IRA FLATOW, host:
This is TALK OF THE NATION/SCIENCE FRIDAY. I am Ira Flatow.
You know, reproduction has always been the exclusive talent of organisms containing genetic material--but not anymore. A team of researchers from Cornell University has created a machine that can make an exact replica of itself. It's sort of a feat that brings self-replication repair from biology into the robotic realm. You science fiction enthusiasts who have waited for the day when machines would walk amongst us--well, looking like people? Well, you're going to have to wait a little bit longer because the Cornell team's robot looks more like a stack of four toy building blocks, the type you might have played with in kindergarten more than it does looking like the Borg or something. Here to talk about the project is the head of the research team. Hod Lipson is assistant professor of Mechanical & Aerospace Engineering and Computing & Information Science at Cornell University.
Welcome to the program, Dr. Lipson.
Dr. HOD LIPSON (Cornell University): It's a pleasure to be here.
FLATOW: We're going to talk a bit about what these cubes--the machine looks like. Let me send everybody to sciencefriday.com, to our Web site. We have a link there to your robotic machines; it'll be easy for people to go there and look at it while we're talking about it. Give us an idea, for those of us who haven't seen the pictures, what the robot looks like.
Dr. LIPSON: Well, the robot is essentially a demonstration of principle. It's made out of four cubes; it's very simple, as you described. Each cube is composed of two halves than an swivel, one with respect to the other. So imagine a set of cubes connected to each other. When one of them swivels, then the cubes connected to it will change conformation. If you have a tower, for example, and some of the cubes swivel, it will change from a tower into a L-shaped machine and then it can change into an S-shape and so forth, so it can move around, change its configuration. And one of the things it can do is it can bend over, pick some cubes from a feeding source, so to speak, and put them at a new location gradually reconstructing a copy of itself.
FLATOW: So it builds a copy of itself right next to it.
Dr. LIPSON: That's right.
FLATOW: And you give it new cube sources to do that. What are inside the cubes that allow it to do this?
Dr. LIPSON: The cubes are pretty simple. They are--contain a motor which allows these two halves to swivel, and it contains a microprocessor that executes commands, and each of these microprocessors exist in each of the cubes and they can--they have a blueprint, if you like, of what the robot should look like and what each cube should do in order to achieve that new shape.
FLATOW: 1 (800) 989-8255. We're talking about these robotic cubes at Cornell with Hod Lipson. How is this different--we mentioned before that usually an organism has to have some sort of DNA/RNA, some sort of genetic material to reproduce itself. Do you give it--is that what the computer program is basically, the genetic material inside of it?
Dr. LIPSON: Oh, exactly. The computer program or the blueprint, if you like, is in some way equivalent to the DNA in a cell. It contains information about what the shape should look like, and more importantly how to achieve that, how to change conformation and what actuators to move in order to reconstruct a piece of the robot.
FLATOW: So it looks around its immediate area for a source for some new material to build a reproduction of itself, and then it knows what to do once it sees the stuff, the raw material sitting there?
Dr. LIPSON: It's even a little dumber than that. It really expects that the cubes will be fed to it in a very particular location, and it goes there to fetch the cube and it waits to sense for that cube, and when it gets that sensory feedback, it picks it up and puts it in the right place. And if the cube is not there, it really can't reproduce; it will fail. It really can't look for anything. It's all reprogrammed and it's fairly rigid in its ability. It's interesting to note that the robot that is being built also needs to help in its own construction, because the parent robot cannot reach across the child robot, so to speak, in order to put down the cubes.
FLATOW: Oh, so once it gets going, it helps build itself.
Dr. LIPSON: That's right.
FLATOW: And the child robot then has to be able to make its own copy of itself, so, etc., etc., so to speak.
Dr. LIPSON: Yes. When the child robot is complete, it starts executing this construction program--look for cubes again and makes another copy, and so forth.
FLATOW: Why do we want to build robots that do this?
Dr. LIPSON: Well, I think the bigger picture is the point to remember. These robots that we made are instances, examples of what we call large space of possible machines. There are many different kinds. We've shown larger machines composed of eight, and more interesting structures that are not towers, but they can fork--forking structures in 2-D and in 3-D. There's really lots of these different machines, so this is just an example. So the main--I think, the big picture/point here is that if you--that most people imagine robots as these metal machines, and if you think about robustness of a robot, you would expect that you just make the robot tougher, you make it more durable. But an alternative that we're talking about here is to make robots that can self-repair, and that's really where we're going with this. You can imagine a robot, let's say, on a space mission, it arrives at the planet, it breaks down. If there is another robot there that can fix it or a robot can create another robot more suitable for a specific task, and then dismantle the original robot and reuse its parts, that can be an alternative way to make machines more robust. So instead of these tough robots, we look at self-repair.
FLATOW: So you, like, bring all the modular pieces with you. If you want to dig something you have a modular piece with a shovel. If you want to do something else, say, you have some other modular piece and the robot can make whatever it needs to be at that point.
