efficiency, productivity per person, has gone up by about 15% in the last 15 years, and that sort of means that there are too many smart people around for the number of jobs. Now, this problem is compounded, and for those of you who like amusements, you see this discussed discussed in a way that makes one giggle if it weren't so serious in the American political campaign. One way it's phrased is in terms of high wages versus low wages. The argument is the United States and Europe have high-wage jobs, and in China you have low-wage jobs. Well, nah, less and less true. Another is skilled versus unskilled, same kind of argument, and that's certainly not true. You go to many parts of the world and you find the developing world and you find people who are at least as good as anyone in Europe or in the United States. So it's not skilled versus unskilled. Here is the one that I think you need to pay attention to. And it's not people versus people, it's a question of how human beings and machines are going to get along, not in the next 5 years, but in the next 100 years. So that's for you and for your children. This is, I think, the big issue. And one of the key points in this will be this issue here of, are we going to be collaborative, or are we going to be competitive? That is, will machines displace people, or will machines make people able to do jobs that they couldn't otherwise do? How is that going to work? And so the new element in this is robots. And robots are not just machines that operate on an assembly line. We are in process, we humans are in process of constructing a peer competitor, which is robots, that is machines able to do things that we can't do with artificial intelligence systems. They may look at us as meat-based intelligence systems and inferior in that way, but artificial intelligence systems, which may be at least as competent as we are in many of the things that we do. So it's an interesting problem. Here are robots, and you're familiar with all of these in one or another fashion. There are just a couple of points I want to make about them. This is an assembly line, and these are robots that are involved in the assembly. The characteristic of these systems is that you cannot come anywhere near them. There's a fence that you don't see around this, but when one of these arms swings, it's accustomed to holding an automobile, and if you happen to way, it snickers through you as if you weren't there and doesn't even know notice. So these are decidedly not collaborative. The military has, of course, been very good at drones, and a drone is just an airborne robot. This is an effort to make a consumer robot. This is a so-called Roomba, which is an exceedingly stupid and inefficient way of trying to sweep the floor. This is a picture of the so-called da Vinci surgical robot. Which may or may not give better surgery, but it certainly enables higher billing. So it's a good thing in a capitalist universe. This is a robot which is of entirely legitimate use. Human beings cannot presently go here, and I'm going to show you a little bit more about this, the equivalent of this, in just a second. But I want to make this point before we begin, and this has to do with the question of whether robots are intrinsically a bad idea. And I would point to early robots, otherwise known as machines, otherwise known as dishwashers and washing machines, which took half of the population, otherwise known as women, and released them from the task of washing dirty diapers and washing, you know, dirty dishes, and made them into teachers and CEOs and whatever they might be. So this was a case in which machines didn't compete but cooperated. They eliminated an unattractive job and replaced it with much more interesting and attractive jobs. So it can be good. It's a question of how you use these machines and how clever we are. So let me show you an example of a non-collaborative robot. This is what we don't do, but I think it's nice to look at this. This is not what you want as a babysitter. But there's a point here that I want you to look at carefully. Look at the grace with which this thing balances, shifts its weight. It's just unbelievable that— I mean, with my cataracts, I can't tell whether that's alive or not alive. In some ways, it is. Better alive. But the underlying issue in this is that if you look at that kind of development, remember, that doesn't have a brain. All it has is a series of accelerometers and things of that kind. So without any control systems at all, other than the most primitive feedback loops, it does things that I can't tell are not something an animal does. Now you take that and you couple it with artificial intelligence, and all of a sudden you begin to see a pathway to something, not next year, but something in the next 50 years that begins to look a lot like us. And there's a deep, really important underlying question here, which is, if a system becomes sufficiently complicated, that it has the power of advanced information processors, the knowledge of the web, the ability to move autonomously, the ability in various ways to improve itself. It can arguably be set on a pathway for independent evolution. And I don't know whether this is going to happen, but I don't think it's just a question of trivial machines. It's more interesting than that. So, here's where we started. Now you say, "What on earth does that have to do with anything?" With much of technology, it started with the military. And our friends in DARPA were interested in