C'est un meetup Hors Série exceptionnel organisé conjointement avec l'Ambassade du Canada en France. Nous serons accueillis dans les locaux de Xebia (merci à eux! et leur événement dataXday) Nous commencerons le meetup à 19h00 et ouvrirons les portes vers 18h30. La video en streaming est ici (les presentations seront sur cette page avant le meetup)
Voici le programme technique:
- “ How we solve Poker” by Prof. Mike Bowling, Computing Science, University of Alberta (15 minutes).
- « L’AI at the University of Alberta » by Prof Mario A. Nascimento, Full Professor and Chair of the University of Alberta's Department of Computing Science.
- Présentation de Element AI par Sarah Larose, Public Policy & Government Relations chez Element AI. Site: http://element.ai
- Deep Learning, Artificial Evolution and Novel AI Behaviors by Vadim Bulitko, Associate Professor at the University of Alberta, Department of Computing Science. Site: http://agi-lab.net
- Understanding the role of exploration in heuristic search by Martin Müller, Computing Science, University of Alberta.
- "Natural Language Processing at the University of Alberta" by Prof. Greg Kondrak
- Présentation des opportunités de bourses pour étudier et faire de la recherche au Canada par Fevronia Novac Chargée de programmes, Service Enseignement supérieur et Jeunesse, Ambassade du Canada en France
Le détail des quatre présentations techniques:
Title: “ How we solve Poker”
SPEAKER: Prof. Mike Bowling
Cepheus is our new poker-playing program capable of playing a nearly perfect game of heads-up limit Texas hold'em. It is so close to perfect that even after an entire human lifetime of playing against it, you couldn't be statistically certain it wasn't perfect. We call such a game essentially solved. This work just appeared in Science. You can read the paper. You can query Cepheus about how it plays and play against it. Or you can read the many news articles on the result. Site: http://poker.srv.ualberta.ca/
SPEAKER : Vadim Bulitko, Associate Professor at the University of Alberta, Department of Computing Science
ABSTRACT: Artificial Intelligence is rapidly entering our daily life in the form of smartphone assistants, self-driving cars, etc. While such AI assistants can make our lives easier and safer, there is a growing interest in understanding how long they will remain our intellectual servants. With the powerful applications of self-training and self-learning (e.g., the recent work by Deep Mind on self-learning to play several board games at a championship level), what behaviors will such self-learning AI agents learn? Will there be genuine knowledge discoveries made by them? How much understanding of their novel behavior will we, as humans, be able to gather?
This project builds on our group's 12 years of expertise in developing AI agents learning in a real-time setting and takes a step towards investigating the grand yet pressing questions listed above. We are developing a video-game-like testbed in which we allow our AI agents to evolve over time and learn from their life experience. The agents use genetically encoded deep neural networks to represent behaviors and pass them onto their off-springs in the simulated evolution. A separate deep neural network is then trained to watch the simulation and flag emergence of any unusual behaviours. We expect to study emergence of novel behaviors such as development of friend-foe identification techniques, simple forms of communication, apprenticeship learning and others.
SPEAKER: Martin Müller, Computing Science, University of Alberta
ABSTRACT: I will give a brief overview of recent work in my research group. While the applications are diverse and range from games and Monte Carlo Tree Search to SAT solving, a common goal drives much of the work: to better understand the use of exploration in very large search spaces.
Liked this entry ? subscribe to Nuit Blanche's feed, there's more where that came from. You can also subscribe to Nuit Blanche by Email, explore the Big Picture in Compressive Sensing or the Matrix Factorization Jungle and join the conversations on compressive sensing, advanced matrix factorization and calibration issues on Linkedin.