The ‘bit’ transformed how we compress, transmit and store information. But as more information is produced by machines, for machines, do the same rules still apply? Professor Deniz Gündüz, Professor in Information Processing, explores how information theory can help answer questions emerging in the age of intelligent machines. Please register to attend in person. A live stream link for online attendance will be available here shortly. We look forward to seeing you on Wednesday 11 November! Imperial Inaugurals are term-time lectures that celebrate our newest professors, recognising their academic journey and showcasing their research. Abstract In 1948, Claude Shannon showed how to put a number on information: how much a source can be compressed, and how much of it a noisy channel can carry. Those limits turned communication from a craft into a science, and everything digital: the Internet, mobile telephony, data storage, has been built around them. Shannon also gave us the ‘bit’: common currency of the information age, which let compression, transmission, computation and security be treated as four separate problems, solved by different engineers. Soon, most of the information crossing our networks will be produced by machines, for machines. How will billions of them communicate with each other: must they agree on a language in advance, or can they learn one? And do those old boundaries still make sense, when a network can compute, a compressor can learn, and the whole chain from sensor to decision can be trained as one? Biography Deniz Gündüz is Professor of Information Processing at Imperial College London, where he leads the Information Processing and Communications Lab. In his inaugural lecture he will argue that information theory is still the right instrument for answering such questions, drawing on his group’s work on machines that learn their own codes and protocols; private and secure computation, in which machines compute together without revealing what the

Description supplied by the event organiser.