So have you guys been following the rise of our new robot master?
If you haven't been following, Watson is IBM's 5 years in the making super computer. The goal of Watson was to essentially make the ship's computer from Star Trek, a computer that can give direct answers to questions stated in normal conversational English. On the surface it sounds like it should be simple enough, right? A computer has access to way more information than the average human and is tops when it comes to sifting through data quickly. Turns out though that more often than not it isn't having info that's Watson's problem. The clues on Jeopardy, like all language, are full of subtle meaning, irony, riddles and other linguistic complexities make teasing out the meaning of the question the biggest trick for the computer. Unless Watson understands what's being asked, it can't answer.
As an example, one question was,"From the Latin for end, this is where trains can also originate." Jennings gave the correct answer, "terminus." Watson answered wrong with "finis." Basically the computer saw the part about "From the Latin for end" which seemed to call for a cut and dried translation and missed the deeper, trickier meaning of the question.
Ultimately, I think the effort to make machines that match the abilities of humans only serves to make it clear how amazing we actually are and, honestly, how little we have to fear from a robot uprising. Watson cost billions and took years to build. It is made up of ten server racks holding 90 servers and 2800 processor cores giving it the computing power of about 2800 very advanced computers. And with all that power in can hold it's own with folks sporting a mere 3 pounds of squishy gray matter. Go team human!
(Keep in mind, I'm writing this Tuesday night and the final match is Wednesday. It's entirely possible that on Wednesday night, Watson trounces the humans, achieves self awareness and launches WWIII, killing us all and completely invalidating this blog post.)
So what do you guys think? Is trying to make machines in our own image a fool's game? Can we ever really make something that's smarter than us? Should we?
The Eleventh Plague
Coming from Scholastic, September 1, 2011