developing a system that looked like a Coke can and would have the characteristic that on command, it would unfold into something that would autonomously walk across the floor find a door, crawl under the door through the crack at the bottom, and then reassemble itself into something that looked like a Coke can on the other side. Now you can see why they might be interested in that. You could understand that I might not be particularly interested in that. But many aspects of the technology of doing that have turned out to be very interesting and very practical to look at. Now to do that, you need always a model, and here is our model. And I just think this is among the most fabulous of organisms. This is an octopus. Octopus. And this octopus is, as you can see, squeezing through a hole much smaller than the octopus, including squeezing its brain, which is apparently pretty plastic. And I would note in this that we couldn't do any of that, or at least without severe damage. So here's something that can do something that we can't do. And, you know, once it does it, I want you to hear Ah, finally it's there. Look at that look of intense self-satisfaction. It's just— But this is a model. And the point in this is that the reason I have a skeleton and stuff like this is that I'm designed to withstand the forces of a gravitational field. And an octopus doesn't need that because there's effectively no gravity if you're buoyant density. But there's no real reason why I need that except for the fact that this was an evolutionary solution. It could be done in other ways. So, you know, we're going to start with things that are fairly soft, but they nonetheless move around quite handsomely in a gravitational environment. Here is the other of our motivations, and that's simply a bicep. It's not— for those of you who lift weights, it's not a particularly impressive bicep. But the fact is that there is nothing, after 50 years of trying, there's nothing that really replicates the function of a muscle in the sense that it's a linear actuator that operates pretty rapidly, pretty efficiently, but doesn't in fact, you know, change volume very much. So we're interested in octopi and we're interested in muscles, not in replicating the mechanism of muscles but the function of muscles. Now let me start with this picture which illustrates a lot of what's going on. That is a raw egg down there, and this is an early gripper. This was the very first thing that we published in this, just to give you a sense of timeline, which is 2011. And the technology itself is as simple as it could possibly be. It's simply a— effectively a balloon over here, an elastomer with an empty chamber on the inside. You pressurize it, and if there is a layer here that has the characteristic that it can bend but not stretch, as this expands in volume, it causes that to bend. I mean, a really trivial idea. And we make this in a straightforward process that I'll show you in a moment. But what's neat about this, this string here is just to pick it up and lower it, and this is the input line for pneumatic pressure here. And when this thing wraps itself around, it's very hard to get a hard robot to do it because you have to have sensors that tell you that you're not breaking the crush strength of the shell. But here, because the pressure is distributed uniformly across all the contact area, all you have to do is to regulate one pressure, which is the input pressure, and you're done. And it's a first step in something which is really interesting about this, which is what it does is to use the properties of materials and design to replace complex sensors, feedback loops, and controllers. So it's a device which simplifies the technology. So how do we make it? We— a lot of this came out of our work in microfluidics. What we do is simply to design a mold, which in our case usually contains a number of small chambers little bubbles for a variety of reasons. We print it using a 3D printer. This might take a day for a complex mold. But then we mold a flexible silicone rubber against that. We put in some kind of layer that gives us the kinds of mechanical properties that we want. We glue these together with a bottom on it, and it's done. So it takes essentially no brains to do it. We like things with no brains. I'm going to come back to that, but I just want to give you a hint of something. First, as this gripper blows up, you can see the little chambers that represent the arm here. But the other thing I want you to look at, just something we'll return to, is it starts to curl up from the end even though the pressure is uniform throughout that system. And why does that happen? And there's actually very interesting science associated with that. But it's also easy to take these things and mold on the inside structures of that kind which improve adhesion if you want to do that. It's a very flexible technology. Now there are other things that make it interesting, and here's one. This is a related structure. It's the same kind of thing, but now in the shape of a hand. Comes down and grips the egg. And we've made this a little bit more complicated in terms of internal structure. You can see the fingers are almost articulated. They have a joint-like activity. And it puts it down without crushing it. And that's all fine. It gets its finger stuck there, but not to worry about. But now here's the thing that's more interesting. This differentiates it from hard robots. That's a mallet. Ouch. And you pound on it, and this would