Artificial Intelligence

I'm about a fourth of the way through The Singularity is Near and the major thing I still have doubts about is the future of artificial intelligence.

What if we can emulate data processing and memory, but never get to that point where a machine can learn in a situation beyond the parameters programmers set for it. Maybe after more reading, I'll be convinced.

This might not matter, however, if we're able to interface the human brain with computers. At that point there might just be an expansion of our own thinking power, probably allowing us to overcome any obstacles that stand before a sentient computer.

I think we have a pretty crazy future ahead of us - if we don't blow ourselves up first.

Re: Artificial Intelligence

“This might not matter, however, if we're able to interface the human brain with computers. At that point there might just be an expansion of our own thinking power, probably allowing us to overcome any obstacles that stand before a sentient computer.”

actually, i'm more convinced of this route than artificial intelligence. why create an artificial intelligence when we can expand the intelligence that is already here? but who knows. that's why it's called the singularity after all.

Re: Artificial Intelligence

here's what Stephen Hawking has to say about this. i still don't know but like i said, i haven't closed any doors of perception

“Some people say that computers can never show true intelligence, whatever that may be. But it seems to me that if very complicated chemical molecules can operate in humans to make them intelligent, then equally complicated electronic circuits can also make computers act in an intelligent way. And if they are intelligent, they can presumably design computers that have even greater complexity and intelligence.”

– Stephen Hawking via Singularity Summit at Stanford

Re: Artificial Intelligence

This is one of the key points of the “big AI” groups - that our minds are just incredibly complicated computers.

There's a counterargument that poses the question: “If supercomputers contain as much processing power as our brain does, why aren't they intelligent?”  They believe that there's some kind of inperceivable item at work in our brains - a scientific version of the soul.

Check out Rodney Brooks' Living Machines lab for more.

 http://www.ai.mit.edu/projects/living-machines/

Re: Artificial Intelligence

As long as humans control the dialogue, we will define what machines do as “artificial” intelligence, or a “simulation” of intelligence, or “only doing what they're programmed to do”, or some other term that devalues machines.

We do this because machines have no souls, and therefore no afterlife, and we want desperately to believe that we are different.

 The truth is that a $4 calculator exhibits human-level intelligence.  If you don't believe this, ask a dolphin or a chimp to give you the square root of 7.

Re: Artificial Intelligence

I don't think you can frame the argument around calculation - there are hundreds and thousands of things we can do that computers cannot (or do badly) -

Pattern recognition

Adapting to unforeseen circumstances

Manipulating our environment in ways to help us

and many many more.

Many creatures on this Earth display both the storage capacity of a computer but with a flexibility one can only find in “minded” creatures.  This is the real goal - as soon as a computer can speak intelligently about a variety of topics that interest it, as well as have the free will and intuition to go out and make it happen, they are really just tools, not sovereign entities. 

Re: Artificial Intelligence

There are many things that computers can't do,YET.  No question about that.  But you're doing the same human-chauvinist thing so many of us do, by dismissing what computers do as “just” calculation, and then listing the things that computers can't do.

I'm old enough to have learned how to do square roots in grade school, like long division only more complicated.   When I calculated square roots, was I exhibiting intelligent behavior?  Or was I “just” calculating?  Was I “just doing what I was programmed (taught) to do”? Did my parents reward my effort by reciting a long list of all the things I couldn't do?

No, when a child calculates a square root with pencil and paper, we call that intelligent.  When a machine calculates a square root, we devalue the behavior.

Re: Artificial Intelligence

There's a difference between intelligence and training.

 Intelligence is flexible in the face of changing conditions - if an equation was missing a symbol or someone skipped a step, an intelligent device can still solve the problem.  Computers aren't there yet.  Flexibility is the key component.

Re: Artificial Intelligence

I could tell you some tech support stories about real, live human beings who had been trained to use computers but didn't have the flexibility to fill in the missing steps or deal with changing conditions.

I suppose we could talk about the difference between intelligence and training and conclude that these people were not intelligent.  But that would be nonsense.  These were regular people, of more or less normal intelligence.  And if following instructions by rote without understanding them was the limit of their abilities, that just means that following instructions by rote IS human-level intelligence, whether it's done by humans or machines.

 Believe me, if chimps could do the work, their employers would have hired chimps.

Re: Artificial Intelligence

That's a limited subset of the human experience.

The intelligence comes in when we stop doing the “computer” task and seamlessly transition to a myriad of other tasks.  We use the same systems in new and novel ways.  I suppose the true argument comes to “what is the definition of intelligence”. 

Re: Artificial Intelligence

Of course it's a limited subset of the human experience.  My own behavior is a limited subset of the human experience.  So is yours.

 It does come down to a definition of intelligence.  My original point was that humans will try to define intelligence in such a way as to exclude machines.  You, being human, have done exactly that.

I do think it is possible to find objective measures of the complexity of behavior, for example via Algorithmic Information Theory (Chaitin).  And certainly more intelligent entities exhibit more complex behavior.  The converse also holds; we ascribe greater intelligence to people who exhibit more complex behavior.  

The obvious conclusion is left as an exercise for the reader.

Re: Artificial Intelligence

Well, if the brain has a bunch of analog tools that help generate human intelligence, a computer could be vastly bigger, and process vastly faster, and still not duplicate human intelligence.

If we understood those analog tools are recreated them for the supercomputers, that might be a solution to the ai problem.

Re: Artificial Intelligence

That's a very good point.  The architectures are fundamentally different, so machines have certainly not duplicated human intelligence.  It's a matter of degree; my own intelligence is not a duplicate of my teachers' intelligence either.

My view is that machine intelligence is more or less human-level intelligence, certainly compared to chimp-level intelligence, but it is a non-human intelligence, almost an alien intelligence, something that we are psychologically inclined not even to recognize as intelligence.

In principle, if we understood the analog circuits of the brain, we could either simulate them digitally in software or implement them directly in hardware with analog electronics.  I tend to think that human intelligence and machine intelligence will continue to diverge.  We could build calculators that are as slow and error-prone as humans, but we don't.