Ah, I see the confusion now: you're conflating NP-completeness and non-computability. They aren't the same thing. For example, the subgraph isomorphism problem is known to be NP-complete, but that doesn't keep Google, Orbitz, etc. from using it or math equivalent to it. Simple descriptions of actually non-computable functions even in principle (vs. non-computable as a practical matter) can be found here.

After you guys figure out what "information" means, your next assignment is to figure out what "consciousness" means. Thanks in advance for helping out on that. :-)

Amen there.

Nobody knows what consciousness means. Not only are we still unclear of everything that happens, we still haven't come to an accepted meaning of what it is! 

Yeah...that's what raised my eyebrow when Steve Meyer said "information suggests intelligence, a conscious being." That's why I got amused by the involved back and forth on this posting just about the word "information". I can't imagine how scientists and ID adherents can agree on "consciousness". It would seem that consciousness would have to include dreams and hallucinations at some level. Does the Intelligent Designer hallucinate? If so, that actually may explain the Copenhagen Interpretation. :-)

It isn't about the word "information." At least, that's not what I'm discussing. I'm discussing Information Theory, a scientific—in the narrow Popperian sense of making falsifiable claims—theory the results of which you rely on every time you use Ricochet. In particular, I'm talking about Algorithmic Information Theory and its role in epistemology, since the launching-off point for the discussion was the claim that a "new information theory" is needed, a claim that can only be tested against current Information Theory and, when so tested is, in my opinion, found wanting.

I haven't heard the speeches or read the papers at the conference, so it's possible they would have argued something that would persuade me otherwise.

• I believe in God, and 
• I hold that God created the regularity and complexity found in nature.
• However, I also hold that there are rules to what counts as compelling proof, and the mere fact of complexity doesn't compel anyone to believe that it must only have come from God. 

I'd happily agree that information suggests a consciousness, but "suggest" doesn't mean "prove." 

Before anyone thinks we're off into intellectual masturbation, let me bring it down to earth. In my day to day life, I'm a database developer, and I specialize in data analysis. It goes by the name of OLAP, Online Analytical Processing. It includes simple querying all the way up to data mining. In other words, I get paid to discover what data predicts. 

I used to merely study the theory of knowledge. Now I get paid for using it.

I don't think I'm fetching far to claim that these questions will become more important in the future. Look around. As a society, we're collecting huge amounts of data. The vast majority of it is unused, un-analyzed, but it's also just begging to be exploited. The discipline that's required to analyze it is still in its infancy. 

Databases aren't like DNA, where perfect conclusions can be drawn. What counts as a legitimate conclusion is going to be crucial. 

I can imagine how: the scientists would have to give up their unfounded philosophical aversion to the notion of teleology and the ID folks would have to quit using the notion of teleology as shorthand for God.

Claire, I'm a bit surprised at your comment about the participants at the conference. It's one thing to claim that The Origin of Consciousness in the Breakdown of the Bicameral Mind, Consciousness Explained, How the Mind Works, or How to Build a Person: A Prolegomenon (to name just a few of the titles on my shelves) aren't sufficiently persuasive. It's quite another to dismiss the entirety of work on the subject by saying we can't even begin to define the term "consciousness." We can, and have, begun.

Interesting! I believe in God also, but I don't agree that "information" suggests a "consciousness." "Information" is just lack of entropy, and contrary to the popular misunderstanding of the 2nd law of thermodynamics, entropy-reducing events do occur without a consciousness to bring them about.

That's not to say that homo sapiens sapiens isn't a truly awe-inspiring entropy-reduction system, though, nor that it isn't true that God wants it that way. In my belief, the Jewish observation that humanity consists of co-creators (that is, providers of order/information) with God is among the foundationally true religious insights in human history.

You might mean something by "databases" other than the traditional SQL databases, because SQL databases are actually based on a (flawed interpretation of) relational algebra, where indeed "perfect conclusions" (that is, sound logical inferences) can be drawn. This admittedly takes some discipline with SQL, as documented in SQL and Relational Theory. A lot of data mining in this context is logical because of that.

DNA, on the other hand, admits no such straightforward logic. Work with DNA today is overwhelmingly statistical in nature. Uncertainty is rife, and the whole toolbox of dealing with uncertainty is brought to bear on it.

With that said, much of the reason I run around talking about [Jaynes 2003] and MDL is precisely because they bridge the artificial logic/probability gap. If you're doing machine learning and haven't checked them out, I highly recommend them!

I agree that we've begun to define what consciousness could be but the point being that getting consensus on what it is may prove as challenging as understanding or getting consensus on various theories of quantum mechanics. The inquiry will obviously lead down a myriad of paths and one still then has to ascribe this definition to an Intelligent Designer.

