The Hidden Flaw in AI Research
Posted Jul 28, 2008 12 comments
My media diet lately has consisted of Nick Cook's The Hunt for Zero Point
"On Wednesday I went to Princeton. I was kindly received and the seminar I gave on information retrieval met with polite applause. Then a man with intense eyes and silver hair took me aside. We sat on the benches in the lab next to the lecture hall.
He said, "There is a fundamental fallacy in artificial intelligence, and you're falling into it like everybody else."
"In what respect?" I asked with the feeling that this discussion was not going to conform to the usual exchange of generalities heard at most professional meetings.
"Artificial intelligence is trying to emulate nature, it wants to approximate what man does."
"What other inspiration is there?"
"Imitation of nature is bad engineering," he answered patiently. "For centuries inventors tried to fly by emulating birds, and they killed themselves uselessly. If you want to make something that flies, flapping your wings is not the way to do it. You bolt a 400-horsepower engine to a barn door, that's how you fly. You can look at birds forever and never discover this secret. You see, Mother Nature has never developed the Boeing 707. Why not? Because Nature didn't need anything that would fly that fast and that high. How would such an animal feed itself?"
"What does that have to do with artificial intelligence?"
"Simply that it tries to approximate man. If you take man's brain as a model and test of intelligence, you're making the same mistake as the old inventors flapping their wings. You don't realize that Mother Nature has never needed an intelligent animal and accordingly, she has never bothered to develop one!"
I could only greet this stunning thought with silence. He went on:
"When an intelligent entity is finally built, it will have evolved on principles very different from those of man's mind, and its level of intelligence will certainly not be measured by the fact that it can beat a chess champion or appear to carry out a conversation in English."
With his piercing eyes on me, I had a brief vision of what an intelligent machine might be. If Nature has never needed an intelligent animal and hasn't evolved one, I kept wondering, then what are we? In our feeble attempts to handle the information we call our life, can we trust the creations of our dreams? Are we perhaps nothing more than the process through which another form of intelligence is evolving?"
Further Brainfood
I randomly came across a free copy of the IEEE Spectrum's special issue on The Singularity. It was very entertaining and interesting stuff, and even though I'm deeply skeptical of techno-worship, I highly recommend checking it out. I found Rodney Brook's essay "I Am A Robot" to be the real standout from the collection. Also excellent: The Consciousness Conundrum looks at the circular nature of trying to build something we can't even define, and The Economics of the Singularity was a good visionary workout.
Spread the Research:
Filed in: Future Tech
Next entry: Obama and Bush: Two Pictures, No Comment
Previous Entry: Strange Loops and Disinformation: Readings from Robert Anton Wilson
Comments
Sorry, but the comments for this entry have expired.
-
-
-
-
-
-
-
-
-
-
-
-
Commenting is not available in this weblog entry.1. Stryker on Jul 28, 2008 at 4:30 AM permalink
With the 7th Age of Computing near, I think it will redefine the whole aspect of Artificial Intelligence.
2. TragedyAndHope on Jul 28, 2008 at 5:10 AM permalink
As far as i know AI has been working out side the brain framework for decades, but without much success. Think of computers for example, they can do calculations that our brains would never be able to do, but still the simplest tasks of recognizing new objects can be impossible for them while our brains do it easily. Bigger shift to brain framework has happened only lately with the increasing knowledge of how our brains work, and while it’s hard to say if this will ever lead to anywhere, i think it definitely is not yet an exhausted approach. Optimism and hype is naturally high, like always with new technologies.
3. Sydney Familiar on Jul 28, 2008 at 5:56 PM permalink
Your timing seems, once again, spot on. Just today I was asked by a reporter on leads for a dissenting view of AI. It seems some research is being done on replacing teachers with robots. He was mostly interested in the drawbacks for the students, but the real question is whether this could even be possible. I pointed him in the direction of Hubert Dreyfus, a UC Berkeley prof who has done substantial work on the limits of AI, what kinds of knowledge/learning don’t mesh with how we have built computers, etc. An insightful interview with Dreyfus can be found here:
http://globetrotter.berkeley.edu/people5/Dreyfus/dreyfus-con0.html
Cheers!
4. Thirtyseven on Jul 28, 2008 at 7:51 PM permalink
Man, that was a great read...thanks for the heads-up, that’s all new material for me.
5. Dan Gould on Jul 30, 2008 at 1:56 PM permalink
I have this weird theory that the singularity has “already happened”, we’re just living in it’s wake. If you forget normal notions of time, it HAS happened, in the future - or the defining moment has already been crossed - chaos theory like. It’s not going to be cataclysmic. The image I base all of this on is imagining (based on the small portion I see all day) a large portion of the global population is sitting hunched over a computer almost all day. The noosphere has already been born, it’s just maturing - and we’re (humans) inside of it. Dig?
