 Next up I'd like to invite Peter Anderson from the ANU College of Engineering and Computer Science and the title of his three-minute thesis tonight is The Language of Sight. Imagine that you were walking down this street all around you people are shopping, talking, drinking coffee but what would it be like to walk down this street if you couldn't see? Could you still find your way? Could you find an ATM or a table that was free at a cafe? Worldwide there's 285 million people with impaired vision that face challenges like this on a daily basis and despite all the advances in medical science many cannot be cured. So in my research I'm using artificial intelligence or AI to help. I'm building intelligent visual assistants that can help in ways the guide dogs and white canes cannot by describing what they see. Looking at this image my visual assistant describes it as a group of people walking down a street with several shops. Now that's helpful but of course there's much more to a scene than this. That's why you can also ask questions. Questions like is there an ATM here? Can you see a street sign or where's the coffee shop? And each question should receive in response a helpful correct answer. Now you and I can describe our surroundings and answer visual questions like these with these but this is still very difficult for an AI system but we're making rapid progress. To tackle this problem I'm using a deep artificial neural network a computer model made up of millions of connections inspired by the vast network of neurons in our brain. In my particular approach I train the network to first identify every single visible object then inspired by our own visual system the network learns to focus its attention. That means it looks only at those objects that are most relevant to either describing the overall scene or answering your specific question. Using this technique right now our work leads the world in terms of describing images and visual question answering. So it's my hope that in a future filled with self-driving cars and autonomous robots that we can also offer greater autonomy to millions of people with impaired vision.