I am pleased to know that you have Julian Jaynes' book on your shelf. One of the more enjoyable and lucid inquiries into the subject. 

I appreciate this. Data analysis requires some degree of familiarity with different algorithms and models, but that doesn't necessarily satisfy my philosophical curiosity. And whereas I don't think that Information Theory and the theory of Knowledge are exactly the same, they've certainly got plenty of parallels, and the cross-pollination serves both.

Thanks! 

Gentlemen and Fair Lady,

Of course I'm confused, as is everybody who has actually worked in this field. Perhaps I didn't word my discussion of np-complete clearly. I meant to say that I strongly suspect that the problem of defining information is of the order np-complete.

As to consciousness, Penrose has spent many years and written three books examining the problem. His conclusion is that no algorithmic machine can be conscious.

I suggest you read up on him.

http://en.wikipedia.org/wiki/Roger_Penrose

Thank you for your kind attention.

Roy

I'm so glad to hear that your curiosity is piqued! I completely agree that information and knowledge aren't the same thing. For example, Algorithmic Information Theory tells me all information in a deductive logical system is already present in its axioms. Yet I am frequently surprised, and feel like I gain new knowledge, from some of the theorems derived from the axioms. I find that observation significant.

With respect to knowledge specifically, there is (a) modal logic that treats it, and the foundations of it are documented well in Reasoning About Knowledge. Also highly recommended.

I guess I'm still confused: is your claim that the definition of "information" provided by Information Theory is somehow ineluctably complex in spite of the successful application of it? Is your claim that Algorithmic Probability is non-computable, and this somehow invalidates our understanding of "information?"

What's worrisome to me is an ongoing set of references to the word "information" without any links to, let alone critiques (a noun, pace Claire) of, the well-established, extraordinarily successful field of Information Theory. In other words, some people seem to me to be writing as if what "information" is is a complete mystery. It isn't. It's an extremely well-defined physical concept.

I can assure you I'm familiar with Dr. Penrose's work. :-)

Dr. Penrose's assertions regarding the inability of algorithmic machines to be conscious are closely connected to his belief that the human brain operates by principles not adequately explained by our best current understanding of quantum mechanics. They're fascinating speculations, to be sure. But speculations are exactly what they are: they enjoy no experimental support whatsoever, and in fact his arguments, when divorced from his speculations in physics, are isomorphic to the Lucasian anti-AI arguments, and are (regrettably) dismissable just as trivially.

I was with you, right up until modal logic. I agree with Quine - - modal logic is a scam. Modal logic confuses the notion of necessity/probability - or any other mode (good/bad, sometimes/always) with the reality of the observed object. One is a statement about the observer's subjective knowledge, the other is a statement about the object. You cannot mix them. The one has no bearing on the other.

 They both affect the willingness to believe, but that effect happens independently of each other.

The task in data analysis, information theory, and knowledge is all the same: How do we prudently manage the relationship between what is observed to what can be known?

A new can of worms is open! ...

Gentlemen,

I am thoroughly familiar with the discussions on logic and consciousness of all of the referenced works. The fact of the matter is that there are many gaps in the theory. For example, recognizing and interpreting natural language is proving to be far more difficult than many had anticipated.

A few years back I developed a system for The New York Hospital - Cornell Medical School anatomical pathology department, one of the largest in the world. The aim was to translate the natural language of pathologist's reports into SNOMED (Systematized Nomenclature of Medicine -- Clinical Terms). Even given a very restricted vocabulary we were plagued with ambiguities, The solution was to return to the pathologist a report that listed as many as 8 ambiguities and select what they really meant.

Again, I rely on Penrose. He has received numerous prestigious prizes for his works on General Relativity, Quantum Theory, Mathematics, Multi-dimensional Geometry and other fields. He has the office next to Stephen Hawking. He is credited as co-author on many of Hawking's works.

Roy

KC: To clarify, I wasn't recommending modal logic as the definitive answer to the question "What is knowledge, vs. information?" I only wanted to indicate that the subject is under study and offer a touchstone—one that I see you're familiar with. :-) You're probably also aware, then, that the current logical effort in knowledge goes under the rubric of Justification Logic.

Roy: I don't recall anyone here claiming that computational linguistics is easy. Nor do I see how the observation that computational linguistics is hard warrants relying on Roger Penrose via argument from authority. When Dr. Penrose has experimental support for this theories, they'll be worth paying attention to for more than their intellectual derring-do—which, I hasten to point out again, is considerable.