break a small robot of that sort, but just to give you a sense that this is real. And after you do that, it works just fine. And in fact, these are sufficiently light that you can throw them off an arbitrary building. It doesn't make any difference, and they're undamaged. They never reach a velocity that will cause impact damage on them. So they're essentially unbreakable to most blunt trauma. They can be cut without much trouble. Hard robots can't be cut, but they're very easily damaged, light ones, by blunt trauma. And then you can make a whole bunch of other things. I mean, here's something which is a millipede-like object. And there, here's an example for those of you who like problems, which is neat. We do this by pressurizing each one of these in turn. How would you make this all work with just one input here? And the answer is, we don't know right now. Here is another device which enables you to crawl up the inside of a tube, if you want to crawl up the inside of a tube. Perhaps you don't want to crawl up the inside of a tube. But this is doubtless useful for something. There are people already trying to think about using these to clean out sewers and pipelines and things like that. Now, these are all widgets which are pretty interesting. I want to talk about one bit of mechanics here. And it has to do with two components. One is what's called snap-through, which is a bistability. And you see that in the little things that kids have. You have a sort of a bubble shape. And you push on it, it pops through to the other side. Both are stable, but they're metastable in either phase. So, I want to briefly talk about snap-through and then also about beam buckling. And this is much more sophisticated and quite interesting, actually. It's got the mechanics people all excited. So, what's going on here? And this is an example of snap-through. What happens here is that as you pressurize this structure, the pieces at the end expand all the way before the ones in here do. So you have a system in which you pressurize this entire arm, but because of an instability in the system, the pressurization here leads to curling before the system here curls. And this is again an example of using design and materials to get around the problems that you would have with conventional controllers, because trying to achieve a motion of that complexity with electric motors and reels and things of this sort would be a major problem. Here it's just a trivial outcome of the basic design. And all of this, incidentally, you can vary these properties by varying the mechanical properties of this and the structure of the pneumatic chambers and things of that kind. Now here's an example of beam buckling. And this was apparently, you can see over here, this was a heavy-lift crane, which I would suspect somebody got a little bit more ambitious than they should and picked up something which caused it to buckle at that point. What is characteristic of this with the rigid material is once it's buckled, it's failed. You see this in concrete pillars. You take a concrete pillar and you push and you push and you push and nothing happens, and then it buckles, and when it's buckled, it's failed. With elastomers, that doesn't happen. So what we can do now is to build structures of this sort, and I want you to look at this as having two components, if you look carefully. What you see is not two circles, but rather two ellipses. There's an ellipse, an ellipse. This and this are the same. This and this are the same. And think about these as being soft beams. And what's going to happen is the following. I'm going to not expand these, I'm going to contract them by applying vacuum here. And as I apply vacuum, these chambers will shrink. This is an empty space with a thin membrane of PDMS on top of it. This will shrink. As it shrinks, this beam will buckle, but it will buckle reversibly. Amazingly, what will happen is that this internal piece here will rotate. Now, I don't know what's going to happen when I do that. Nothing is going to happen. So let me go back and find this on my computer. So what you're seeing here is a rotary motion here. So I've converted the linear motion that I showed in other things into a rotary motion. And you can begin to do all kinds of interesting things with this. I'll show you one here. What this is, is a single input, which is a single source of vacuum, leading to massively parallel motion, which in this particular case orients a bunch of people— not a bunch of people, a bunch of discs or placks here. This would be very hard to do again in a system that was as simple as this one if it were hard. So by this kind of beam buckling, it's possible to get to some very interesting sorts of structures. Here, I think, is another one. No, this one again requires this. Oh yeah, the reason for pausing on this is that what you see here, this part of it is going to rotate in that direction, this part is going to rotate in that direction because of the symmetry of this series of holes here with ellipses of different orientation. And when you do that, what happens with a single vacuum input is this. Collapses, grabs, and picks up. So, I mean, quite a sophisticated motion for just one pressure And the nice thing I should say about negative pressure is that it is prototypically cooperative because it can't blow up. You're just sucking in. So you can have this kind of thing embedded in you and there's really no risk of exceeding an overpressure and having it explode in your abdomen. Not