Make robots smarter - Ayanna Howard

Transcriber: Andrea McDonough
Reviewer: Bedirhan Cinar

So how many of you have a robot at home? OK, I see about 20, 30 hands. That's actually pretty good. How many of you would want your own personal robot at home? I know I would! OK, so why doesn't this exist? Why can't I go to the convenience store or the department store and, you know, go up to the cashier and say, "Yeah, I want my personal robot"? Well, I'm going to talk to you about how to make that happen.

The thing that we need to do is to make robots smarter. Now, no one will argue that we don't have robots. We have rovers that are going to Mars and are getting science data and expanding our understanding of the world. We have manufacturing robots that are helping to build our cars that we drive today. We even have robots that are helping our military, that are out disposing of bombs so our soldiers can come home safely. So we have all this, so why don't we have the personal robot? Why don't I have my robot chef? Because I can't cook. (Laughter)

So, here's one of my robots, this is a simple walking robot, but it is by no means smart. And so, what we need to do is we need to change the definition of what a robot is. How do we do that? Well, the first step, before we even start designing and getting our hands dirty, we have to come up with rules, kind of the laws, rules of conduct. And why is this? Because if these robots are smart, they might be capable of more than we want. And so we have to come up with rules.

Thou, robot, shall not harm a human. Thou shall obey me, and only me. Thou shall always protect me at all possible times. So we have to lay the boundaries, the rules of engagement, before we actually start designing. And then we have to come up with tools.

So I believe that the way to make robots smarter is to mimic people. Now, our brains are complex, there's a lot going on in there, and so, it'd be hard to try to open up the brain and actually figure out how to mimic humans. The best way is to observe, is to actually watch people do things, and figure out what are they doing, what are their thoughts, what are their actions, what are their emotions? And so, part of making robots smarter is actually trying to mimic humans, mimic how we do things, so maybe they can do it a little bit better.

And so, some of the tools are varied. And so, I'm classically trained as an electrical engineer. I never thought I'd have to understand things like child psychology, infant development. So, understanding that the way infants develop to children, develop to adults, and how they learn and interact is actually important for robotics. I didn't understand that I'd actually have to watch tapes of monkeys interacting and communicating, because they have a whole social kind of mechanism where they learn from each other, and so that's really good to make robots smarter.

And, of course, neuroscience, I've always been fascinated with neuroscience, but I never understood that I had to figure out why do the neurons fire, what about the environment helps us to learn, and all of those really contribute to making robots a little bit smarter.

And so, some of the things that I do -- and this is just a little snapshot -- one of the things is mirroring. So they say our ability to look in a mirror and wave and actually recognize that the person on the other side is us, that self-awareness, is a sign of intelligence, and that allows us to then look at someone pitch a ball and figure out, "OK, I know how to pitch a ball, I'm going to mirror their improvement."

And so I actually have a robot where we are trying to design a robot health coach. And so, I have an exercise physiologist showing the robot how to do some exercises. You know, we want to get strong. And then, the other thing is learning. So, learning is important. We do this as children, we do this even as adults, we do this as elder. And yet, one form of learning is muscle memory.

So how many of you play an instrument? OK, so when you start off, for example, if you think about the violin, you start off and your instructor might actually come and move your hand a little bit or maybe move your bow a little bit up. So they actually touch you in order to give you muscle memory. And that helps you understand how to do things a little better.

And so we actually have a learning methodology where of course, we're not going to take the motors and move the legs, and so we have to Nunchuk to give our robot muscle memory in terms of how to do dance moves. And then, lastly, is creativity.

So, you might ask, "Robots? Creativity? I don't get this. Why does the robot have to be creative? What about creativity makes them smarter?" Well, creativity and imagination, those are the things that allow us to create problems when we don't know how to attack it. They allow us to make something out of nothing.

I mean, if you look at the apps that are out there and the tablets, and the iPads, and the iPhones, and the Androids -- 20 years ago they didn't exist. So, how is it that we got from something where there was nothing and expanded? It was our imagination. It was our creativity. And these are the things that allow us to figure out new things.

And so, I have a robot that is creative, it plays piano, is a composer, and if you listen, it plays "Twinkle, Twinkle Little Star." (Music)

So, all of this together, the last thing is interaction. So, you have a robot, you want it to be your playmate, your teacher, your instructor, you want it to interact. And isn't it so cute? (Laughter)

So, interaction is key, it is key to understanding how to work in our world with us, and so the interaction piece is very important. It deals with communication, it deals with understanding, it deals with gaze, it deals with attention. All of these things together allow that interaction and our robots to be smart.

And so these are just some of the tools that we use in order to make robots smarter. So, I want to leave you with one thought. So, I'm all for robots and smart robots. I mean, that's what I do, I'd be out of a job if I didn't believe in that. But yet, where does it end? How far do we push it? How far and how smart should we make our smart robots? Thank you. (Applause)