Can we trust the judgement of machines that see? Computer vision is being entrusted with ever more critical tasks: from access control by face recognition, to diagnosis of disease from medical scans and hand-eye coordination for surgical and nuclear decommissioning robots, and now to taking control of motor vehicles.
The latest technologies for visual decision-making use neural networks, transmitting signals from input to output in many stages, where the signals at intermediate stages are not easily interpreted. This makes it harder to understand and therefore trust the decisions. Moreover, it has been shown that decisions from neural networks can often be reversed surprisingly easily – by so-called “adversarial” counterexamples, suggesting a certain fragility in a network’s operation. How sure can we be that the computer makes good visual judgements and decisions?
Andrew Blake is a pioneer in the development of the theory and algorithms that make it possible for computers to behave as seeing machines. He is especially interested in segmentation as optimisation, in visual tracking as probabilistic inference, and in real-time, 3D vision. He trained in mathematics and electrical engineering in Cambridge UK and at MIT , and studied for a doctorate in Artificial Intelligence at the University of Edinburgh. He was an academic for 18 years, in Edinburgh and Oxford, ultimately as Professor of Information Engineering at Oxford University. He joined Microsoft in 1999 to found the Computer Vision group in Cambridge, before becoming Director of Microsoft’s Cambridge Laboratory in 2010 and a Microsoft Distinguished Scientist. Currently he is a consultant in Artificial Intelligence. In particular he is Chairman of Samsung’s AI Research Centre SAIC in Cambridge. He is consultant and Scientific Adviser to the FiveAI autonomous driving company, and serves as an adviser to Siemens. In 2010, he was elected to the council of the Royal Society and was appointed to the board of the EPSRC in 2012. He was Director at The Alan Turing Institute 2015-18. He has been Honorary Professor of Machine Intelligence at the University of Cambridge since 2007 and is a Fellow of Clare Hall. He has been a Fellow of the Royal Academy of Engineering since 1998 and Fellow of the Royal Society since 2005. He twice won the prize of the European Conference on Computer Vision, with R. Cipolla in 1992 and with M. Isard in 1996, and was awarded the IEEE David Marr Prize (jointly with K. Toyama) in 2001. The Royal Academy of Engineering awarded him their Silver Medal in 2006, and in 2007 he received the Institution of Engineering and Technology Mountbatten Medal (previously awarded to computer pioneers Maurice Wilkes and Tim Berners-Lee, amongst others.) He was named a Distinguished Researcher in Computer Vision by the IEEE in 2009. In 2011, with colleagues at Microsoft Research, he received the Royal Academy of Engineering MacRobert Gold Medal for the machine learning at the heart of the Microsoft Kinect 3D camera. Exactly 80 years after Einstein, in 2014, he gave the Gibbs lecture at the Joint Mathematics Meetings – the 6th British scientist to do so in 90 years. The BCS awarded him its Lovelace Medal and prize lecture in 2017. He holds honorary doctorates at the University of Edinburgh and the University of Sheffield.