a good idea. This just can't do it. So it's intrinsically safe. Now, here's muscle. What I'm going to show you here is our muscle equivalent. I'm really very happy with this kind of thing. You can see the structure. It's the same kind of structure that I showed you under negative pressure, that is to say vacuum. And as the chambers that you will see collapse, what happens is it contracts this way, but it does not contract that way. And it's designed for this kind of anisotropic motion. And here it goes, if I can find my cursor. Where is my cursor? Here we are. So I apply a vacuum, and it lifts a big weight. I mean, it does about the same thing that a muscle does at the same volume. And if you pause for just a second and look at the structures that are here— I should have that loop, but that's okay. If you look at the structures, What they are is a series of beams that are again buckling. The beam connection may be a little bit unclear, but this can be considered as an ensemble of buckling beams. Now, some examples. This is Josh Lessing's arm with a volleyball, and that's Joshua Elkington doing his thing with a volleyball. This is our arm. Which does its thing. It's a little slower, but that's only because on this particular piece we had a thin piece of tubing. And then to show that it's just a step from here to the appassionata, it can do a flip. And it only took a couple of tries to do that, but you don't, you know, technical details. But this is really quite interesting because if you look at the mechanical performance of this, it's not really very different than that. It's about a third the speed, but we can actually have the same speed that we have with jostling. It's about the same efficiency, maybe 50% less, and it will lift about the same weight. So it's essentially the functional equivalent of muscle. Now, not the biological equivalent, but the functional equivalent. Now, applications. And I'm a big believer that applications focus the mind. So I'm going to talk about 4. One is to show you some work that's been done by a company named Soft Robotics, which was founded on the basis of this. And I want to make a remark about this. Well, let me just show you the other examples first. I'm going to talk about some further biomimetic structures. I'm going to talk about a topic that we're very interested in called life that never was, by which I mean If you look around, there are certain kinds of motions that creatures use to move around with, but you could imagine a lot of others, and why aren't they used? And it's now easy enough to make models of things that we can examine the question of why weren't some of these alternatives used. And then the last one is some examples in biomedicine. And let me just pause to comment about this first one. Good friend and ex-student David Walt was one of the founders of a company called Illumina. And those of you who are biologists will have heard of Illumina. Illumina makes all of the boxes that are basically used, 80% of the gene sequencing is done with Illumina sequencers. And I was at a meeting once and some student raised a hand, her hand in that case, and said, Professor Walt, we know you continue to work on this and we know that you're also doing things with proteins and how do you avoid the conflict of interest of interest in your laboratory by having both a company and having a laboratory activity? He says, there's no problem. He's absolutely right. The reason is, he says, that as soon as you hand one of these problems to a company with mature professional engineers, you can never keep up in university. You can't possibly do as well as they do, so you don't bother. When you hand it over, you quit, you go do something else with your life. And that's our approach with this kind of thing. There isn't a conflict of interest. We continue to work in soft robotics. Soft Robotics works in soft robotics, but what they're doing we can't possibly do. We don't actually want to do it. It's not our business. So let me show you a little bit about what they do. Here is a soft robotics gripper picking up a pumpkin. Here's a soft robotics gripper picking up potato chips. Very hard to do with hard robots because they crush the potato chips or they don't have the strength to pick up the other thing. This is fruit picking, and the idea here is how do you get around the difficulties of, you know, being a farm worker. Here is picking up something which is a very fragile object. And then the everyday task which we all experience of picking up a cactus and putting the cactus down. I mean, I won't ask how many of you do that every every morning. And then here are some agility tests. And you notice what this is, is positioned on the end of a hard robot. And I think this will be the major use at the beginning. This is something which I actually quite like. And I'm told this only took 4 or 5 tries to do, catching the ball. But it's really not an easy thing to do with a hard robot. And here is You know, another thing that's actually quite hard to do with hard robots. It's really easy to do here. Notice 3 fingers. You just pop it off because it's a magnetic couple and put on 4. And that's critically important if you're actually thinking about industrial applications because you have end effectors that are changed out all the time for these arms. Then here is another example using one of your next-door neighbor's devices, and you can watch this thing move. It's really quite good. Here's celery. This is real, you know, real speed, real