In India, not many would know a popular American game show called ‘Jeopardy!’. So, to set the context – Jeopardy! is a quiz show that touches art, history, world play, pun, science, sports, movies and everything else. The twist is that instead of being asked questions, the contestants are given answers for which the contestants have to give questions. Merv Griffin created this show in 1964 and now it is in its 27th season on CBS in the United States. For instance, one of the clues given to contestants was – “This mystery author and her archaeologist hubby dug in hopes of finding the lost Syrian city of Urkesh”. The solution was – “Who is Agatha Christie?”

Now that you know the show – I am sure you understand how difficult it is for a computer to get the solutions for the clues that are presented. Not, if it is Watson.

Watson, named after Thomas J. Watson, the founder of IBM, is the latest Artificial Intelligence (AI) machine created by IBM to answer questions posted in natural language. Very similar to the onboard computers featured in Star Wars and Hitch Hikers Guide to the Galaxy.

Besides the really smart AI software that took four years in the making, the machine is powered by 10 racks of IBM Power 750 servers with close to 3000 cores running at 80 teraflops and 15 terabytes of RAM. Over the last four years, its developers have fed all kinds of information, including books, encyclopaedias, dictionaries, novels, plays and lots more into it to help it solve all kinds of riddles thrown at it with ease. The machine uses multiple algorithms to come at a solution; the solution that is computed by most number of algorithms is given the highest probability rating. This is what is given as the answer.

In 2007, when the team tested its capabilities for the first time it gave correct answers only 15% of the time. Now, when it went against two of the all time best Jeopardy champions in a demo round, it came out trumps with winnings of $4,400, the two champions Brad Rutter and Ken Jennings went home with earnings of $1,200 and $3,400 respectively. During this game, none of the contestants got a question wrong but it was the contestant hitting the buzzer the fastest, won. Watson, had its own pneumatic finger to push the button. A full show of Jeopardy! is planned for February 14, 15 and 16, 2011 where Watson will play the champions again.

Now, imagine a resource like Watson available to your enterprise to solve problems or give faster insights into data that was difficult to think of before. Currently, you fight with dashboards and reports to give you all the BI data that you need, and creating a new query can be quite difficult if not sometimes impossible.

For BFSI and Telecom companies, such a resource could be invaluable to ensure that they are always ahead of the competition and are able to get valuable insights into their customer data. How about fraud detection?

The possibilities are endless. What IBM would need to demonstrate is that this machine is fully capable of cross indexing and providing solutions to enterprise data and not only smart answers to seemingly impossible trivia questions. Also, it will be important to understand where and how will this machine get its data from in the organisation, and how long will it take to configure this to individual needs.

While you ponder over this question, and try to imagine the capabilities of the machine – I’ll go back to solving today’s Sudoku. Or rather I would leave it to Google Goggles to do it for me. Yes, my phone may not answer Jeopardy! questions, but it can certainly solve this Sudoku puzzle.

Geetaj Channana