IBM's Watson Computer System appeared on Jeopardy Game show in three episodes from Feb 14 to Feb 16, 2011 and beat two contestants, the greatest winners of all time. It generated a lot of interest again in the power of artificial intelligence. It has been a ever-lasing wish of scientists to create a machine which can match the power of human intelligence. Well, Watson is one step forward in that direction, but we are nowhere to achieving that. Before this event, Deep Blue has been in news in the last decade of 20th century. Deep Blue won a six-game match against world champion Garry Kasparov. These are the most popular events where machine was pitted against human and it won. It is not too difficult to build a machine, which can beat human in a narrow knowledge domain, but to build a machine, which has to face questions from open ended domain requires immense computing power and very advanced algorithms. Another challenge is to design search and retrieval algorithms to mine the huge data in a meaningful way. The answer to a question has to be represented in human terms. It should be understandable to a human as if it has been answered by another human. Machine has to deliver the precise response and synthesize, integrate, and go through the breadth of knowledge.
Watson came out of IBM’s DeepQA project, whose aim is to illustrate easy accessibility of natural language contents. The project encompasses a range of research topics: Natural Language Processing, Knowledge Representation and Reasoning, Information Retrieval, Machine Learning, Massively Parallel Computation. The project is led by Dr. David Ferrucci and a team of 25 researchers drawn from various fields of computer science.
An interesting question, people have asked is why we need Watson like machine when all the information is available through a search engine. We have to remind ourselves that web search engines are designed to return ranked results based on search engine’s algorithm. A human has to filter through these answers and select the ones appropriate to the question. This is where Watson beats search engine. Watson had its weird moments; famous one, when it answered, “What is leg?" reference to a clue about an Olympic oddity in which the athlete was missing a leg.
Watson is designed using UIMA (Unstructured Information Management Architecture), which is now open source and available at uima.apache.org. Watson needs only 500 GB to compete on Jeopardy. It doesn’t seem like a lot, but it doesn’t require any image, audio, and video files, just plain text. DeepQA high level architecture is given in AI magazine's Fall 2010 issue [1].
What powers Watson:
AI Magazine article:
Future of Watson (TED webcast):
FAQ on Watson:
Watson Link:
http://www-03.ibm.com/innovation/us/watson/
[1] Building Watson: An Overview of the DeepQA Project, David Ferrucci, Eric Brown, Jennifer Chu-Carroll, James Fan, David Gondek, Aditya A. Kalyanpur, Adam Lally, J. William Murdock, Eric Nyberg, John Prager, Nico Schlaefer, Chris Welty, vol. 31, issue 3.
[1] Building Watson: An Overview of the DeepQA Project, David Ferrucci, Eric Brown, Jennifer Chu-Carroll, James Fan, David Gondek, Aditya A. Kalyanpur, Adam Lally, J. William Murdock, Eric Nyberg, John Prager, Nico Schlaefer, Chris Welty, vol. 31, issue 3.
No comments:
Post a Comment