It's Elementary, Watson: The Robot Uprising Begins (Ken Jennings is First on the List)

First in a two part series brought to you by Ben Templeton.


- 44 Maagnum


“We’re going to revolutionize industries at a level that has never been done before.” That quote came from the Senior Vice President of a little company called IBM. You might have heard of it. Forbes lists IBM as the 33rd largest company in the world.

So what fantastic product was Dr. John Kelly discussing? Perhaps some exciting new social media technology, all the rage in the tech industry today? Or some fancy multi-touch screen that can double as a coffee table (nope, Microsoft has that one covered (quasi-pun intended)).

No, IBM (admirably, in my opinion) stays away from the trendy topics in “pop technology”, the Facebooks and the Smartphones of the tech world. They stick to their strengths, and devote their billions of dollars to weighty computing problems on the scale of mainframes and supercomputers. Past game-changers have included System/360, RISC computer architecture, and the invention of DRAM, amongst countless others.

IBM’s new innovation which will, in their words, surpass all of these is the hardware and software necessary to perform one task: play Jeopardy. Yes, that’s right, the unprecedentedly revolutionary technology that IBM is debuting on Monday (7:00PM EST) is designed to play a game show. “Watson”, as the computer is named (after the company’s founder, and as a reference to the sidekick of the famous information synthesizer Sherlock Holmes) operates without the internet or any human interaction. It hears the clues as its opponents do and responds in a clearly computerized but acceptably non-creepy voice. It mechanically presses a button to buzz in, like all of the other contestants.

You (as well as IBM’s investors, I imagine) are probably curious as to why playing a game show is worth the mountains of money and manpower that were devoted to this project. In today’s post, I’ll discuss the significance of this breakthrough, and in the next one we’ll look into the technology that powers it.

The IBM engineers describe Jeopardy as “a playing field upon which we could do some science.” The significance of Jeopardy is the way in which the clues are presented. Computer Scientists and AI experts call it the problem of “natural language”. The challenge is to understand the hints that are posed in normal English. Especially difficult is the style of Jeopardy’s problems, with subtle puns, jokes, and categories. Consider the following clue, from the ironic (to a supercomputer, at least) category “Chicks Dig Me”. “This mystery author and her archeologist hubby dug in hopes of finding the lost Syrian city of Urkesh.” The difficulty posed by this clue is apparent. Watson has to parse the category and the clue to determine that the response is a female mystery author. Given the tangential connection between Urkesh and the female mystery author, running a quick check of the Encyclopedia Britannica entry on Urkesh is unlikely to yield especially useful results. In his human brain, a trivia-master like Ken Jennings has a neurological database of information. If Mr. Jennings knows of a female mystery author married to an archaeologist, that is almost certainly sufficient for him to answer the question. Watson doesn’t have the capacity to make this connection so easily. Despite those linguistic pitfalls, Watson buzzes in confidently, and with a hint of digitally syncopated sass, responds “Who is Agatha Christie,” in this bit of video released.

In the words of Dr. Kelly (Senior VP), “The rate of growth of information is surpassing our ability to understand it and extract knowledge from it.” The breakthrough of Watson is that it can synthesize true knowledge from a mass of information the way that the human mind can and (if it proves superior to the greatest Jeopardy champions) possibly even better. In this day and age of information overload (mostly due to the internet), a computer that can process excess information and provide insight is an unbelievably powerful tool.

This capacity for synthesis has incredibly far-reaching implications. For example, a fairly high percentage of medical errors come at the diagnostic stage, and many happen simply because the diagnosis is too slow. A computer like Watson could plausibly listen to a description of symptoms, analyze its database of information, and provide a diagnosis. Teamed with a human doctor, this system could avoid time-consuming consultations and provide a concrete statistical analysis of certainty rather than simply a doctor’s gut feeling.

The possibilities are endless. A computer with a capacity to synthesize insight from as much knowledge as Watson could detect economic trends that lead to recession in a way that a human wouldn’t be able to, or (less glamorously) it could power the customer relations process.

So how does this technological marvel perform the magic of natural language processing? Tune in next time for a quick tour of the computing techniques used to make Watson a Jeopardy champion. Coming some time before Monday, to a blog which is stored on a server to which you may or may not be geographically proximal.