It's Elementary, My Dear Watson

IBM’s Watson robot triumphed on Jeopardy recently. While the geek guru of trivia stumbled a bit on the first day and had a tendency to answer really, really wrong when it was wrong and bet in odd, seemingly random cash sums, the computer came roaring back to leave the champion human contestants in the dust.

Did any of us really question that the computer would beat the humans at trivia?

Computers are like little digital elephants. They don’t forget. I once memorized all the U.S. presidents, their terms in office, their vice presidents, their major accomplishments and their major failures for a college class on American history. But, I’d be hard-pressed to remember much of that list today. The human brain doesn’t retain knowledge it doesn’t use often---at least, not very well and not completely or I wouldn’t have such trouble with maps of Africa.

If a computer “brain” had the gaps in it that my brain has (especially for presidents between Lincoln and Roosevelt #1), I’d be sorely disappointed in that computer. I’d think it a colossal failure if its digital circuits sparked and burnt out the way my organic ones often do.

But, Watson triumphed, and at least IBM cheered. Personally, I’m not sure what to make of Watson, exactly. I hear that the next stop on Watson’s smarter planet world tour, of sorts, is to help with medical diagnosis. Watson can take in and hold a lot of factors, facts, stats and charts that a human doctor cannot. I can see how that might be useful. But, Watson is lacking one thing you can’t ever teach a computer -- human reasoning.

Yes, you can program Watson with all sorts of data. You can tell him that A + B + Ru + the speed of light over the square root of pie + a fever equals Z and in Watson’s world, the answer must always be Z given those factors. Humans are a little better at realizing that factors and data aren’t the whole picture.

Still, Watson can be a great help. In our industry, IBM has pitched that Watson can assist with energy efficiency in the field and in educating consumers about use. (And, oh yes, kids and geeks alike will love learning via Watson. He’s like the world’s most advanced See ‘n’ Say.)

IBM also posits that Watson can help make decisions in utility control rooms, mostly due to his ability to understand “natural language” questions. This means you can ask Watson questions in English, and he can translate it, basically, into the digital format he works in. But, bear in mind that, essentially, Watson isn’t a native speaker of English. We natives talk weird. We drop syllables. We don’t form complete sentences. In talking, we often get the gist of things, not every word. We grasp the overall concepts with language. We work in big pictures, and Watson sees in singular, linear details.

Singular, linear details are great for a computer, but not so great in a world built by humans for humans with human-caused and human-impacted problems which careen off the logical line in nifty and tangential ways.

IBM says that for energy field personnel, Watson’s natural language interface can help with troubleshooting and safety. Personnel can use their mobile phone connection to query databases of information and "Watson can suggest the correct action to fix a power disruption and identify causes of certain problems in the field," the company said in a press release.

I will be very interested to watch Watson evolve in the field as he works with doctors and energy personnel. I’m sure he will continue to add to his extensive bit of knowledge -- those details of facts and data and trivia. But, I really want to watch Watson move past the elementary item of natural language interface to something more analytical and more profound. If IBM can truly accomplish that, then as Ken Jennings noted during final Jeopardy with Watson, “I, for one, welcome our computer overlords.”

OK. Well, maybe I’ve seen the Terminator series too many times to go quite as far as Ken. But, I would welcome all the intelligent help we can get in power control rooms.

Bring on the rise of the power robots, IBM.