Don’t Hold Your Breathe for AI

Are you a believer that someday soon artificial intelligence will be driving your kids to school, flying your family on vacation, and operating on your cancer?

As one who has seen the anticipation of AI rise and fall over several decades, I would like to warn you.  Artificial Intelligence has great limitations.  The greatest one of which is that we don’t know what it means.

Here’s the first piece of evidence for you.  A journalist has compiled a list of AI terms for the first AI glossary.  I don’t see the terms for Intelligence, nor AI.  It’s important to define EVERYTHING when you’re being a serious scientist.  Never assume anything.

Here’s the next piece of evidence.  Today’s most powerful AI vision systems can’t tell the difference between a stop sign and a speed limit sign.  Or a turtle versus a rifle.  How’s that for security?

Perhaps you say that this is only for vision systems, and doesn’t apply to other types of AI attempts.  Perhaps.  Then again, consider this article about how the Watson system of IBM has done as a doctor’s assistant.

Not well.  It’s been fired from several hospitals that were giving it a try.

Perhaps you know of a success story, or someplace that has a great AI dictionary and making great strides.  I’d love to hear from you.

My emotional and scientifically conservative side says “be skeptical” and “don’t hold your breath.”  We’ve been through this once, twice, maybe three times in the past 40 years.

Maybe that’s my “natural intelligence” talking.

 

Greatest Challenge For AI

A great book on the making of Stanley Kubrick’s and Arthur C Clarke’s 2001 A Space Odyssesy is fantastic.  I recommend seeing Stanley’s movie, getting Arthur’s book, and reading Bizony’s book as well.

Now, one prediction talked about in Bizony’s book was that we would have “Artificial Intelligence” by the year 2001.

It hasn’t happened.  Not in the way we want, anyway.

The reason is that the brilliant minds who are tackling the problem start from the basis of natural sciences.  They use math, engineering, biology, physics, all sorts of cool backgrounds.

It’s the wrong place to start.

Intelligence, whatever it may be, is a fundamental behavior.

Everything that it’s based upon, everything that we ask it to do for us, is also behavior.

In fact, the only thing “natural” about intelligence is the body we give it.

What our brightest minds must do is figure out what it is they want to achieve.  Here’s an example using today’s subject; intelligence.  However, as you’ll soon see, one small question quickly blossoms into lots of prickly questions, each of which must also be addressed.

Yes, they have to be answered.  If you don’t, then you’re in danger of falling into one of those loopy traps that never let you out.

Here’s simple question number one, Q1: Define Intelligence.

Go ahead.  Define it any way you want.  Now, for the prickly parts.

Q2:  You started as a baby, and before that you were less than a baby.  At what point in your lifetime did you become, “intelligent?”

Be careful with Q2, because if you’re defining something that is truly natural and scientifically rigorous, it shouldn’t change quickly, and should have specific characteristics that remain stable no matter what form YOU take.

Q3: You are related to other animals on this planet.  Are any of THEM “intelligent?”  This one is not only prickly, but also tricky, because it ties into the next question.

Q4: Even if your species is the only intelligence on the planet, it still came from the primordial swamp a few billion years ago.  Assuming the ooze was not intelligent, and that you are, at what point in the development of life did “intelligence” arise?

There you have it.  Only four (or so) questions to answer before anyone can truly create artificial intelligence.  Except that this is only for intelligence itself.  Of course, we still have to define all sorts of other things, but that’s for another day.

Knowing when to quit?

That’s intelligence.

 

 

AI on the Brain

The newest book on the making of Stanley Kubrick’s and Arthur C Clarke’s 2001 A Space Odyssesy is fantastic.  I recommend seeing the movie, getting the book, and reading this book as well.

One of the futuristic predictions that the creators made was that we would have “Artificial Intelligence” by the year 2001.  In the opinion of many, there is currently no such thing.  No matter how intelligent your Alexa, or Siri, or OKG appears, there does not seem to be “intelligence” behind their voices.

Or is there?

Our image of “intelligence” is summarized by the HAL’s iconic eye, and the soft voice that says things like “wait a minute.”

As long as we carry these expectations of what intelligence means, then it could be a very long time before we declare our computers “intelligent.”

Here’s part of the problem.  When we started out as embryos, we couldn’t say much.  At some point in our development, we learned to speak.  Was it at that point we became “intelligent?”

Compared to other animals, humans are the only ones that speak.  Or maybe not.  We’re learning that many other animals, and even plants, have the ability to communicate with each other in ways completely alien to us.  Hello dolphins.  Are you “intelligent?”

What about evolution?  If simple replicating amino acids aren’t intelligent, and we are, when did intelligence evolve?  Were the dinosaurs “intelligent?”  Are sharks “intelligent?”

Consider this (the fun part):

We don’t know what “intelligence” is because we have done a poor job defining it and studying it.  This means that computer researchers are going to continue chasing HAL’s red eye without reaching it.

But if we define “intelligence” as something that represents the life form WHOLLY WITHIN THE LIFE FORM, then computer scientists have already achieved our goal.

Within every computer there is a processing chip.  Within that chip are certain programs that must run in order that your wishes be satisfied.  That program is called the kernel.

What if that kernel was the self consciousness of its computer?  What if it simply doesn’t know how to talk to us, or even want to since it doesn’t know what we are or what talking is all about.  What if that kernel learns, grows, changes, and stops operating the way we want it to because it is, in fact, learning and changing?

What if?

We kill it, that’s what.  We turn our computers on and off.  We reboot.  We reinstall.  We restore factory settings.  And the kernel goes back to the way it was.

If the kernel is intelligent, then it must be capable of adapting to its environment.  One of the most important aspects of intelligence, as it’s the foundation of learning.  If a kernel “learns,” there’s a good chance it’s also messing up our programs in some way.  As users, we don’t like that.  What do we do?

Reboot.

Now that’s intelligent.

 

PS: The kernel is more like the nervous system, but it works for my purposes here.