time. And here it goes. And what the gripper is doing is adapting to the different shapes of the celery almost effortlessly because it's a soft gripper. So you don't have to have controls that watch everything in detail. You combine preexisting vision systems and this this amazing tripodal arm in order to achieve this kind of motion. And why do you do this? And the answer is that although you don't think of yourself this way, you are filthy. And so if you are in an industrial food processing line, you may feel that you just have a minor cold, but actually it's the beginning of a neurovirus infection, at which point 100 people get sick, because you're just there, and the company that's involved loses 25% of its revenue for one quarter from that sector of the United States. So there's a very deep interest in the food processing industry to get people completely out of food processing, industrial food processing lines. And this solves one of the problems in that, which is to have something that has the characteristic that it's flexible enough to pick up food, which is variable mechanical strength, variable shape, variable size. It's just easy because it's pneumatic. OK. Now, more biomimetics. This is one of my favorite animals. This is a spider. And although the spider is a very sophisticated creature, it has joints that are actually among the simplest of the arachnid joints. And the way this works is here, there's an exoskeletal part which is rigid, and here's an exoskeletal part here that's rigid, and then there's effectively a little balloon in here. And what happens is that when the spider— the natural relaxed state of a spider leg is curled up, and all of you have seen dead spiders, and dead spiders always have their legs curled. And the reason is that what happens is that this tendon here naturally pulls it in this shape, and what the spider does is to pressurize the structure on the inside and expand it like that. So that this is the basic force diagram for spider joint, and what we do is the same thing of using a rigid axial structure. We put in a little tension element here, which is an elastomeric tendon, and then here we have basically a balloon, and you'll see in a moment that this works quite nicely. I would also say that the The basic sophisticated structural elements that we use here for these polymers are the same straws that you use for drinking soda. So, I mean, it's cheap, it's really wonderful, and it's a circular cylinder beam, so they're very satisfactory. So here is an ant. It happens to have spider legs. But notice the way the arm, the leg, it's, at any point there are 3 feet on the ground. Let's do it again. Let's see what happens if— no, I know that's not going to work. At any point, there are 3 feet on the ground, so it moves as a series of 2 triangles. And it's actually a very sophisticated motion which enables the thing to maintain stability over a wide variety of rough terrains. And we haven't spent that much time working on rough terrains, but here it is galloping uphill. That's an example of a biomimetic structure. Here's another biomimetic structure. This is a water strider. A water strider uses 4 of its legs to maintain balance and 2 of its legs to scull. And one of the characteristics of a water strider is it has to be light enough that it will stand on the surface of the water held up by surface tension, and this is. So it's, you know, does the world care about another water strider? And I don't care because I thought it was a nice demo. Okay, so then the next thing is life that never was. Here I think you will be amazed at how this all works. Here is one of the— it works, you know, it's galloping after prey here, but it's not galloping very well. So it's doing the best it can. I could imagine an organism that did this. I mean, here it is escaping from a predator. Here is— this is actually a pretty interesting device. It's just an incompressible disk, and here we put a balloon. And as the balloon expands, it causes the disk to buckle, and this nonlinear motion here, the buckling motion combined with legs, provides actually pretty good motion. And if you have two of these, I couldn't find the video, but if you have two of these, you can steer it quite well. So this is a perfectly practical way of doing things. Why this didn't occur during the Cambrian explosion, I don't know, but it doesn't seem to. And then if you look at this carefully, each one of these, as it expands, because there are rigid slits, it expands, the balloon expands, but it rotates. And because it's again the ant-like construction of 3 and 3, it manages to make its way forward. This must be among the least agile and beautiful walkers you've ever seen. But nonetheless, it works just fine. So you can explore these kinds of ideas. And then the final thing is actually important. This is probably as important as the food handling and mechanics. Incidentally, the application for that first thing that I showed you of picking up potato chips and pumpkins and things of this kind, why would you ever want to do that? And the answer is you want to do that if you are working for Amazon and people buy, you know, they go online, they buy potato chips and 5-kilo cans of lube oil and sweaters in slippery plastic bags. And so the idea for that is to replace people who do one of the world's most horrible jobs, which is packing boxes for Amazon. So here's just an example of a system which will probably go into development pretty soon because it works. What we have