Watson: What it Means to Me

I cannot recall the last time I tuned into Jeopardy! for more than a couple of questions, yet over the last three nights I found myself glued to the television for the “exhibition match” of the century. IBM’s supercomputer “Watson” defeated the two most successful players in Jeopardy! history in a three day, two game, cumulative dollar challenge. Humanity was defeated in the arena that was designed specifically for humans. I should be devastated, rather I should be digging a massive hole in my backyard to house my future Robageddon bunker.

But for some reason I am fairly happy with the success of Watson. There are of course a couple of qualifiers to my somewhat reserved glee. First off, I am assuming that Watson is not the first massive node in a computer network that could achieve sentience. Secondly, I unlike Ken Jennings (sore loser: http://slate.me/fTBtVE), do not believe Watson is a harbinger for the same kind of layoffs in knowledge based work as automation was for factory jobs.

As far as the apocalypse is concerned I believe Watson, impressive as it is, is no where near powerful enough to think independently of its human programmers. Watson’s quirks, minimal as they were, revealed the limitations of its capabilities. Watson was never 100% sure of anything, how could it be, it has never truly learned anything. The people who program Watson learned how to refine and improve the Watson algorithms which in turn made the computer better at correlating information.

However, as any psychology major can tell you, correlation does not equal causation, and it is causation that represents the wall of Watson. Watson knows that a blanket is related to warmth because the words appear near each other a significant amount, but Watson cannot understand “why” they do, because it has never been cold. Watson can see the connections between kiss, hug, love, and need but it could never produce an original sonnet about the joys of love because Watson does not experience emotions or feelings, it reports on them.

Now that I’ve spent some time tearing it down, I ask what is Watson good for? Watson, I believe is good at what we are not. Human beings are fantastic at extending our knowledge base. People have invented, reinvented, combined, recombined, composed and then improved upon ideas since the dawn of our species. Some times our progress has been slow (Dark Ages) while in other times it has been unbelievably fast (Renaissance). The key to progress has always been access to knowledge. No one person can read through all key works on any topic of importance to human advancement and then synthesize the results into an original work because the sheer volume of material would take lifetimes to consume. The amount of knowledge we currently have as a species is mind-numbing, Watson has data banks of 15 terabytes and that only scratches the surface.

What Watson could do (if it works as advertised) is comb through our knowledge base and look for information that correlates to a particular problem. It could then itemize the information according to relevance of the query thus giving humans a useful starting point at the genesis of their research rather than years, or worse lifetimes, into it. Watson can keep us from reinventing the wheel with the same exact problems of the last one. It can show us the best wheel in existence and then we can improve upon that design. Of course, that is my vision for this machine, maybe IBM has other plans for it.