Dr. LIPSON: Exactly, and the question is what's the smallest set of building blocks you would need to really...
Dr. LIPSON: ...have a very large repertoire of possible functionalities.
FLATOW: And do we know what that is yet?
Dr. LIPSON: No, but that's one of the things that defines, in our opinion, the ability of something to self-reproduce: how many modules it has, how versatile is the function that it can achieve using them, how simple the modules are--there's all different aspects that it will improve the self-replicability of these types of machines.
FLATOW: Robert in Burlington, Vermont. Hi. Welcome to SCIENCE FRIDAY.
ROBERT (Caller): I have a really weird question, but are you remanufacturing these microprocessor chips that you're creating the new robots with, so that they have their own microprocessor that they can in effect create more robots? I can't see how you can possibly be doing that.
Dr. LIPSON: No, the robots take a supply of cubes. They are fed with cubes. The cubes are the basic building blocks...
ROBERT: And the microprocessors are part of those cubes.
Dr. LIPSON: That's right. The microprocessor, the motor--all of that is part of the cube.
ROBERT: I don't think that can be compared, then, to the reproduction that happens in biology because, you know, in biology we don't have--we don't eat little brains and stuff like that to reproduce more people. We eat much more, you know, fundamental food, much more basic food than that. It's like--to be comparable to the biological reproduction, it'd be like your robots would have to eat silicon and eat plastic and eat metals and stuff like that, and somehow assemble them into microprocessors in order to do that. That's what would make it comparable to biological reproduction.
FLATOW: Well, what about if you cut a worm in half and it grows the rest of itself back again?
ROBERT: Well, if it grows another brain, yeah. Yeah. You know, if it creates an entire thing from fundamental food. But see, the thing is if he has food that is really like, you know, part--it's like if I were to eat another human being but somehow integrate that other human being's brain and then create another human being out of that. That's what the comparison would be. It is not nearly comparable to the reproduction that happens in biology, if you have food that's full of microprocessors.
FLATOW: Dr. Lipson?
Dr. LIPSON: Yeah, I think this is very good point, and one of the things that we put forward is really this notion that self-reproduction is not a binary thing that either you can do or you cannot do, and all forms of self-reproduction are equivalent. Absolutely not. And what we do say is different, that there are different levels of self-reproduction. They are all based on the same principle of taking basic building blocks and assembling them. The level at which this is done is different and gives rise to different extents of self-reproduction, so certainly people self-reproduce and animals self-reproduce by taking in very simple building blocks, amino acids, and making bodies out of that. Our robots take in much more complex building blocks and use fewer of them; nonetheless, I think it is a--you can compare these processes, but obviously our machines are very, very low in terms of their self-replicability compared to biological systems. There's no doubt about that.
FLATOW: Thanks for the call, Robert. So where do you go from here?
Dr. LIPSON: So, well, there's a number of possible directions, but one that is very interesting--a number of years ago we worked on evolutionary robotics. I think it was also described here on this program. And in that projects we used ideas inspired from biological evolution in order to adapt machines over time, so they evolved in simulation and they gradually improved under kind of artificial selection. And those machines could not self-reproduce, but they could adapt very nicely. The machines we have today can self-reproduce, but they cannot adapt. The one direction that would be very interesting is to combine these two traits into a single system and see--and get a little bit closer to what biology can do. And again, it's just in that direction. We're clearly far away from capabilities of biology.
FLATOW: When you say adapt, it would look at its environment to change and suit it?
Dr. LIPSON: Exactly. You can imagine a robot that you put on the floor and you scatter the cubes around, or some other building blocks, and it looks for them and it finds them and uses them, so it will be much less structured, and we'll gradually learn how to do that. It will learn how to self-reproduce; instead of being programmed like we did today, it will actually learn that...
FLATOW: And you need to...
Dr. LIPSON: ...over time.
FLATOW: ...get together with the "I, Robot" people to vacuum those cubes up...
Dr. LIPSON: Right.
FLATOW: ...so you can...
Dr. LIPSON: Yeah. That would be more approach. Yes.
FLATOW: Well, thank you very much for taking time to talk with us. Is there a lot of money going into this, or are we basically doing this kind of research on your own?
Dr. LIPSON: Well, there's a little--a little going into this. I think NASA, for one, is one federal agency that's interested in the possibility of these, and I think it's a viable alternative.
FLATOW: All right. Well, thanks again, and good luck to you.
Dr. LIPSON: Thank you.
FLATOW: Hod Lipson is assistant professor of Mechanical & Aerospace Engineering and Computing & Information Science at Cornell University.
We're going to take a break, and when we come back we're going to talk about sexual abuse. There's legislation, there are laws in states around the country--in Florida, other places--that have really made stringent laws against sexual abuse. We're going to talk about them. And the science of sexual abuse--very little science in defining how we deal with it. So stick around. We'll be right back after this short break.
NPR transcripts are created on a rush deadline by Verb8tm, Inc., an NPR contractor, and produced using a proprietary transcription process developed with NPR. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.