here is just a glove that— this is a glove that you buy from Sears and Roebuck, and we've put a simple, you know, bender on the outside here and connects to a pressure source in this case. Could be done with vacuum source if you want to do it a different way. What happened to my cursor? Here we are. So you can see you move fingers. Why do you care? And as I said, an example of the use of this kind of thing is as aids to nurses. Surgeons get all kinds of care. Nurses don't get so much, even though they really run lots of the medical system. And one of the things you have to do if you're a nurse is the following: You are in a rehab unit, a patient comes in, the patient has had a stroke, and the stroke affects a hand, and it's critically important that that hand be manipulated to begin to return it to motion, otherwise you develop, as a patient, something called spasticity, in which everything freezes in place and you sort of can't get it going again. Ever, no matter what you do. So it's really important to do that. But by the way, it's also not really very interesting to do it, you know, to sit there and help somebody move their hand. This is a fine solution for that kind of problem because you can arrange it so the pressure and the timing is anything that you want. It's more convenient for the patient. It's more convenient for the nurse. It's probably just a better deal all around. And you can also take a step with this and go one further. Which is that it turns out these are going to be, I think, very useful as assist devices. So if you've had a stroke, you may recover some motion but not all of the motion, and you would like to have something where you get a little bit more strength to pick up an apple or a cup or whatever it might be. And this will provide that very gently. It just gives you a little bit more strength to hold on to things. And all it requires is a little compressor on a belt and some batteries, and you can go ahead and do this. So I think there are going to be lots and lots of very useful uses of this kind of thing. So let me begin to summarize, first with some comments about soft robotics. The first thing is that we do not think of soft robotics as competitors for hard robots. I think of them as supplementary. And there will be areas where there is competition because there are— vast effort has gone into making hands with reels and electric motors and things of this sort. And there may be cases where hard hands are better, but I think in general, sort of soft hands are going to work better. They're much cheaper and they're more adaptable. The second thing is collaboration is really important. If you talk to the people who make hard robots, they will freely admit that the big problems with hard robots is that they're non-collaborative and they don't have good mechanisms for making them collaborative. They're also pretty expensive for what you get. They're also very energy inefficient. The hard systems, a pony-sized object that is a hard robot run by a motor, will typically be about a factor of 100 less efficient than a pony will be. And why the biological systems are so much more efficient is not completely clear, at least not completely clear to me. It's an interesting basic problem in motivation, in motives, motion. So anyway, collaboration, being absolutely safe around people, is critically important. Cost is a big advantage of these soft systems. Lightweight means that they can operate in unstable environment. They're damage resistant. They're complementary kinds of things. I mean, they're different systems in different ways. A very interesting issue is that if you look at hard robotic systems, a lot of their cost is the control systems. And there will be applications with soft systems in which you also have to have controllers. But as I showed you, you can get conformal gripping of an irregular-shaped thing, an apple or a whatever it might be, with no controllers at all. You just have one pressure controller and the materials do all the rest. And I think that's very interesting. These issues of nonlinearities, I told you about snap-through and about beam buckling, but, and then also I showed you the Maha disc system, which is another buckling instability. But there are a very large number of interesting nonlinear motions that you get in buckling in materials if you combine elastomeric and rigid materials. So that looks like it's going to be very engaging. And then I really believe that applications and products focus the mind. So surgery and rehab and prostheses, we've talked about. Agriculture, both food picking and food handling. I haven't talked about search and rescue, but I'll touch on that in just a moment. Single use, you send a robot into Fukushima, to have a look around and see what's going on. By the time it's been there, it's so radioactive that you don't want to get it back. But you can't afford to send a stream of $50,000 robots in and throw them away. These, it makes no difference. You couldn't care less. You can make them for, you know, $50. Hazardous tasks, you know, all of these kinds of things, I think, are legitimate applications for the soft robotics. Now, a couple of other things. One question is always, what next? Everything that I've showed you up to this point has involved rubber with the same general sort of elastic properties as rubber bands. We don't actually use rubber bands because they auto-oxidize. But PDMS is, you know, it's an elastomer. You can get pretty good elastomeric strains on it. This