Are We too Harsh on AI?

The search for true AI has been a long, slow, and bumpy one, and the more years pass it seems the slower progress has become. And yet, maybe we are being too harsh on our computers.  We expect our AIs to know everything, to do everything, perfectly, as soon as they are switched on.  Isn’t that a little unfair?  More importantly, isn’t it a little misguided?  In our search for the “perfect solution”, are we actively discouraging approaches that could lead us to machines that are truly intelligent?  Humans, after all, are neither perfect nor omniscient, nor are we born knowing everything.  In fact we are born knowing very little: most of the qualities of human intelligence that make it a desirable thing to replicate–such as language, creativity, and perception–are things we have to learn how to do, not things we are born with.  There is even evidence that such fundamental aspects of perception as object permanence and visual closure must be learned!

Babies aren’t born knowing how to talk.  Infants aren’t even born knowing how to focus their eyes.  We have to learn how to walk, for goodness’ sake!  And it takes us years.  Isn’t it asking a little too much of computers to expect them to be able to perform, a priori, tasks it takes us years to perfect?  Take, for instance, the seemingly simple task of recognizing a car from different angles.  Not an easy task for a computer, but for us it is.  Or at least, it seems simple to us; we do it all the time, without thinking about it.  Like walking, we’ve practiced it so often we don’t even have to spend conscious thought on the process.  Yet isn’t that very practice our biggest advantage?  How many times in your life have you seen a car?  Ten thousand?  A hundred thousand?  A million?  How many times have you heard, read, or spoken a sentence?  A million times?  Ten million?  A hundred million?  How many times in your life have you seen a human face, or used your hands to pick up and feel an object, or listened to music?  The number of experiences a human accumulates in a single year puts most so-called computer “learning” algorithms to shame.

Not only have you been exposed to these elements of our world countless times, you’ve received important feedback every single time.  Computers have some obvious advantages over humans (speed and precision, primarily), but we often forget the advantages that humans have.  A computer does not have parents or teachers, it does not have a body it can use to explore and manipulate its surroundings (at least, none anywhere near as capable and sensitive as our own), and as fast as its processing power may be it can not hold a conversation any faster than a human can, as long as humans remain the only things that know how to talk.  Human children, on the other hand, have the advantage of decades of careful, focused training and constant feedback from their environment, their peers, and the adults around them who have already mastered the skills they need to learn.

Intelligence, it turns out, does not mean knowing how to do everything.  Rather, true intelligence seems to be about learning how to do things.  It’s a slow process, but a flexible one–much more flexible than the artificial “intelligence” we usually expect from our computers.  A computer, for instance, can be programmed to play perfect Chess in a matter of hours, assuming an expert is doing the programming.  But if you want that same computer to learn how to play Go, not only do you have to change the software completely, you probably have to get someone else to do the programming!  A human, on the other hand, takes years to master a game like Chess–but that same human can just as easily learn to play Go, or Checkers, or Backgammon, or Capture the Flag, or Let’s Pretend, or Super Mario Brothers–and they don’t have to get brain surgery to do it.  If they are in any sense “reprogrammed” when they learn these skills, it is the program itself that is doing the programming!  This is the key to real intelligence, and until we master it, AI is doomed.[1]

[1] For a possible alternative approach to the problem of machine intelligence, I recommend the excellent On Intelligence by Jeff Hawkins.

Leave a comment

Filed under Science

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s