The annual university competition is dedicated to accelerating the field of artificial intelligence (AI).
The UdeM group, MILA Team, is the only Canadian university involved in the competition.
“It’s a mixed feeling. I’m proud that our team is the only Canadian team to be accepted, but at the same time I am surprised and disappointed that our colleagues and friends at McGill University did not get selected to participate in the competition,” Iulian Vlad Serban, one of the UdeM students, said.
“These are truly great researchers, who we have been collaborating with extensively the past two years.”
The teams have to build a socialbot that can converse and and engage on popular topics for 20 minutes.
“The response to the Alexa Prize challenge has been incredibly positive, with applications from over 100 incredibly well-qualified teams from 22 countries,” said Rohit Prasad, vice-president and head scientist at Amazon Alexa.
“We are eager to see what students from the 12 sponsored teams around the world — including one team of undergrads — create with Alexa.”
Alexa is the voice service that powers devices such as Amazon Echo, Echo Dot, Amazon Tap and Amazon Fire TV.
The selected teams are:
- CMU Magnus from Carnegie-Mellon University in Pittsburgh, Pa., U.S.
- (Team name TBD) from Carnegie-Mellon University in Pittsburgh, P., U.S.
- eClub from Czech Technical University in Prague, Czech Republic
- WattSocialBot from Heriot-Watt University in Edinburgh, Scotland, U.K.
- Princeton Alexa from Princeton University in Princeton, N.J., U.S.
- BAKAbot from Rensselaer Polytechnic Institute in Troy, N.Y., U.S.
- Machine Learning @ Berkeley from University of California, in Berkeley, Calif., U.S.
- SlugBots from University of California, Santa Cruz in Santa Cruz, Calif., U.S.
- Edina from University of Edinburgh in Edinburgh, Scotland, U.K.
- MILA Team from Université de Montréal in Montreal, Que.
- Roving Minds from University of Trento in Trento, Italy
- HuskyBot from University of Washington in Seattle, Wash., U.S.
All teams will receive a $100,000 stipend for participating and the group with the highest-performing socialbot will win a $500,000 prize.
An additional $1 million will be awarded to the winning team’s university if their socialbot is able to converse coherently for 20 minutes.
“When we first heard about the competition, we were very surprised but also enthusiastic,” explained Serban.
“We spent a few days discussing whether we should participate or not. After all, taking part in a one-year research competition is not an easy decision.”
The UdeM team consists of 12 PhD students and professor Yoshua Bengio:
- José Manuel Rodríguez Sotelo
- Sarath Chandar Anbil Parthipan
- Iulian Vlad Serban
- Zhouhan Lin
- Sandeep Subramanian
- Saizheng Zhang
- Chinnadhurai Sankar
- Taesup Kim
- Alexandre de Brébisson
- Sai Rajeswar Mudumba
- Tong Che
- Nan Ke (Rosemary)
“Our research lab is very international, so it’s not surprising that our team comprises students from different countries including India, China, Mexico, New Zealand and several European countries,” said Serban.
He said the idea of artificial intelligence has puzzled him for years.
“I suppose my interests in AI first started when I was kid watching movies like A.I. Artificial Intelligence and the original Star Wars movies from the early ’80s,” Serban said.
“In a sense, one could say that my current research is leading towards building characters like David and C-3PO.”
When asked how the team plans to build a socialbot capable of having conversations, Serban was purposely coy.
“As our competitors may be reading what you print, I can only give you the big picture,” he said.
READ MORE: Do humans trust robots too much?
“Current systems are nowhere near able to converse naturally with humans for a long period of time. They do not really understand how the world ticks and they cannot maintain an interesting conversation topic.”
“We will need both new research and engineering to build an intelligent socialbot in order to win the competition.”
Serban explained AI is already integrated in our everyday lives — from smartphones to self-driving cars.
Montreal is a growing hub for AI research, he added, and it’s not just the universities that are pursuing the research, there are also start-ups dedicated to it.
“AI is being used to detect faces and smiles on your phone, and to process millions of images and lines of text on popular websites. As you might have heard, self-driving cars — powered by deep learning — are now driving around in the U.S.,” he said.
“I believe it’s only a matter of time before we get intelligent personal assistants — like Siri, but much smarter — which can help us in our daily life.”
Winners will be announced next year at a special session of re:Invent 2017.