Friday, September 30, 2011

While computers get smarter...

Once I got on I-94 in Detroit I didn't need to keep the GPS on, but since it was going to be a long drive home and it was getting late, I thought I'd keep Marianne, my name for the machine, around just to keep me company. After nearly flawless directions, (the only mix-up was when Marianne mistakenly recalculated after a 15 foot detour in Gary), I got to the major east-west street closest to our home in Chicago when Marianne told me to turn right on "TAU-hee" Avenue. I immediately corrected her (OK I was getting a little punchy after a long drive on top of a full day's work): "It's 'TOO-ee' Avenue you dumb machine!" I said.


I returned the car the next day and took my son along so he could see my cool rented car and meet Marianne. "Listen to this," I said to him as we approached Touhy Avenue. "Turn right on 'TOO-ee' Avenue" said the machine. I said to my boy: "wow this really is a smart machine after all, she must have listened to me!"

Marianne was my first experience using a GPS, which is short for a global positioning system, a machine that is able to pinpoint your exact location then determine the best route to your destination. It's a marvelous bit of technology but I've always felt confident enough in my own navigational skills that I've never been tempted to get one. Once I was picked up at the Phoenix airport by a friend of my parents who insisted on using her GPS, even though I knew exactly how get to where we were going. She insisted on listening to the machine instead, which led us on a wild goose chase and got us hopelessly lost. But that was ten years ago and the machines are much better now. When presented with the GPS option as I rented the car a couple of weeks ago, I thought I'd give it a whirl, besides the company would be paying for it. Like any relationship there were some bumpy moments mostly due to my own inexperience. But pretty soon we were in the groove so to speak and as they say, it was the beginning of a beautiful friendship.

In the Detroit area, my work was about 15 minutes from the motel and Marianne got me between the two via a dependable, if somewhat convoluted route as I would later discover. But after making the same trip three days in a row, I still depended on Marianne getting me there, I never committed the route to memory. What's more, I never got to know the territory I was covering. "Did you see such and such?" I was asked by a colleague who had been there before. I hadn't because, as far as I was concerned, following Marianne's directions to my destination was my only mission. Had I used a road map, I would have had to figure out the route myself, which as I have learned from experience, would have imprinted itself on my brain much quicker than simply following Marianne's directions. Plus a map would have shown me other features of the neighborhood I was driving through, places my curiosity would have taken me.

Which got me thinking about what we lose when we let our machines do the thinking for us. To put it more bluntly, as computers get smarter, are we getting dumber? Some might argue that computers are now employed to do many of our routine mental chores, freeing up our brains to think about more important things like our fantasy football roster. Others would argue that the very process of figuring out simple things like how to get where we are going, go a long way toward keeping our brain synapses active, healthy, and alive.

Let me say at this point that I love computers. To me they are amazing creations of the human spirit, limited only by the creativity and intelligence of the people who control them. I have done some programming of my own as a hobby, and have come to greatly respect the work of software designers. I thoroughly understand their drive to make computers perform tasks that people only a generation ago could not have dreamed. The holy grail of software design is to create intelligent computers, that is machines that can learn, correct themselves, and create and teach other computers, in short, computers that can do essentially what we do, only better. No one quite agrees whether we're there yet, computers can do truly remarkable things, but they don't do them the same way humans do and because of that, can't be truly called intelligent.

Take the computer that beat World Chess Champion Gary Kasporov. There have been chess playing computers for a long time but until fairly recently, none of them have been able to beat the best human players. It's not really intelligence that computers use to play chess, it's shear processing muscle. In chess a good player must look ahead several moves on both sides of the board to determine a successful strategy. The better the player, the farther ahead one can look. But even with the limited number of pieces on a chessboard, there are millions of permutations of moves, far more than even a grand master can possibly see. That's not a problem for a computer, even the old beat up one one I'm typing on now can perform about a billion calculations per second, give or take a few hundred million. A computer can take those possible moves, play them all out to their completion, and determine which will yield the best results, in far less time than it took me to type all that. Still it took a massively powerful computer to beat Kasparov, one that was was devoted entirely to the game of chess. In other words, you or I could have beaten it in a game of tic tac toe.

For his part, Kasparov was a bit of a spoiled sport when he lost to Big Blue the IBM computer in 1997. He claimed that it cheated and demanded a rematch, which never happened. To me that seems a little like a world class runner being disappointed at being beaten in a race by Michael Schumacher driving his Formula One race car.

Given that, one would think it would be child's play for a software engineer to design a program to play the game of Jeopardy. Not so. Jeopardy is not merely a game of retrieving information, which computers can do in their sleep. It's a game about human language, which computers are terrible at. Again it was the scientists at IBM who took on the challenge to build a computer to take on the greatest of the Jeopardy champions. Their challenge was to create a machine that could understand the clues which often contain slang words, cheeky puns and double entendre . Some Jeopardy clues are painfully easy for a computer. Take this one:

"He was the pope in 1324."