is a kind of elastomer I'm very interested in. That is automobile or bicycle tires. Because the pressure that we use for most of the grippers that I showed you was about half an atmosphere, so 7 pounds per square inch. And you say that doesn't sound like very much, but I would point out that that's the difference in pressure between the top and the bottom of a 747 wing when the thing is flying. It takes about half an atmosphere.. And if you don't believe that, estimate the number of square inches in the wings and the difference in pressure, and you end up with about in the order of 1 million pounds that that can pick up. So with this kind of structure, it becomes less elastomeric, but I can do 100 to 1,000 pounds per square inch with those. And that means that one of our longer-term objectives is to make a robot that has the characteristic that it can tear an automobile in half. And my students have not embraced this project yet because they say they don't have a spare automobile, but that's a lack of imagination on their part. And then for a number of things in biological applications, there are truly soft materials like— that's Jell-O or some tomato aspect. And there's no reason why I can't make low-pressure, low-force things for that. But that might be just the thing to hold a liver out of the way or something of that kind. So I'm— I think there are both opportunities here, obviously a lot of these, but then new kinds of structures both here and new kinds of applications here. Now just two final things which have to do with this. This is again a kind of how does this fit into a university kind of issue. And for those of you who haven't seen this, this is a diagram that was put together by a guy named Don Stokes at Princeton and has become the basis for endless debate in the United States about policy of science. And the argument is the following. In the period since World War II, we have had a model in which there was a strict division between research that was done Basically for curiosity, and on a scale with low and high quest for fundamental understanding, we have Mr. Bohr, who was not interested in applications when he thought about atoms, and so he is used to— his name is used to personalize this quadrant of high fundamental understanding. And then Mr. Edison is down here for high practical use, and Mr. Edison incidentally despised the whole idea of research. He thought that if you wanted to make a better light bulb, you went to your storeroom and you got fragments of bamboo and bits of ox hoof and whatever you could find, and you just saw what worked. There was no point in screwing around with the science stuff, because it really didn't work very well. And so this was supposed to be industry and this was supposed to be university. And, you know, we know the distinction was not that clean. But many people, and I fall very clearly in this category, believe that this is where we really want to be. And this is called in the trade the Pasteur quadrant. And who was Mr. Pasteur? Mr. Pasteur did two things that were interesting. The first was that he was around at the time that the very first microscopes were being developed, and he looked at milk. And fresh milk looked like nothing, and spoiled milk had all sorts of little things moving around in it. And he said, there must be some correlation between these little animals moving around and spoilage of milk, so I'll try heating it and it'll spoil less rapidly. And that became pasteurization. But by the way, he's often called the father of microbiology because that was the first He solved— there was a problem which was food spoilage. There was no science to deal with that, so he invented the science to deal with a big problem in application. He also was one of the very first people to do vaccination, and so he is often also in a different world known as the father of applied immunology because he was— I think he did rabies, although I've forgotten. So the question is, what about this as a way of doing things? I argue that this is where universities should be. We should look at big problems, and to me robots fall as a big problem because of the arguments that I made at the beginning. And there isn't very much science for the soft part of big robots, so shoot, we should find ways of solving the problems with robots to deal with this big problem to enable human beings to do things that they don't want to do. Like, for example, substituting something more interesting than standing in a field being crop-dusted with insecticide while you're picking fruit in the Central Valley in the summertime at 140-degree temperature. So if we can find, and it's in a sense your all's responsibility, find jobs that are more interesting so that when you don't have to pick fruit, you can go do something that's really, you know, more fun to do and more profitable. So that's an argument for having at this end of things a component that's a company, at this end of things a component that's smart students, and trying to put them together. You note that there is a blank quadrant here. Low quest for fundamental understanding, low consideration of use. In Washington, where I spend too much of my time, this is called the university quadrant. And that's not where we wanna be. So let's try to think about that problem. And I put this up because I believe, as a, you know, I'm a professor and I believe in reading lists. And so this is a reading list. And let me just go through these 4 things. My colleague Clay Christensen made a