IBM’s Watson supercomputer became famous when it beat the top “Jeopardy” champions on TV. What folks didn’t know is that Watson did that with significant handicaps that wouldn’t be applied were the same system actually deployed to answer real questions.
What Watson is particularly good at is providing the right answers to questions, which apparently, humans (at least, individually) are really bad at. As you watch the U.S. Congress circle the drain, you’d have to think that Watson would make stupid politicians stick out and make it far harder for either side to sustain stupid partisan positions. Let’s consider a world governed with Watson’s help and hope it doesn’t turn into Colossus: The Forbin Project.
I’ll close with my product of the week: an offering from OnLive that will make your iPad or Android Tablet into a gaming PC or console.
The Power of Watson
Watson is one of the first big steps toward machine intelligence using existing technologies. What it was designed to do was to parse a question into component parts, analyze the components, find matching results, and finally attach a value to those results so that the most likely answer could be brought to the surface.
Watson was crippled to play in the “Jeopardy” TV show because it needed to be as limited as a human. So it had a physical rather than electronic connection to the buzzer, and it was only allowed to use information loaded into local storage. It couldn’t use the Web or even IBM’s vast knowledge databases. Neither of those limitations will be applied to the eventual productized system.
A division was created around Watson and an initial market identified for it — the medical market. When given a series of symptoms, Watson will identify the top likely causes, which will then define a series of tests to confirm the actual diagnosis.
This is as opposed to current diagnostic practices, which involve a doctor or doctors listening to patients describe their symptoms and then guessing, based on knowledge and experience, which causes are most likely. The problem with this way is that people suck at this approach, which is why so many people are misdiagnosed.
Argumentative Theory and Confirmation Bias
Apparently, we are hardwired to make bad decisions, because our approach to setting a position is to first guess what the answer is and then look for information that supports it. This is exactly the opposite of how a decision should be reached — starting with research and ending with the answer.
This is also likely why we often have researchers who publish what is later found out to be fraudulent research. They are so convinced they are right, they can’t accept that they aren’t. The science behind this is called argumentative theory and the direct cause is confirmation bias.
The first suggests that it is more important that we win an argument than it is to be right, and the second suggests that we take positions first and then ignore anything that disagrees with that position. Makes us look like a race of idiots.
You can catch yourself doing this when you have an argument with a sibling, parent, friend or spouse and find yourself looking for hours on the Web for that one thing that proves you right. Still believing, you start to think that the Web is biased, or you just aren’t searching right.
You can also see this when you look at someone you generally disagree with and can’t see anything they do that is right even though it isn’t possible for anyone to be 100 percent wrong any more than it is for them to be 100 percent right. You saw this happen in politics when Herman Cain was asked about Obama’s Libyan policy. Even though what he eventually said he would do was what Obama did, when framing his initial position, he tried very hard to disagree with it.
What he should have said — and he would have looked more intelligent saying it — was that Obama did the right thing there, but instead he looked like an idiot trying to disagree with something he agreed with. Granted, he has had other issues since — but even there, it appears he took the early position that his alleged dalliances wouldn’t be a problem, and boy was he wrong.
Watson and Making Congress Obsolete
Can you imagine a political debate with Watson on stage? Let’s take global warming. There is only one most probable answer, and the nine next-most-probable aren’t likely to be the opposite. The two politicians state their positions, and Watson points out, with overwhelming backup data, that one of them is an idiot.
You want to argue tax policy? You want to pick on someone’s birth certificate? Argue healthcare, abortion, whatever? Watson would have behind it most of the world’s current and historical information. You want to talk about what the forefathers would have thought about Christmas and politics? (Congress used to work through Christmas.)
Any question on virtually any subject, including whether we are likely to be invaded by space aliens or whether a congressman was behaving inappropriately, Watson would likely know the top high-probability answers — and not based on party, religious beliefs, special interest money, blackmail or sexual favors.
Of course, if I were building this type of system for political debates, I’d build in the raspberry sound which, I expect, it would use a lot for politicians who said really stupid things. I can think of several folks in both parties, in and out of office, I could name it after.
There’s at least one error in the story (maybe Watson would have caught this). Congress did NOT work through Christmas. I’m not sure where this idea came from. They may have not have taken an extended recess like today’s Congress, but they always took off Christmas day and usually Christmas eve.