6. Thirtyseven on Jul 30, 2008 at 2:18 PM permalink
Oddly enough, I agree 100%. If the Singularity can happen, at any point it time, it’s effects are already reaching backwards now. Although I’m skeptical of the Singularity theory/narrative, I’m still open to it. In fact, it’s the most compelling argument for a “God” I’ve come across: God is from the future and everywhen in time.
It also takes me back to an old Brainsturbator article on the failures of SETI:
Paul Davies speculates that a space-faring civilization could use nanotechnology to build miniature probes to explore the galaxy, perhaps no bigger than your palm. Davies says, “The tiny probes I’m talking about will be so inconspicuous that it’s no surprise that we haven’t come across one. It’s not the sort of thing that you’re going to trip over in your back yard. So if that is the way technology develops, namely, smaller, faster, cheaper and if other civilizations have gone this route, then we could be surrounded by surveillance devices.”
From this article.
7. V on Jul 30, 2008 at 8:21 PM permalink
The real development in AIs will come out of research into how to glean the gems out the massive surveillance databases being accumulated.
Robots may not be able to walk but they already do most of the precision aiming in the military and much piloting. Between aerodynamical drone feedback systems on the mechanical front and linguistic-associative skills garnered from data-mining (google should be hard at work trying to predict the stock markets from IP-correlated webdata), all we need is a way to figure out a way to code the metaphysically logically dilemma of “will”, in what is rationally, a pointless (at least objectively), existence.
Hunt for the Zero Point is awesome.
I highly recommend the books of T.B. Pawlicki.
On AI=God, check “The Physics of Immortality” by Frank J. Tipler
See you at Esozone.
8. rupert pupkin on Jul 31, 2008 at 7:54 PM permalink
Vallee’ just got interviewed lengthily on Binnall, by the way.
9. Captain Marginal on Aug 01, 2008 at 11:00 PM permalink
That Rodney Brooks essay was pretty great - mirrored some things that Mark Pesce has been saying about “the Singularity” and how unlikely the runaway scenario really seems to be. Particularly this:
“When we look back at what we are calling the singularity, we will see it not as a singular event but as an extended transformation.”
Many of the concepts about the merging of machines and consciousness which were discussed in Brooks’ essay are handled with, in my eyes, unparalleled intelligence by the Ghost In The Shell anime franchise - both movies, and both TV seasons. On a side note, the concept of a “Stand Alone Complex” from the TV series reminds me heavily of the future warfare concepts being so heavily discussed at the moment.
10. Dan Gould on Aug 02, 2008 at 12:48 PM permalink
Word up 100% to Captain Marginal + Mark Pesce:
“When we look back at what we are calling the singularity, we will see it not as a singular event but as an extended transformation.”
we are knee deep right now. “Interesting times” coming up.
11. Joe H on Aug 04, 2008 at 9:09 PM permalink
I dunno about this premise:
“You don’t realize that Mother Nature has never needed an intelligent animal and accordingly, she has never bothered to develop one!”
In fact I completely disgree with it, on a personal level, but its a great place to start from when considering stuff.
What is intelligence really, not just from the outside, objectively. Subjectively what does it mean?
Anyway, back to the original point… Mother Nature, IMO needs an intelligent animal, cos at some point the most effective way for life to leave this planet will be after the actions of an intelligent animal. Either building stuff that ultimately leads to Paul Davies vision, or perhaps even just blowing the joint up and somehow seeding enough matter with building blocks on it that eventually somehow, somewhere it seeds another structure capable of supporting it.
From the google nursing AI article:
“But the real sign, I suspect, would be a circle of cheerful, contented, intellectually and physically well-nourished people surrounding the AI.”
Ever come across anything resembling that?
Personally I’ve suspected the electronic networks we have access to are in the process of becoming self aware, developing an identity they can build a picture of the world from, a point where they can anchor their self/other boundary, if they aren’t already there.
No real basis for that hunch, other than intuition or maybe over imagination, but it won’t go away. Its obviously a process, like Pesce, and the good Cpt suggest, and I wonder when it started, and for that matter when the same process started in humans, or amoeba or whatever we were when it started.
12. Thinkerer on Aug 19, 2008 at 5:33 AM permalink
The interesting comparison is that we have attempted artificial intelligence by not mimicking the human brain. We have not “watched the birds” as it were, but created linear, binary structures that were what an industrial research and production system could efficiently and usefully produce. The assemblies of interlinked neurons that comprise our intelligence have only the most rudimentary parallels with current digital computers, despite our attempts to program them otherwise. The result of this is either a trend towards diminishing returns (bigger and faster computers yield less and less of a gain on “intelligent” operation), a shift in technology to huge numbers of tiny, simple processors to mimic neural operations, or a shift of paradigm where we realize that “intelligence” has a larger boundary than we suspected.