Computer Vision News - July 2021

Ann Kennedy 43 Best of CVPR 2021 of animals versus pictures of non- animals, and it worked! It was kind of just magical! You gave it all of these pictures, and suddenly, it was able to distinguish these two things. That was just wild to me, that you could build something like this to solve a task and not really understand how it was solving it. It would just magically work. Does being a researcher in your field make you believe more in magic? Oh no! It’s not magic. It’s just simple learning rules implemented in a big network of neurons. Do you notice any common mistakes that junior researchers make? Attending talks and conferences and getting out there is something that made a big difference to me early on in grad school. It’s common to have the feeling that “I shouldn’t go to this meeting until I really understand what people are working on”, until I know the field. But you go to these meetings because you don’t know the field! Attending conferences and asking stupid questions is how you get to know an area of research. You shouldn’t hold "It’s not magic. It’s just simple learning rules implemented in a big network of neurons."