Phil: You also have to factor in the carbon footprint of an average American, not just their idling power output – and that includes raising and killing a lot of meat, veg, and all the concomitant transportation of said food, plus transportation of the humans, etc. Watson would still beat that by a mile, to be sure…

But IBM has stated that they expect a Watson class system to be about as computationally intensive as the average desktop will be around 2020 (give or take a couple years). And cellphone class in 5-10 years after that, assuming IBM can keep pushing the advances in computational infrastructure the same speed we’ve kept up the past hundred years or so (and I see no reason to doubt them, especially if they can figure out a way to ask Watson ‘How do we make a faster you?*’)

Watson doesn’t have voice rec because plenty of the clues require the human to read the clue to understand it’s true meaning – homonyms, puns, etc. Certainly IBM will be adding this, I believe they’ve stated they plan to spend the rest of the year tuning the system and adding bells and whistles before making him commercially available.

It bears repeating, Watson is programmed to not buzz in until he’s got an answer rated ‘confident.’ Jeopardy! champs like Jennings has stated that a common tactic for champs is to buzz in as soon as the man activates the if you think you can think of the answer and use the extra seconds to find the answer.

And congrats to IBM and Watson for such a bang-up job. It’ll be very interesting to see the uses Watson systems get put up to when other humans get their hands on them and start getting creative (or as William Gibson might say, ‘start getting lateral’)

*Yes, this is a joke… for now.

But IBM’s point is to show that Watson can process human language. I think they have done a great job.

Watson has no voice recognition capability, and that’s a whole ‘nother set of problems.

The application would be achieved by using Watson’s language parsing and search algorithms on information provided by to it by another subsystem with voice recognition capability.

That said, I don’t know why anyone’s saying that this will improve services provided via phone. Watson has no voice recognition capability, and that’s a whole ‘nother set of problems.

Big freakin’ deal.

In their advertising, IBM likes to talk about a smarter and more efficient world — a greener world. What’s the carbon footprint of their Colossus compared to those two tiny organic creatures? I researched that question this morning:

Each of the Power 750 Express servers consume a maximum of 1,949 watts, or 175kw for the whole computer, and that’s not even including the power for the cryogenically-cooled quantum hyperdrive shunt to actuate the solenoid connected to the Jeopardy! buzzer. A human being standing still would consume at most about 150w — less than 1/10 of 1% of Watson — and he just holds the buzzer in his hand.

Each of the servers generates a maximum of 6,649 BTU/hour. All together, the machine would generate about 600,000 BTU/hour and require about 50 tons of air conditioning. Presuming heat removal takes about as much energy as the actual computer, Watson would consume over 350kw of power. It can only operate in an environmentally-controlled server room. A human being can operate anywhere and would just sweat a little bit if it got hot.

IBM: when you’re able to make a green version of Watson that consumes less than 1kw of power, I’ll be impressed. Until then, I’d rather watch humans.

Notes: The Watson config was found on the Wikipedia. The power usage of the servers was found on IBM’s spec sheets. The cryogenically-controlled quantum hyperdrive shunt is entirely fictional.