Want to keep learning?

This content is taken from the University of Birmingham's online course, Good Brain, Bad Brain: Basics. Join the course to learn more.

Skip to 0 minutes and 6 seconds Hi, my name is Nick Hawes from the School of Computer Science and this is one of my robots, Bob.

Skip to 0 minutes and 16 seconds So I work on artificial intelligence which is the field of creating intelligent machines, or intelligent software, or any kind of intelligent artefact. Bob is just a machine. Bob is mechanics, and electronics, and all sorts of stuff. But the role of artificial intelligence is to try and put a brain in his head, or actually in his stomach where the computers are. Well, in artificial intelligence, we’re inspired by the brain in many ways. And personally, I’m inspired by it as a, sort of, proof of concept.

Skip to 0 minutes and 45 seconds We know that there are these artefacts, these creations, this bit of physical stuff that creates all sorts of intelligent behaviour, whether that’s actually in humans or in animals, we know that it’s capable of a whole range of things. So we’re always pushing to try and build robots and programme robots that can operate in human environments and operate in the ways that humans do, and particularly can learn and adapt in the way that humans do. And that’s the crucial thing. So we’re actually now very good at programming robots to do very simple tasks like assemble cars in a very, very controlled environment, or even mow the lawn or hoover your carpet.

Skip to 1 minute and 20 seconds Those things are very, very easy to do, because the world is very predictable and simple. What we can’t do is build or programme robots that can adapt themselves or learn the way that humans do. So what’s really incredible about the abilities our brains give us is how plastic and adaptable they are. How we learn over time and develop over time. Robots don’t do that. Or AI systems don’t do that. They’re a fixed rigid software system. Now, you could imagine that in that, some of the kind of weights inside the programme, kind of like the connections between neurons, if you will, can change and adapt over time. So some things get more prominent.

Skip to 1 minute and 55 seconds And that allows robots to adjust their behaviour in small ways. But these large functional changes, we don’t have the ability, or the software, or the knowledge how to programme these things.

Skip to 2 minutes and 11 seconds Some people do programme them as if you’re programming brains. So people create artificial neuron networks with synapses and neurons. And they create these things and train them. I don’t really follow that approach, because I think that gives you a big system and you don’t really know what it does. You’ve just go all these weights, and flashing lights, and things. So I very much like to recreate a functional level so you can know what’s going on. But there you really have this problem of isolated functions. You have these blocks that do one thing and not another thing. And how do you make the bit of the robot that sees, so we’ve got some cameras, they’re the eyes of the robot.

Skip to 2 minutes and 44 seconds How do you hook them into the speech if the robot wants to talk? How do you get them to hook them into memory? All these things happen so naturally in our brain, at least from my perspective as not an expert in the brain. But in a robot or in software, these things are very separate. We have this huge problem about crossing modalities which, again, seems to be very natural in a natural system. So we’re inspired. And we face these challenges. So I’d say the big challenges that we face in robotics that seem to be solved by nature are integrating and learning over time. How do we take all our sensory processing and make a 3D world out of them?

Skip to 3 minutes and 21 seconds And how do we take that and learn from our experiences? How do we learn to get better at things? How do we learn to do new things? You can teach a human just by writing down a series of instructions and give them a series of things they can do in a new environment. A robot, you’d spend a year programming it to do that same task. And then it would only be able to do that one thing. Whereas, a human will learn and adapt over time.

Perspective: robotics and the brain

Dr Nick Hawes discusses the value of applying what we know about the brain in the areas of Robotics and Artificial Intelligence (AI).

Share this video:

This video is from the free online course:

Good Brain, Bad Brain: Basics

University of Birmingham