lew (12/02/82)
It is a cliche that birds totally lack intelligence and behave according to a fixed algorithm. Yet their nest building behavior, including the ability to seek out materials in an unstructured environment, completely dwarfs current AI achievments. I suppose one could argue that various AI achievments are on a qualitatively different plane than bird behavior, but I am leery of this idea. Look at birds' exquisite real-time processing of vision and control of flight, and all high level goal directed. I think it is foolish to dismiss these abilities as insignificant to the understanding of intelligence. I essentially agree with Weizenbaum that the claims of AI achievements to date are ridiculously overblown. Incidentally, I am not a vitalist. That is, I don't think that there is any extra-physical element required to achieve intelligence, or even consciousness (I think these are implicitly confused in many discussions.) But I don't think "true intelligence" can be achieved with current computer architectures. Lew Mammel, Jr. ihuxr!lew
trb (12/03/82)
Birds vs. AI, ha! Brings me back to the days of my youth... I was a young (15) freshman taking a philosophy seminar, and boy, was my prof strict. We were sitting around a table talking about things being animate or inanimate. I suggested that rocks might be animate but with a greater life span than we humans could observe. Prof starts getting upset. She didn't buy the rock biz, so I took a sheet of paper from my notebook and said "Look at this piece of paper. Look at how sedate it is. Happy and well rested." Then I crumpled it up (with much hackeresque arm-waving). "Look at how down-trodden it appears!" (smashing it with my fist) "Look at it writhe!" (as it uncrumpled) "If I set it afire it might crackle and blaze and turn black with anguish, fall apart and eventually die!" (I don't smoke, I had no matches, too bad). "It's alive!" She flunked me for the course; I don't think she liked my attitude. There's a point here. This article is a protest against those folks who claim to be in THIS astral plane AND talk about making real progress in AI in the near future. I know you can have fun and show off and build lisp machines if you're an AI researcher, that's fine. Just don't tell me about Artificial Intelligence until you have enough proof that we'll believe you. Andy Tannenbaum Bell Labs Whippany, NJ (201) 386-6491 (I was a teenage AI project)
pcmcgeer (12/05/82)
If AI is a failure, as Messrs Mammel and Tannenbaum maintain, then we should remember that it is a heroic one. Many of the great triumphs of our field - symbolic computation and language parsing, for example - started off as AI projects. Granted, the AI folks haven't built an intelligent machine yet: however, in their attempts to do so, they gave us Risch integration and every modern compiler. I'd like to fail like that, too. It is interesting that Dr. Weizenbaum should debunk the achievements of AI: he sat on Joel Moses' thesis committee, and therefore should be more aware than any of us of Dr. Moses' unique triumphs. Rick.
trb (12/06/82)
I have never in my life said that the AI community hasn't produced enough fruit to be unworthy of the effort contributed to it. The AI/CS communities at our favorite hacker schools have produced some of the most aesthetically pleasing (and USEFUL!) hardware/software known to computerdom. I am saying that I have never seen an AI demonstration that would impress an audience made up of computer-literate people (who wanted to see only real technical achievement) and lay people (who wanted to see a computer acting like you and me). Even if they never come up with a machine to properly replace relationships with people of the opposite sex (some people claim that today's computers do that, that's not what I mean) I will still be forever grateful for the AI community for inventing the tools that have and will make our jobs pleasant to do. (I can't resist this...) I wish people would stop intermingling my ideas (in this case that the AI community has produced no real AI fruit) with ideas that I never dreamed of claiming (that the AI community has been totally fruitless). It is tough enough to defend oneself from wielders of flawed logic in face to face conversation, it's even tougher to do it gracefully over netnews. Andy Tannenbaum Bell Labs Whippany, NJ (201) 386-6491
leichter (12/07/82)
Re: AI has given us every modern compiler Sorry, that won't wash. The parsers used in modern compilers are quite unrelated to anything done for natural-language processing. Parser theory goes back to ALGOL-60 and was advanced to its current state of sophisti- cation mainly be some compiler heavies and some theoreticians. The problems of natural- and programming-language parsing actually have little to do with each other; the kinds of very difficult disambiguation algorithms and just "general state-of-the-world-and-this-conversation" considerations that are central to natural language parsing - which is quite inseperable from natural language understanding - just have no analogue in the programming language world; and the programming language issues that arise - like recursive constructs - tend to be rather trivial in natural languages. -- Jerry decvax!yale-comix!leichter leichter@yale
soreff (12/07/82)
Open letter to Andy Tannenbaum: What exactly would you consider a real AI accomplishment? You stated that "... the AI community has produced no real AI fruit". I freely admit that no AI program passes the Turing test. Couldn't some lesser accomplishments be admitted as AI accomplishments as well? After all, no one says that biochemistry has been thus far unsuccessful as a science because no one has synthesized an entire cell starting from inorganic chemicals and elemental carbon. -Jeffrey Soreff
bj (12/08/82)
(I can't resist this...)
I wish people would stop intermingling my ideas (in this case that the
AI community has produced no real AI fruit) with ideas that I never
dreamed of claiming (that the AI community has been totally
fruitless). It is tough enough to defend oneself from wielders of
flawed logic in face to face conversation, it's even tougher to do it
gracefully over netnews.
Andy Tannenbaum Bell Labs Whippany, NJ (201) 386-6491
This is a big problem. A large number of responces that appear on the
net (and of the responces I have received by mail) are written by people
who have not read the original article carefully. They often just skim
the article, pick sentences at random, and mis-read them. These articles
(along with original articles by people who don't know what they are
talking about) makes up much of the garbage on the net.
Please, *READ* and *THINK* before you flame.
Not afraid to agree with an AI project,
B.J.
decvax!yale-comix!herbison-bj
Herbison-BJ@Yaletrb (12/08/82)
Open answer to Jeffrey Soreff's question: What exactly would you consider a real AI accomplishment? Ask your mother. I'm not really kidding. Round about the time I was busy being born there were intelligent researchers with stars in their eyes thinking about how they could use this newfangled digital computer technology to think and act like flesh-and-blood people, only faster and better. I'm sure those guys weren't talking dreamily about the AI accomplishments that have come to fruition as of 1982. Have you ever seen Hymie in "Get Smart" or HAL in 2001 or any of the many robots in SF and comic books? The fans just wouldn't have bought some guy hacking an expert system or working on a natural language interface with no human-like (or better) smarts behind it. This is what I mean by "real AI fruit." I didn't say that the AI community has produced no real fruit, I said "no real AI fruit," that is, no *artificial intelligence*. What do I mean by AI? A Turing Test would do just fine, really. Spend 1/2 hour talking to me over a tty link, and then talk to some other AI project's favorite baby, and I'm sure you'll have appreciation for the progress I've made. (I hereby challenge you.) Some lesser accomplishments are certainly appreciated for their merits, as I was glad to emphasize in the article you were replying to, but I was just drawing the line between the original intent of AI and what's come from it. If I set out to turn water into wine and I come up with Tefl*n instead, I guess I would argue that I produced no real "turning water into wine" fruit, and you would argue that I certainly had been productive. Lesser accomplishments are admitted today as "real AI fruit" only because of disenchantment with the original goals. Assuming that producing Hymie's and Hal's is possible (and I think it is possible), the AI community has thus far not borne its intended fruit. Andy Tannenbaum Bell Labs Whippany, NJ (201) 386-6491