Slashdot

Today I installed a couple important updates to my workstation at work.

They were for French grammar checking in Office apps.

They required reboots.

Yay. WSUS wins again!!!

Gosh sigs are great, arent' they?

I've been toying around with this new one lately:

I would rather read the worst book in the world than read the best web page. . .unless I'm masterbating.

RE: Vulcans and Godel

In light of Godel's Incompleteness Theorem, would there really be a race totally centered on logic? It would seem illogical to do so. . .at the very least it would be incomplete. Logic seems insufficient for living.

Hofstadler and Searle

Speaking of Godel, I think Hofstadler and Searle are saying the same thing regarding strong AI.

Before I give Searle too much credit, I want to say that I think the Chinese room "thought experiment" is a load of crap. However, Searle's summary of the experiment is worth considering:

• Programs are purely syntactical.
• Minds have semantical content.
• Syntax by itself is neither constitutive of nor sufficient for semantics.
• Therfore, programs will never have semantical content.

I buy the summary, though not the thought experiment.

Now, this really seems to stick a fork in the Strong AI that believes that consciousness is mere symbol manipulation. And after a few months of meditation on the subject, I have concluded that Searle is right -- this form of Strong AI has abstracted away "too much from the actual physical implementation of the computational processes involved in thinking."

But what about the so-called "active symbols" that Hofstadtler proposes?

I think that's the key. Essentially, a non-human intelligence would probably be realized in a machine that, on the whole, was NOT a Universal Turing Machine. It's probably made up of a lot of pieces that are UTC's, but on the whole it isn't one.

The sections in Hofstadler's book on the location of meaning are exactly what I'm talking about. I've never made a conscious decision to only analyze the "visible" portion of the E/M spectrum. But because of that very limitation(and sufficient time to "learn" about it), I am now able to associate symbols in my mind with external objects that I "see."

Thanks to Godel, we know that truth is bigger than proof. All Searle has is a proof, though not a proof that Strong AI(in general) is impossible, just impossible on a(in a) UTC.

Vinge's Singularity

The preceeding thought is not so much a slam against Searle as it is an affirmation of Vinge's Singularity.

I can't believe we "re-elected" GW!!!

Artificial Intelligence, and Intelligence in general, are not primarily concerned with "eneumerating truths" or "deciding propositions." If you think about it, in a very practical sense, one's intelligence is not a reflection of how many or how fast one eneumerates or decides the truth of a particular proposition. Especially when intelligence is used in the vernacular sense(or Turing Test sense).

However, there is, after a suitable explanation of Godel's Incompleteness Theorem, a nagging suspicion that the very idea of "Artifical Intelligence" is a contradiction. All programs are explicitly formal, definitely formal enough to fall prey to Godel's construction, so how could they possibly be intelligent with such an obvious and severely limiting property? Isn't it paradoxical to think that there could exist a formal system(program) that understood its own Godelian limitations?

Ultimately, it's no more a paradox than Zeno's paradox(es).

The knowledge that one can or cannot establish truth in the face of infinity is irrelevant IF there is any non-volatile level. Regarding Zeno's paradox and classical mechanics, the non-volatile level is the compact set that is any closed interval of the real line. Regarding AI. . .well that's the content of the ~750 pages of Hofstadter's book.

----
The Location of Meaning
This was one of the more illuminating sections for me. It really sets up a nice argument against intentionality.

The question is: How does one know when any given "message" was "understood" "correctly"? What does it mean for a message to be understood?

Hofstadter outlines a 3 tiered structure for any message: The "outer" message, the "inner" message, and the meaning. The "outer" message serves the purpose of presenting "Hey, here's a message!" The "inner" message serves the purpose of presenting "Hey, here's how you decode me!" The meaning is the actual information contained in the message.

Regarding sending out LP's made of gold into space with the hope that they will be replayed by an alien race, the outer message has the form of a perfectly circular gold disc. A form so odd in nature, ANY intelligence will recognize that it must be more than just a shiny disc. The inner message is the grooves in the LP. Bumps and grooves that spiral from the outside towards the center. . .what to do with them. Again, just about ANY intelligence will recognize the isomorphism between this oddity on the surface of the disc, and the "information" this disc probably holds. Obviously then, the message is the audio stored in the grooves.

I think it's a fairly straight-forward way to explain general communication, especially when looked at from an informatin theory standpoint.

My question for people who believe in intentionality as a necessary condition for mental processes is: at what point did you intend to only decode messages from the visible light spectrum or the audible spectrum? The answer is it is programmed into you via your DNA. I don't believe that any human has existed that made the deliberate choice to only monitor those small parts of the spectrum for messages.

Thus, there exists a non-volatile level, our senses, that guide and limit our own mental activities, regardless of our intentions.

----
misUnderstanding Zeno
Godel's Incompleteness Theorem, once properly understood, is a lot like Zeno's paradox(but on philosophical steroids).

Godel's Incompleteness Theorem applies to all formal systems with sufficient descriptive power. Those systems will either be inconsistent(unlikely by definition) or incomplete. But more importantly, if one adds those "missing" propositions in an effort to complete the system, the system remains incomplete. Much like Zeno's paradox, where one can never take a first step to even begin moving since one can never define a first step because of the infinite divisions.

----
Koans

It seems that there is a very logical reason for teaching Zen through paradoxical koans. In order to fully grasp the paradox, one has to stop thinking in a decidable truth sense, and step outside the system to see the greater truth.

This is analogous to the sensation of looking at most Escher drawings. . .or constructing a Godel numbering. . .or playing an endlessly rising Cannon. . .or naming something without naming it.

----
Godel's Incompleteness Theorem isn't a limit on what will be possible with computers. . .if anything it guarantees that what is possible with computer is at least capable of handling most decidable situations, and whatever emergent behavior comes along as a direct result of the program existing(running?) will obviously supplement those abilities.

I use my Titanium PowerBook mainly for two things:

Digital Performer for recording music and

CLI

Both of which are things you either know how to use or you don't; the quality of the OS doesn't really matter.

So I'm always pleasantly suprised when the computer performs a "consumer" task with the greatest of ease, thanks to the fine people at Apple. iPhoto works great.

Here's what I wanted to do:
I've got 84 raw images from a digital camera that I want to scale down from 1900x1900 or something to 640x480.

Import photos into iPhoto. I already had the images on the hard drive. Importing took about 5 minutes.

Apple-a to select all photos

File-Export from the file menu. Choose scale to 640x480. Choose JPG format. Click export. Select destination folder. Click choose.

2 minutes later, 84 scaled down pictures in the destination folder, ready to be emailed.

Whole operation took less than 10 minutes. I thought it would be longer since I would have to export each image individually. Too easy. Good job, laptop!