Physics Colloquium: David Johnston
David Johnston from ThoughWorks will be the featured speaker at an upcoming physics colloquium. His talk is titled, Deep Learning, Artificial Intelligence and the Internet of Optimization.
Recent advances in the fields of optimization and machine learning are ushering in an age where algorithms, connected in global networks are involved in sophisticated decision making and are taking over a larger part of the human economy. Deep learning, a renaissance in the field of neural networks, is now fully competitive with humans on many tasks in voice and image recognition, natural language processing and translation. Reinforcement learning systems, aided in part by deep learning, are now capable of piloting cars and aircraft, orchestrating robotic motion, out-competing humans at sophisticated strategy games such as Poker and Go and can learn to do this largely without human interaction and training. While the excellent performance of such systems are demonstrable, the reasons behind the effectiveness of deep learning is still being worked out. The field of mathematical optimization is experiencing a boom of it's own. Optimization, the workhorse behind much of machine learning, is just as in important in it's own right. New algorithms such as the Alternate Direction Method of Multipliers enable the concept of an /Internet of Optimization/;a network where parties not only exchange information but seek out and discover economic trade-offs in order to arrive at an optimal solution toward a shared goal. Applications include the running of the smart grid, distributed routing of vehicles, operations of new markets and the automated running of supply chains. In this talk, Johnston will discuss these advances, how they are being applied and some of the open questions remaining. Besides applications in industry, he'll emphasize how scientists can benefit by mastering this new set of tools to increase their own productivity.
Admission is free. No ticket required.
Department of Physics and Astronomy