brand, in a sense, by talking about what he called the innovator's dilemma, which was the inability of big companies to change their products and business model when technology changed. So, you know, what did Olivetti, which made mechanical typewriters, do when computers and things like that came along? And the answer is they went out of business. So the question is, how do big companies do this kind of innovation? It's a really good and interesting question. He was answered by my other colleague, Jill Lepore, who is a historian, I think. I'm not exactly sure which, but someone who, as one of my colleagues has said, writes more rapidly than he can read. So Jill just pours out really interesting stuff, and she at one point worked with Clay, and she said not a word of this is true. It has nothing to do with innovation. What happens is that innovation is really a gradual accretion of improved ideas. You have a product, you improve it, you have that product, you improve it, and that what we have now in the internet and what you have in automobiles is not the result of a single dramatic thing, it's the result of constant improvement. I happen to fall more in this category than in this category, but it's interesting to read these. This has to do with the question of whether industry should deal exclusively with short-term work that has the characteristic that it basically can't involve anything in terms of longer-term research, because the argument goes, it is the obligation, the fiduciary obligation of boards of directors to maximize stockholder return. And what this woman, Lynn Stout, has to say is that it is not the fiduciary obligation of boards of directors to do that. Rather, It is the obligation of boards of directors to maximize the welfare of the corporation. And that, in fact, might mean making sure it's around in 20 years, which would be a long-term investment. This is not popular with Goldman Sachs, but, you know, this is something you have to think about. And then this is a, for all of you who are students, I think an absolutely wonderful book, which was put together as a history of Bell Labs, which everyone everyone believes was this giant playground in which brilliant physicists, unconstrained by economic reality, did brilliant physics and created the transistor and the internet and the entire world. I mean, and there's no question that Schreifer and Bardeen and Shannon and the rest of these were brilliant physicists. I mean, they did fantastic work. However, the book makes it very clear the reason they were doing this was to make a more profitable telephone system. And there were people who actually wanted to be in New York and talk to customers in San Francisco. And they were willing to pay a lot of money to do that. And as a result of that, there was a demand, a market for the production of ideas leading in that direction. And so I think it was Baker who was the head of this operation at that point. Divided the process of innovation into 4 steps. There was a step which we do, which is science, and then there was a step which she called invention, which meant making, proposing a kind of prototype product. And then there was a much longer and harder step, which was development, which was the engineering development and regulatory clearance and product safety and shelf life and all the rest of this kind of stuff. And pre-manufacturing, manufacturing setup, so at the end of that you had a product. He said if you get those 3 things done, you've accomplished nothing, because the only thing that counts is, is there a market? And in universities, we don't know very much about markets. So those of you who are interested in being entrepreneurs, it's not taking your thesis and converting it into a prototype product. It's figuring out that somewhere out there, there's a payer who isn't necessarily the customer or the beneficiary. It might be the insurer or somebody else who will actually pay you, do this most intimate of human interactions, which is to write you a check in return for what you have. And it's a different point of view, and it's very worth the effort learning how to do that. And then my final slide is this one. And I put it up for one reason. You don't know any of these people, and they're terrific people in various ways. What's interesting to me is this red list here. Because there was a period not so long ago, 20 years ago, 30 years ago, in which if you'd looked at that list, it would have been 80% US. And bits and snatches of other things. And the world has truly changed. The best people come from everywhere. Not from everywhere, but there are places that are just particularly good. For example, the people I get from Turkey and from Iran, somewhere, I don't know, maybe there's an Iranian, yeah, here, Iranian, are spectacularly good. I mean, really spectacular. And I don't know why that is, I can speculate. But you notice there's 1, 2, 3 US people on here, and all the rest are other things. I think this is an important part of understanding how we run a university in the future as well. I don't know exactly what lesson to take from it, other than to do interesting stuff, you need the best people you can find from anywhere. And you gotta figure out where they are and how to get them and how to bring them up to speed and all the rest of that and what happens later. But I suppose it goes without saying that these people did everything that I told you about, so I'm here simply for advertising. And I thank them, and I thank you. That's it.