ACM A.M. Turing Award 2018: Yoshua Bengio, Yan Lecun and Geoffrey Hinton

we're willing to be wrong to be completely humble to the fact and they'll be completely Hardy to man do you mean by that well haughty to man no otherwise you get laughed out of a right idea in 1943 neurophysiologists Warren McCulloch and mathematician Walter Pitts wrote a paper modeling the first neural network using electrical circuits for several decades into the early 1980s research and interest in artificial intelligence was almost non-existent while most remain skeptical Geoffrey Hinton yoshua bengio in yawn laocoon were convinced in the potential of neural networks there were a bunch of crazies like me and Jana Yahshua who all believe that great big neural networks with many layers of feature detectors could actually learn everything but it just seemed utterly implausible especially to people who thought the most knowledge was innate that you could learn all this stuff with know just from random weights and I think people quite not unreasonable to think that this stuff was crazy but actually we were right but it's when I started reading those connectionist papers these are early in your met papers of the 80s that I realized this was the area I wanted to do research in I was I was definitely thinking I was I was right the whole time it's just that you know it would take time to convince the the you know all of our colleagues because what are the main issues is that the computers were not powerful enough really to kind of scale this up and then the software that was required to run those things was fairly complicated to write at a time we we had invested a time in in writing them but nobody else wanted to basically you know put the effort the field of AI experienced a second winter from 1987 to 1993 but the winter for neural networks lasted even longer into the mid 2000s enthusiasm and investment in artificial intelligence once again disappeared however Hinton Ben Jill and laocoon continued to brave the storm and in 2012 their work finally received wide mainstream acceptance after deep neural networks were used to identify objects fast sir than its competitors during the imagenet challenge fast forward to the present in Hampton Ben geo and laocoon are now the 2018 ACM a.m. Turing Award recipients for the work on artificial neural networks but there is still much work to be done in making AI smarter which means - they have less common sense than a house cat and we need to fix that and I can see many places where I would say current AI is stupid and is lacking an understanding of the world around us so the bulk of my research at the moment is on something I call self supervisor running it's kind of a new mode of learning that is picking up within the community which the hope is that we will be able to train machines and have them learn in ways that are similar to the way animals and humans around and so a lot of my current work is about how do we train systems so that they better understand how the physical world works how the world of humans can be understood by machines so that they can interact with humans the thing I'm most interested present is an idea called capsules which I've been working on for the last few years the idea is to try and build into a neural network a little bit more knowledge about the world so that a neural network automatically can recognize things from different viewpoints the future of AI looks promising and what many are calling the next big thing with endless possibilities [Music] what economists are telling us is that AI is what they call a general purpose technology gbt sort of like you know the steam engine back in the old days or you know the electric motor and things like this is going to progressively disseminate in in all corners of the economy and lead to more productivity which means more wealth produce per our work so another thing that gets me really excited is what is happening right now in Montreal around Miele you know I started from a small lab with a handful of students and they grew and grew and grew and now it's a huge thing it's whole community it's hundreds of people in this institute it's a whole ecosystem of BI companies and startups and universities that are working together a journey that has spanned more than four decades has finally come to fruition and has now been recognized by the ACM AM Turing award it's very nice to win it as a group it's always more fun being part of a successful group than sort of being all on your own it's not a price for a person it's not a price for the three of us who got it it's a price for the whole community that has made it possible to achieve progress we've seen in the last decade or so I'm so glad that that whole area is being recognized if you have an idea and it seems to you it has to be right don't let people tell you it's silly just ignore them learn more about the Turing award recipients in the June 2019 communications of the ACM

Channel: Association for Computing Machinery (ACM)

Views: 10,496

Keywords: Turing Award, ACM Turing Award, 2018 ACM Turing Award

Id: HzilDIhWhrE

Length: 5min 46sec (346 seconds)

Published: Wed Jun 19 2019