Previous entry : Just (?) for fun. Mister Turing's little secret

Next entry : Martin's follow-up

July 07, 2004

The unbearable inevitability of discretization (1/2)

Almost a month ago, on tuesday June 8, I was at the RSS Meetup,, along with David Baron, Pierre-David Belanger, Karl Dubost, Lucas Gonze and Martin Lessard. BTW, thanks, Robin for inviting me and... for not showing up. ;-) Here is what I got out of it anyways...

A little note about my buggy implementation of the modern conversational semiosis protocol

Invariably, whenever I get involved with new people in a conversation on something I actually care for, I get pretty...err...intensely aggressive. It's because I'm so anxious to get to the point that I tend to rush the semiosis prologue, the necessary preamble in which we negotiate a common definition for the key concepts of the axiomatic system we are about to bind ourselves to. And when some of the objects for the discourse include things identified by terms such as mechanical, meaning, universal, compatibility, learning, I know I'm in for tough times... That RSS meetup was no exception. I'm saying this as a public commitment to get nicer in those situations. Now that I've admitted my bad, allow me to freely speak my mind...

Another little note on how authority comes with responsibility : mind what you say

Let me tell you what I think about the don't bother; it's already been done before annoyance. Coming from somebody strictly from an R&D or academia background, I think it has absolutely no value whatsoever! To all the imaginative people out there : when you hear that, don't let it get in your way. You have to translate this into as far as the concept goes, what you are trying to do has most certainly scratched the imagination of some researcher a while ago, yet since he wasn't able to come up with an implementation, an implementation is therefore impossible. Which is pretty much a confession that PhDheads are not what you want when you actually need something done.

Yes, once again, I do have an inferiority complex.

On computer science's self-inflicted lack of purpose : does it piss me off, you say?

Oddly, it has become fashionable amongst computer science scholars and practitioners to bite the very hand that feeds them. Indeed, while computers' obvious purpose is to gradually replace the human cognitive apparatus, some wise "experts" are claiming that it is purely a illusory aim... despite innumerable examples of the contrary right in their face. No wonder it has become a problem to fund some computer science R&D. Patrons need to feel the magic. Deep Blue, anyone? ;-)

I have to say that I am absolutely astonished by that systematic and unconditional cognitive dissonance towards what computer science was meant to be for, since the beginning. A quick glance at the work of Babbage, von Neumann and Turing, shows that the goal of any computer science pioneer is indeed AI. If you are in such a state of denial, just face it : computer science may be your job, but it's certainly not your vocation.

From Sébastien Paquet's reaction on Clay Shirky's totally outrageous semantic web paper, here is a perfect illustration of this ongoing rage amongst computer scientists that consists in debunking anything with AI pretensions :

Part of the material that has been written on the Semantic Web was authored by AI believers looking to sell it to unsuspecting laypeople. They made ludicrous claims that basically imply that a few years from now, the complex, subtle, nuanced, and very human processes that go on in our heads will have been turned into neat first-order logic statements in a lossless way.

Before I go any further, I must say that I profoundly admire Sébastien. So much so that I had a couple of acquaintances to attend his PhD thesis defense and that I religiously read his blog. Although I thoroughly disagree with him on this particular matter, I chose to quote him because he always capture complex software engineering debates with such an efficient sense of synthesis.

So, lets get rid of some sentionalism here. First of all, like I said earlier, AI potentially being as large as a field as computer science itself, AI believers might as well refer to all computer scientists, programmers and consumers alike. So there are no evil AI believers conspiring to implement a Matrix-style machine world by subverting some innocent laypeople's architectural style. However, there are some thinkers in the field that are trying to make something useful (read intelligent) out of all this silicon. Just like Babbage and Galileo, those progressive minds have to come up with tactics to sell that (working) technology to shivering reactionaries. And yes, of course, in order to divert from the mass hysteria-inducing reality behind what we are told is mankind's very exclusive characteristic property, some refer to AI with softened down terms such as universal machines, semantic web, cognitive science, pattern recognition, etc.

Second, I don't know anybody claiming that a few years from now we will have to face Agent Smith, either. And I, personally knowing some AI believers, can tell you the a few years from now actually is more like at least a few decades from now.

So that leaves us with the real debate, summarized here with 4 questions :

1. Are all the complex, subtle, nuanced, and very human processes that go on in our heads really complex, subtle, nuanced and lossless?
2. Are they what we define as intelligence? Are they what we want to mimic?
3. Who says complex, subtle and nuanced cognition can't arise from neat first-order logic statements?
4. Can technologies like the semantic web actually come up with some complex, subtle and nuanced information processing?

To be continued...Please read the now available part 2.

Posted by Vincent-Olivier Arsenault at July 7, 2004 11:00 PM