Computers aren't really that dumb, are they?

In April, I used my computer to complete and file my taxes (because when it comes to taxes, I don't really know what I'm doing). Sure, I gave it some necessary information, but it compiled it for me in such a way that would make the IRS happy. I couldn't have done it myself. Another thing I can't do myself is make a list of the movies that Arnold Schwarzenegger has appeared in. Seeing as how I can only think of a handful (most of which are the Terminator movies), my computer would undoubtedly do a better job than I. Why then, can't my computer tell me what Schwarzenegger and his "Twins" co-star, Danny DeVito, are doing in the following photo?

Computers are considered omnipotent juggernauts of processing power but the truth is, they're pretty dumb. They are the undisputed chess champions of the world, but struggle with basic English. IBM's Watson supercomputer defeated two top Jeopardy! contestants a couple years ago, but for the clue "What grasshoppers eat," Watson answered: "kosher." For all of Watson's computing power, for all of its brawny bits (and bytes), it looked pretty silly.

Consider how difficult it is to get a computer to do just about anything. For instance, what if we were physics teachers and as a last task before summer break, we needed to get a class average for the final exam. In order to do that, we might feed the computer the following code:

It's not particularly lengthy code, but the point is... it's code. We can't just give the computer some tests scores and write, "collect the class average." It simply does not understand that. Similar to the above example with Arnold Schwarzenegger and Danny DeVito, the reason why my computer can't comprehend a request to compute a class average is because it is missing a certain human element. We can understand emotions, expressions, and a host of other human signals that a computer cannot because it has no point of reference for them.

What can be done?

Remember, a computer is a machine comprised of hardware and circuitry, it doesn't have a human brain. That being said, we are getting closer and closer to the point where Artifical Intelligence will be able to perfectly mimic human behavior. Whether that's a good thing or bad thing is debatable, but I certainly think there are costs associated with a machine that understands my wants and needs. Sure, it would be awesome if I could give my computer a host of test scores and tell it to give me a class average, but what if it could sense I was angry or being mean to it? Would it erase some of my files? Would it fail to load my contacts on purpose? Of course these are rather extreme examples but I bring them up merely to illustrate the fact that we want computers to be smarter, but how much smarter? And how smart is too smart? While I wait for those answers, you'll have to excuse me, I have some class averages to calculate.