It is used to be a trope of AI criticism to point out the fact that whereas high level tasks seemed to be amenable to AI techniques (e.g. playing chess, or other intelligent games) tasks trivial for a human child (describing what is in a picture) seemed many decades away. As it turns out, since 2012 there has been a sharp increase in the quality of artificial image description, with a steep influx of works still improving the performances and versatility of image analysis algorithms. We shall try to review and explain the specificity of the class of algorithms that triggered this progress (Deep Learning).
Sunday, June 4,18:00. Building 502, Room 43. For further information e-mail orenharman@gmail.com or phone 03-5317756
