Blog entry by: Parag Singla
IJCAI, International Joint Conference on Artificial Intelligence, is the premiere annual international conference on topics relating to AI. IJCAI-2015, the 24th in the series, was held in Buenos Aires, Argentina from July 25 to July 31. The first 3 days of the conference were occupied by various workshops and tutorials, followed by the main conference starting Tuesday the 28th. This was the first time that the conference was held anywhere in South America.
Being one of the main AI conferences, the conference showcased publications on a broad range of topics including Knowledge Representation, Planning, Automated Reasoning, Uncertainty in AI, Social Networks, Natural Language Processing, Search, Game Playing, Satisfiability, Machine Learning, Computer Vision, Multi-Agent Systems and AI & Arts. The conference also hosted the AIBIRDS competition where the goal was to come up with an AI agent which could do better than humans in the game of Angry Bird.
In addition to the paper presentations via talks and posters, every day there were few invited talks by prominent researchers and academicians in the area. One of these was by Ariel Procaccia, who also received the IJCAI-15 Computers and Thought award. Ariel talked about his research on computational social choice, the problem of kidney exchange, and advanced techniques for fair division. It was quite exciting to listen to him, especially the practical applications of his work. There was another very interesting talk on Google DeepMind project and the cool deep learning technology they are working on. DeepMind was a startup based in UK which was bought by Google in 2014 for more than few hundred million dollars (number as quoted by WikiPedia). Bart Selman, a recipient of the IJCAI-15 John McCarthy award, talked about the history of the Satisfiability solvers (and the associated complexity) dating back to the later half of the 20th century, and continuing all the way to the modern day solvers. He also talked about how modern versions are becoming more and more sophisticated with a significant amount of focus on structured problems. Manuela Veloso presented her enthralling work on Mobile Robots and how they could be employed to help with everyday tasks.
Two publications got the IJCAI distinguished paper award. One of them was from the University of Washington, Seattle which presented a novel technique for non-convex optimization using the idea of recursive decomposition. The other work, done at CMU, presented a Bayesian approach for posterior estimation in an active learning setting. The conference also featured recent key publications/works from other related conferences such as UAI (Uncertainty in AI), AAMAS (Multi-agent Systems) and KR (Knowledge Representation), among others.
In terms of Indian representation, there were several papers authored by people from India. The host institutions included IIT Delhi, IIT Bombay, IIT Madras, IISc. Bangalore, IIT Roorkee and few corporates including IBM IRL and Microsoft India.
The conference featured two panel discussions on issues relevant to AI. One of them was on “Future of AI”. The panel included various leading AI researchers including Stuart Russell from Stanford (author of the well-known textbook on AI “Artificial Intelligence: The Modern Approach), Manuela Veloso from Carnegie Mellon and Michael Wooldridge from Oxford. It was a very interesting panel discussing about the implications of the latest AI techniques for the society and the perceptions thereof. There was a great deal of concern about the celebrities’ such as Stephen Hawking’s recent statements about AI posing a potential threat to the modern day society with the invention of increasingly sophisticated automated technologies. A strong need was felt to create widespread awareness about the huge positive impacts of AI for the society and how just like any other technology, there were potentials for misuse, but the decision about what purposes the technology would eventually be used for, essentially lay in society’s hands. The second panel was on whether the famous Turing Test for intelligence was still meaningful in the present day context, or if something new needed to be invented as a more meaningful test of human level intelligence.
In addition to the technical program, the conference hosted a few social and cultural programs as well. There were city tours organised with the help of local travel agents. I got an opportunity to partake in one such city tour. It was a great experience to know about various different parts of the city, the history associated, the rich neighbourhoods, the local soccer teams and the tango district. Argentina being home to the famous tango dance, there was a banquet organised on one of the evenings, where the dinner was accompanied by an amazing tango performance by local artists. It was a great fun to be part of this event.
Local people mostly speak Spanish, so language was bit of an issue while going around, but we managed ok. In terms of food, there is a lot of availability of wide variety of non-vegetarian food. Though the choice of vegetarian delicacies was limited, there was enough for a vegetarian to survive happily.
Overall, I had a great fun attending the conference, networking with various other researchers and enjoying the hospitality of Buenos Aires!