Artificial Intelligence may not be so Intelligent

Computers have, so far, shown no signs of evolving.  No cellphone has initiated a phone call on it’s own to start a conversation with a human.

The science of statistics suggests that a room full of monkeys will someday generate a Shakespeare play.  Something to do with random processes and time.  A computer will never generate a Shakespeare play because there is no program that can embody the process of writing human language based on a set of rules or calculations.  It is hard enough to teach English grammer, let alone to try and teach the creative activity of art through language and have it incorporate real meaning.

A computer, if programmed correctly, could emulate the monkeys and typewriters and through a mathematical model of random character generation, create something.  Given the size of the task, an entire play would require millions of years of monkeys trying, let alone the time required to read each of the attempts.  However, it might be a very worthwhile activity to attempt to program a random character generator that is designed to iterate millions of times and see if it can generate a coherent sentence.

To put in the Shakespearean realm, one could use the phrase “Let slip the dogs of War”.  There would need to be a corresponding program that is checking the output of the random generator for anything resembling human language.  Eventually, some meaningful sentences should occur, again, only by random chance.  And eventually, the random process should successfully generate the desired sentence.

The phrase is a sentence consisting of 23 letters with 27 possible values including the spaces.  Based on simple factorial calculation there are 5.84 x 10 exp 36 possible combinations.  5.84 trillion, trillion, trillion.  Clearly within human scope there is no way to accomplish the task.  Assuming a computer could execute 1,000,000 combinations per second, it would only be able to generate 31 trillion combinations in a year.  Meaning it would still take trillions of years to finish.  It might be possible to speed this process up with some really clever parallel programming, but you get the idea.

This leads to 2 important truths.  First, computers are not smart, they are fast.   Second, the statistical notion that is an infinite amount of time is allowed for something to occur is on it’s face not a proof that something can occur.

Will computers learn and exceed their programming to invent something new?  So far there is no reason to think so.  All of the examples of human-like behavior in computers are simply programming techniques to imitate known human processes.  That the computer is able to execute these processes faster than humans does not make them human.  It just makes them powerful tools for human investigation.

Mechatronic Tips