I googled exactly that and Google's first response was Pope John XXII, AKA Jacques Duèse. Pretty damn good.

Here's another clue, one that actually appeared on the show:

"The original head shrinker, he fled to London in 1938 following the Nazi annexation of Austria."

Try googling that and you'll get everything from Rudolph Hess to the Dalai Lama, but not the right answer. For a human player, that's not a terribly difficult clue. "Head shrinker" could either be a person who makes shrunken heads, or a slang term for a psychologist. We quickly dispose of the former since A) not too many makers of shrunken heads are household names likely to be answers on Jeopardy and B) in the clue we have 1938, and Austria, neither a time nor place known for shrunken head makers. So by process of elimination, this term must refer to a psychologist, and if that is correct, an obvious shot in the dark would be Sigmund Freud, which would be the correct answer. The other words in the clue merely confirm the solution. Rather than using this simple string of human logic, a computer analyzes each word in the clue and ranks them in order of importance. Judging from the Google results to my query, I'd say the program deemed Nazi to be the most significant word followed by Austria, then London. The results Google gave me contained all three words, but nothing having to do with a head shrinker, although some results included the word head and others the word shrinker. It's the subtleties and nuances of language that computers can't deal with very well. That is until the IBM guys got to work on the project. After another single use monster computer, a few years of work by some of their most brilliant minds, and tens of millions spent on the project, the folks at IBM did indeed build a computer that beat the best human Jeopardy players.

Now you may be wondering why all the effort to build a machine that can win a TV quiz show. Well its the same reason we learn subjects at school that seem at first to be pointless, like algebra. It's all about learning how to solve problems. While to this day I still can't factor a quadratic equation to save my life, much of what I learned in high school algebra has stayed with me and at least abstractly I use it as a basis for solving daily problems. Solving the Jeopardy problem gave programmers a clear insight on how to create software that comes close to dealing with the complexities of human language, that will help them build systems with far reaching applications. Some of these future applications no doubt will benefit us greatly. Of course as is the case with all great innovations, some applications will cause us harm. In the end, computers are simply tools. As a tool, a hammer can be used to build a house, it can also used to bash someone over the head. It's the same with computers, it's up to us to know how to use them wisely.

There is an ongoing debate in education circles on the use of computers in the classroom. Movements are afoot to provide money to underprivileged schools for the purchase of computers. This is seen as progressive, after all computers, the theory goes, are such an important part of society that in order to make it, let alone prosper, an individual needs to be computer savvy. If learning computer skills is to be part of the curriculum, something has to be sacrificed, as there are limited hours in a school day, especially here in Chicago. In the case of my son's school, as always, art, music, and the rest of humanities, which so many seem to think are the most dispensable, are the first to go. The old fashioned skill of penmanship seems to be a thing of the past, I guess what's the point of handwriting if you can type on your laptop? Arithmetic and spelling are still taught but one can only wonder for how much longer, after all we have calculators and spell check to do that stuff for us. The saddest of all to me, is that libraries are being turned into media centers and what was once space for books, is now space for computers. We removed our son from his first school because the room that was once a library became a broom closet. A seventh grade class project shown at an open house that publicly bragged that all the research was done on Wikipedia sealed the deal.

But like it or not, computers are here to stay. I believe there is a place for them in our children's schools but I don't think they should take the place of books or other traditional subjects that are also necessary elements of life. I just challenged my son to look up the state motto of Hawaii, in Hawaiian (please don't ask me why). He came up with the answer on the computer in less than a minute. That's great. But we need to find more challenging tasks for our children, tasks that teach the elements of problem solving. If computers are going to be an integral part of the curriculum, I would suggest including basic computer programming for children, that is, start with a blank screen, teach the kids a simple computer language, one perhaps designed for such a purpose, and let the kids have a go at creating simple programs, then see what happens. They will fail as everybody does at first. But gradually they'll learn to understand what makes a computer tick, and learn a very valuable lesson in logic and problem solving by being forced to communicate with a device that "thinks" much differently than they do. Most importantly, they will learn that a computer, a tool they have known all their lives, is not simply a magic box where you type in a question, and it delivers an answer.

Any innovation's greatness is measured in terms of how much it benefits vs. how much it hinders society. The automobile gave people at all ends of the economic spectrum unimaginable independence. On the flip side driving has become the primary means of transportation for most of us at the expense of walking and other healthy activities. We are consequently told that we need to exercise so many hours per week to make up the difference. By letting them do the thinking and not challenging us, computers and software, especially dependable ones that are completely intuitive and user friendly I'm afraid are doing the same thing to our minds that automobiles did for our bodies.

Computers may or not be intelligent, but the people who design them certainly are. Maybe we can tap into those minds if just a little to learn how to keep our own minds active and healthy,

And whatever we do, please let's not throw away those road maps.

Just an idea. But what do I know? Thanks to the computer, I just write a blog and as everyone knows, any idiot be a blogger.

No comments: