The problem with intelligence "arising" from mechanical operations
In the comments on this post, rob argues that a bunch or "circuits" (or neurons, I guess) behaving according to deterministic, mechanical laws is exactly what "gives rise" to intelligence, in humans or computers.
The problem with this view is Occam's razor. Let us consider a door lock. If the lock is set, one can't open the door without a key (at least without breaking it). We can see why this is so on simple, mechanical principles. Now, it just could be that the door "knows" when it is supposed to let people in who don't have the key, and when it shouldn't. But generally we reject any such hypothesis as superfluous: once we understand how the door mechanically does its job, we simply don't need to posit any "knowing": it won't "do any work" in our explanation of when we can get in the house and when we can't.
Now let's say we add some biometric feature to the door: the owner can still get in by fingerprint if he forgot his key. Would we say the door "knows" it's the owner? Again, not likely: we'd likely say we'd just added a new mechanism, i.e., we now have a more sophisticated machine. The positing of an "intelligent door" still fails the test of Occam's razor: having explained the operation of the door on strictly mechanical grounds, there is just no need to add in a dollop of intelligence that has no effect whatsoever on what happens.
And so we have a proof by induction: if a machine (even a biological machine) with n mechanisms can be fully explained mechanically, and we add the n + 1th mechanism that is also understandable on purely mechanical grounds, then the new machine is also fully explained mechanically. No piling on of circuits upon circuits ever justifies the adding in of a new element, "intelligence," that is something non-mechanical. If it is true that even human beings are "just circuits," behaving mechanistically, than "the human mind" is just a ghost, and you don't really "decide" if an idea is true or not: some circuits simply did what they do accoding to mechanical laws, and the outcome spit out of your mouth, with no room for whether you "think" a proposition is "true" to have any influence at all on what you say.
And that is something recognized by honest materialists such as Alex Rosenberg, who claim that ideas like "belief" are without meaning, and that consciousness is an illusion. (Although who is suffering from this illusion is not clear.)
Now, if we want to re-define intelligence as "a really slick mechanical contraption," then OK, but surely then Windows NT or Photoshop qualifies as "intelligent." But most AI enthusiasts want to have their cake and eat it too: they will, when pressed, admit that what goes on in any computer is entirely mechanical, but they still want to claim that, say, Watson, "has now achieved real intelligence!"
This is pure mystification.
The problem with this view is Occam's razor. Let us consider a door lock. If the lock is set, one can't open the door without a key (at least without breaking it). We can see why this is so on simple, mechanical principles. Now, it just could be that the door "knows" when it is supposed to let people in who don't have the key, and when it shouldn't. But generally we reject any such hypothesis as superfluous: once we understand how the door mechanically does its job, we simply don't need to posit any "knowing": it won't "do any work" in our explanation of when we can get in the house and when we can't.
Now let's say we add some biometric feature to the door: the owner can still get in by fingerprint if he forgot his key. Would we say the door "knows" it's the owner? Again, not likely: we'd likely say we'd just added a new mechanism, i.e., we now have a more sophisticated machine. The positing of an "intelligent door" still fails the test of Occam's razor: having explained the operation of the door on strictly mechanical grounds, there is just no need to add in a dollop of intelligence that has no effect whatsoever on what happens.
And so we have a proof by induction: if a machine (even a biological machine) with n mechanisms can be fully explained mechanically, and we add the n + 1th mechanism that is also understandable on purely mechanical grounds, then the new machine is also fully explained mechanically. No piling on of circuits upon circuits ever justifies the adding in of a new element, "intelligence," that is something non-mechanical. If it is true that even human beings are "just circuits," behaving mechanistically, than "the human mind" is just a ghost, and you don't really "decide" if an idea is true or not: some circuits simply did what they do accoding to mechanical laws, and the outcome spit out of your mouth, with no room for whether you "think" a proposition is "true" to have any influence at all on what you say.
And that is something recognized by honest materialists such as Alex Rosenberg, who claim that ideas like "belief" are without meaning, and that consciousness is an illusion. (Although who is suffering from this illusion is not clear.)
Now, if we want to re-define intelligence as "a really slick mechanical contraption," then OK, but surely then Windows NT or Photoshop qualifies as "intelligent." But most AI enthusiasts want to have their cake and eat it too: they will, when pressed, admit that what goes on in any computer is entirely mechanical, but they still want to claim that, say, Watson, "has now achieved real intelligence!"
This is pure mystification.
I thought this was a really good post.
ReplyDeleteI agree with your statement.
'If it is true that even human beings are "just circuits," behaving mechanistically, than "the human mind" is just a ghost, and you don't really "decide" if an idea is true or not: some circuits simply did what they do accoding to mechanical laws, and the outcome spit out of your mouth, with no room for whether you "think" a proposition is "true" to have any influence at all on what you say'
(Or perhaps my circuits just cause ,me to write that I agree with this statement).
A world where 'human beings are "just circuits" behaving mechanistically' would indeed be an empty world with no genuine free will - but that doesn't mean it isn't true and that future scientific investigation won't provide further validation. It is certainly a logical consistent view that is contradicted by no evidence I am aware of.
However I don't quite see why intelligence and consciousness could not exists in such a world, even if free will did not. I can accept that I may be just a mechanical device devoid of free will , but that does not mean I have to reject the idea that I have genuine consciousness of this bleak situation.
As you say - intelligence is just a word and there is no reason why it can not be given a meaning that makes sense in a purely deterministic world and it seems somewhat arbitrary if Windows is considered intelligent or the definition sets the bar higher.
While my own intelligence (that I appear to have consciousness of) tells me that it would be cool if the world I inhabit is non-deterministic in a way that was consistent with genuine free will it also struggles with this view from a philosophical and scientific perspective,