Home » Digital Insight

Commercial implications of IBM’s Watson supercomputer

By 15 February 2011 3 Comments

It appears that IBM has pulled off a public relations masterpiece with their computer hardware and software placed inside of the TV game show Jeopardy.  IBM’s “Watson” (which is a cluster of 90 high-end computers each based the four IBM Power7 processors and very sophisticated software), has managed to defeat two of the best human Jeopardy players in history and video clips are all over the Internet and YouTube.

The implication of Watson is that computers might soon have sufficient intelligence and computing power to listen and respond on more of a human level.  The commercial implications of this technology are huge in every sector such as IT and healthcare.  It won’t replace humans making the most critical decisions and working at the highest levels, but it could potentially serve as first level support.  IBM posted a series of videos that discuss the commercial ramifications of the technology used in Watson.

So now that IBM has demonstrated this level of software intelligence and given us a peek into the future, the question is how long before the computing power is cheap enough to deploy on a more massive scale.  We know that each one of the 90 servers that comprise Watson are about 3 times faster than the more common x86 servers used by businesses which means it’s probably 12 times faster than a high-end home computer or 48 times faster than a typical laptop.  But with 90 of those IBM servers, that means Watson is roughly 4320 times faster than a typical laptop.  Laptops will have to double in performance 12 times before they are as fast as Watson.  If we assume 18 months for every doubling of performance, it will take 18 years which means the world will be radically different when today’s children reach adulthood.


  • Nick R Brown said:

    I’ve watched the episodes so far and have enjoyed it as a spectacle, but there are a couple of things that I have picked up on:

    1) Is Watson held back from analyzing the question until it is completely read or as soon as the question appears on screen?

    2) Does Watson have an unfair advantage being that there is a seamless microsecond window between finding the right answer and electronically buzzing in, whereas human opponents must process the answer and then experience a nerve delay as they physically press a buzzer?

    3) As I mentioned on my twitter feed (@hownowbrowncow shameless plug), Watson taking key words and using an algorithm to select the best guess estimate from keyed in stored data is far less impressive in the grand scheme of things compared to two humans who have not only accumulated but understand knowledge even if Watson is faster on the draw. If the Internet has shown us anything, it is that information gathering is vastly different than knowledge.

    4) In the end, as Isaac Asimov philosophized in multiple novels, machines will eventual be able to interact with man, and they will be able to rationalize through statistics the most logical choice in any matter. But they will never comprehend emotion, and never be able to feel. And ultimately this gorge will always allow man to be superior.

  • George Ou (author) said:

    1. Are any of the contestants allowed to read the screen? If so, the computer obviously reads faster than any human. Now if it’s only a matter of listening, the computer would not have much of an advantage here.

    2. The computer can buzz so fast because there are 90 high end servers with over 2000 CPU cores working together to solve the problem. If it were just 1 server, it would take 90 times longer to respond which means it would probably not do well at all.

    3. It’s much more complex than just keywords.

    4. Superiority is subjective. I’m just waiting for Watson to crack an “I can’t do that Dave” joke the first time it talks to a Dave.

  • Bethie Blake-Lopez said:

    @Nick R Brown:

    1. Watson is fed the question (as a computer text file) at the same time that it appears for the humans to view.

    Of course, this process of feeding the file to the computer is near-instantaneous, so Watson begins its processing well before the humans do.

    Even so, the amount of ‘work’ that Watson has to do to ‘understand’ the question is so very great (which is what requires so many computers), that Watson still does only about as well as the humans at understanding the question.

    2. Watson is deliberately constrained by its makers from pressing its buzzer unless it decides that its answer has a high probability of being correct.

    The humans aren’t held back in this way. They can press the buzzer as soon as they are allowed (which is approximately 1/4 second after Alex finishes reading the clue.)

    Watson does have an advantage at the purely physical task of pressing its buzzer. It’s electromechanical setup is simply faster at button pressing than a human eye+brain+thumb.

    But since Watson can only press the button if it’s fairly sure it’s right, it didn’t always press the button before the humans.

    And of course, the humans took advantage of the fact that they could press the button even if they weren’t sure they had the right answer (and use the second or two that it took Alex to call on them to continue searching for the answer.)

    3. Watson is not merely performing a keyword search (like a Google search.) It does semantic analysis on the knowledge it possesses, and uses that knowledge in the context of the clue it receives to generate (and weight the correctness of) an answer. This is far from mere information gathering.

    4. This is an analog to the “god of the gaps” argument that theists use to bolster their belief in a supreme deity. They claim that there’s always something that man doesn’t yet know or cannot do (and that presumably only a supreme deity can know or do) that makes clear the need for a supreme being. This argument is constantly being diminished as man learns more and more about more and more.

    Humans have been steadily giving way to their machines for centuries. Watson merely shows that another gap between man and machine is about to close permanently. Don’t panic, but don’t rest on any laurels that leads you to think that man will always be superior to machines.