Symmetry meets AI
Artificial Intelligence, especially its branch of Deep Learning, is a disruptive technology whose implications we are just starting to fathom. Due to its impressive successes, AI is somehow regarded as a magical blackbox, able to surpass human ability in an inexplicable way. We, theoretical physicists, do not like blackboxes and enjoy breaking things. So in this talk I am going to start breaking one of these boxes and watch an AI learn the presence of symmetries when no information on symmetry was provided. It isn't magic, but it is the AI doing its best to perform a task and finding the smartest route to success. Once we have learned the inner workings of the AI, we are going to use it to develop an algorithm which inspects artistic works and assigns a degree of symmetry to them. This is one step towards a much broader aim of understanding the inner workings of AI, like developing new ways of learning or controlling for the presence of unethical biases.
Friday, 10th February 2023, 14:30, Aula Magna and Online