Wednesday, December 16, 2015

The Nature of Intelligence

The question of the nature of Intelligence/sentience presents intellectual traps that are difficult to avoid; not least the problem of a nested regress!

This post on Uncommon Descent is a fine example of how, on the  Intelligent Design question, I'm probing an entirely different line of inquiry to the de facto IDists. The post is by Denyse O'leary who quotes one of the ID community's gurus, Robert Marks.

Firstly, setting the scene:

Anything algorithmic can be done by a computer. Give me a recipe for doing something, and I can whip it up in the kitchen. There are things which are not algorithmic the most celebrated of which is Turing’s halting problem: there exists no algorithm able tell whether or not a computer program runs forever or halts. (The halting algorithm must work for any and all computer programs.)

But a computer program will halt or won’t halt. But since there is no algorithm to figure this out, the halting problem is undecidable. We don’t know before running the program whether or not it will halt. It could run trillions of years and then halt long after we’re dead. If it doesn’t halt, we may never know (unless we know the so-called busy beaver numbers which is the same as knowing Chaitin’s number which is unknowable. But I digress.)

Clearly there are some elementary algorithms where we can prove they halt. However, the tenor of the halting theorem is that there is no general algorithmic procedure which takes any algorithm as a parameter and is then able to determine whether it stops. That is, we may be able to determine halting conditions in special cases, but not in the general case.  The halting theorem can be proved by showing  that when an attempt is made to submit a hypothetical general halting detection algorithm to itself as a parameter this results in a contradiction. Hence, some questions we submit to algorithms are incomputable. Incomputablilty has the potential to arise whenever an algorithm attempts to flag conditions about itself: Like the well known contradictions arising from Russell's paradox it is not possible for an algorithm in general to talk about itself without raising contradictions. 

But I digress. I'm actually more interested in the following quotes from Marks, quotes which bring to light something I've long suspected would be a position favoured by a de facto IDist like Marks. Basically it's another Intelligence-of-the-Gaps sentiment:

Lastly, Roger Penrose in Shadows of the Mind and The Emperor’s New Mind makes the case the human mind, through creativity and the creation of information, does nonalgorithmic things (and is therefore not merely a computer).

I am starting to believe creation of information requires a nonalgorithmic process, hence intelligent design.

This is not unexpected: As I have made clear many times on this blog, IDists like Marks are dualists who see a sharp distinction between "intelligent agency" and "natural forces". This dualism is embodied in the IDist's  explanatory filter; This filter ensures that when "natural processes" fail as an explanatory device a default is forced to "intelligent agent".  So it is no surprise to find Marks casting around for reasons why intelligence should be classed as an entirely different genus to "natural processes" and fundamentally different.  Marks thinks information cannot be created by "natural processes". This is an error in itself which I will be publishing on shortly. However, if you believe "natural forces" can't create information it is a very easy next step to posit that some kind of mystical unknowable process must be creating information and of course that process can only be the apparent inscrutability of intelligence! Wouldn't it be right then if intelligence fits in the category of the non-algorithmic?

I had anticipated long ago that IDists like Marks would settle on Penrose's proposal that intelligence is non-algorithmic. My own opinion is that this proposal is unlikely and I give my reasons for this here and here. All evidence suggests to me that the human mind is finite and therefore the ontology of human intelligence has the same reflexive limitations that give rise to the halting theorem and incomputability in general: Viz: human intelligence is based in a system that can not make certain general statements about itself without those statements invalidating the very conditions these statements are attempting to comment on; there are certain things we cannot know about ourselves. Ergo, human intelligence doesn't step outside the limitations of incomputability.

From my perspective I feel that the de facto IDists are more than welcome to explore this dualist line of inquiry whereby intelligence is categorized as a different genus of process capable of exploring the realm of incomputability; I'm not trying to stop them and I'm happy that they follow this very different line of inquiry; although if they are committed to this view of intelligence they will find progress difficult.

But the polarized dualist backdrop against which de facto ID plays out doesn't favour the intellectual nuancing needed to explain why some people follow one route and some another. Much more in line with de facto IDism's embattled community is the hunting out of fifth columnists and traitors and then hanging them out to dry; as we will see in my next post!

No comments: