Jeopardy: Human vs Machine

I’m not a fan of Jeopardy. But a (not so) recent exhibition match caught my attention. It’s a match between IBM’s super computer Watson and 2 of the best Jeopardy human contestants ever, Ken Jennings and Brad Rutter.
Imagine IBM’s Deep Blue vs chess grandmaster Garry Kasparov. Except this time, the game is Jeopardy.

Watson is a question answering (QA) computing system. IBM describes it as an application of advanced Natural Language Processing, Information Retrieval, Knowledge Representation and Reasoning, and Machine Learning technologies to the field of open domain question answering which is built on IBM’s DeepQA technology for hypothesis generation, massive evidence gathering, analysis, and scoring.
Watson is made up of a cluster of 90 IBM Power 750 servers (plus additional I/O, network and cluster controller nodes in 10 racks) with a total of 2880 POWER7 processor cores and 16 Terabytes of RAM. Each Power 750 server uses a 3.5 GHz POWER7 eight core processor, with four threads per core.
Wow…. A geek’s wet dream.
Watson had access to 200 million pages of structured and unstructured content worth 4 TB of disk storage, including everything in Wikipedia. Watson was not connected to the Internet during the game. It would be useless for Watson to be connected to the Internet during the games too. Watson needs to compute the answer fast and the Internet is just not fast enough. Everything needs to be stored locally for Watson to compete with humans.

Warning: If you don’t want any spoilers, please stop reading now.

And in the end, the Machine won. Which isn’t really a surprise to me after watching seeing it in action. The only reason why Watson will lose is if it hit a programming bug midway through the game. Else it would be pretty hard for humans to win.
For a start, Watson’s computing speed is amazingly fast. If you notice, most of the questions that human answered are questions that Watson is not confident of answering correctly. Watson will not press the buzzer if it is not confident on the answer. The accuracy of Watson is also very impressive. It only made a few errors. For most cases, Watson knows what it knows and what it doesn’t knows.
At the end of the 3rd day, Watson got $77,147 while Ken Jennings and Brad Rutter got $24,000 and $21,600 respectively.

Is this the beginning of the rise of the machines? Will Watson eventually lead the machines and take over earth? Hmmmm…… nah, not a chance. We watched too much Hollywood movies. There’s no need to run out and buy the Think Geek T-Shirt that says “I, for one, welcome our new computer overlords“. Although I got 1 myself just in case. Hahahaha….
Actually, I look forward to 1 day when Watson have speech recognition capabilities. Right now, the questions are fed to Watson in text form. Actually we already have very good speech recognition algorithm out there now. But I guess it is not fast enough. IBM might need to add more Power 750 servers if they want to add speech recognition. If Watson have speech recognition, it would also be able to hear the answer from other competitors and avoid the same error. It may even learn from other players during the game.
Watson is a milestone in Artificial Intelligence. Natural language is one of the hardest puzzle. Yet Watson demonstrated that it’s able to understand the complex Jeopardy game pretty well.
So what’s next for Watson? Survivor? Hmmm……


Leave a Reply

Your email address will not be published. Required fields are marked *