Close
The Robotics and Artificial Intelligence MEng is an innovative new degree launching in 2023. Professor Simon Julier (Robotics and Computation) who co-wrote the course, shares some insights.
How is the field of robotics changing at UCL?
UCL has been working in robotics for a number of years now, but it’s all been scattered across different departments.
Quite a lot of robotics has been focused in mechanical and electrical engineering, with a lot of use in surgical applications and manufacturing fabrication too.
It felt like the right time to consolidate everything together, and with the space available at UCL East, it gives us the ability to run a large number of robots.
What difference can artificial intelligence (AI) make to robotics?
AI and machine learning is absolutely dominating these days. When deep learning was in its infancy, we were trying to tackle things that people said were impossible. And yet here we are, solving huge problems with machine learning.
AI has come on so far, it’s at a point where it can be applied to robotics in many ways.
Companies like Tesla and Arrival are all investing heavily in robotics; they are also used for space exploration and even increasingly in film and TV production.
There’s a fundamental problem of having data from the real world, but a lack of ability to interpret it to create sensible actions for the robot to take. These are all machine learning problems.
There are also aspects such as processing, natural speech and sentiment analysis where AI can really help.
What are the challenges that still need to be solved with robotics and machine learning?
People need to think about the limitations of these algorithms in terms of the actions that the robots can actually do.
One frustration we've seen is with computer vision. This is when you show pictures to a robot and it identifies what the pictures are – for example, a cat, a duck, a skateboard, a duck on a skateboard and so on.
They might be accurate about 80% of the time. Even when you hear of a new network that gets better results than this, it is often only marginally more than 80%.
This kind of news gets a lot of public attention, but as roboticists, we’re thinking that if you put anything in front of the robot, it will get it wrong a fifth of the time.
This kind of success rate is not good enough to apply to many real life practices, whether it’s in the medical world or self-driving cars.
What are the most common misconceptions about robots?
If you've only seen robots in films or TV, you think robots are really great. There’s an idea they’re like R2D2, or magic hyper intelligent devices.
But when you see them for real, you realise they have absolutely no common sense!
They'll do bizarre things that makes absolutely no sense to anybody. In fact, they're systems that need engineering, care and feeding. They only can put out what you put in.
There can be a similar misconception with AI and machine learning – that you don’t really need to know anything as a human because the machine will solve it all. But you need to know the key critical skills to help machines learn, and to look at data and know if it’s sensible or not.
So what will students learn on this degree?
The true strength of this programme is that it couples theory with practice.
There’s a strong foundation in key critical skills, which has a lot of maths as a basis. We’ll look at a lot of these different skills and see how they work with machine learning.
There's an implementation component, because most of the modules will ask students to write code, understand the complexities of programming, and understand how programmes all fit together.
As well as teaching programming skills, we also teach debugging skills for when things stop working.
There's some learning on how to design and run experiments, because when you create any kind of robotic system, you obviously want to know if it works or not.
Communication skills are key as well – and assessments also include report writing and presentations. There’s a lot of practical work and project work, and students have access to a huge number of robots in the lab.
Throughout the degree, there is a constant focus on ethics too – thinking about what should and shouldn’t be created, and considering what tangible differences new ideas in this field can make to the real world.
Why should students apply for the degree?
There has traditionally been a lack of people in robotics who are really strong machine learning people too. But probably half of the robotics papers published now have machine learning in some shape or form in them, so this combination is clearly dominating the direction it’s going in.
Students I’ve taught robotics to in recent years have gone on to work as robotics engineers within various established companies and start-ups.
In other industries, the substance of what the computer does is extremely isolated from the real world. But bringing robotics and AI together is the total opposite.
Our belief at UCL Computer Science is that computers are not isolated from the real world.
Their whole purpose is to interact with, shape and change the real world, hopefully for the better. This is the idea behind this programme. We’re looking for students who are both interested in maths, and in making it come alive.
Having access to real robots in the lab and actually seeing and shaping what they can do is a very powerful start to see what can be translated out into the physical world.