Hi all, I've been lurking about having just found this fabulous website. I am totally addicted and can't wait for 'The Well of Lost Plots'. I believe waiting 12months is entirely too long !

I must confess to being owned by a very fluffy hyper dog named Butler. My dog does not drool(ok maybe a little when confronted with a whole chicken on the counter), will do anything for food, prefers sofas and beds to any floor for sleeping and always greets me enthusiastically when I get home.

Butler says hello to everyone by the way. Actually, he jumped on the keyboard and what it really said was "9phawp908if". So, hello to everyone from across the pond!

Tracy

Hello back at you ( and to Butler).

************************************************************

"This was willed where what is willed... can get rather silly."

THis always happens when there's a rash of newbies all at once (you obviously all fell out of the same closet that you'd been hiding in) - if I have said hello already ignore me, but welcome to the fforum!



PSD

==========

This is the work of an Italian narco-anarchic collective. Don't bother insulting them, they can't read English anyway.

Terri: yes, I'm certainly interested in those areas, but I don't actually study them as such. I might if I didn't have a full-time job. Doesn't Terry Winograd get a mention in "Goedel, Escher, Bach", by the way?

Hi to everyone who's new here... including Butler!



..........................................................................................

That which does not kill us makes us stranger.
(Llewelyn the dragon, Ozy and Millie)

Sarah

Sarah, yes, Winograd's SHRDLU program gets about 5 or 6 pages in Chapter XVIII, "Artificial Intelligence: Retrospects", plus the dialogue at the end of the preceding chapter (recall that the text of each chapter discusses the dialogue at the end of the previous chapter) is an actual SHRDLU transcript, which Hofstadter has titled "SHRDLU, Toy of Man's Designing". Winograd's commentary is captioned "Dr. Tony Earrwig".

My master's thesis was a SHRDLU-like program in which the "toy" domain was set theory and number theory (instead of blocks), and the program could extend its grammar and vocabulary through the conversation. It was called "Gretel", for "grammatically extensible through example language":

Sample Gretel conversation:

User: Every set is a subset of itself.
Gretel: I don't understand the word "itself".
User: Every set is a subset of itself means if X is a set, then X is a subset of X.
Gretel: OK
User: What is the successor of 4?
Gretel: I don't understand the word "successor".
User: If N is a number, then the successor of N is N+1.
Gretel: OK, then the successor of 4 is 5.
User: For every number N, N is a factor of N.
Gretel: OK
User: Is 6 a factor of 6?
Gretel: Yes
User: Why?
Gretel: Because every number is a factor of itself.
User: Is 3 a factor of 6?
Gretel: I don't know.

All of which has nothing to do with cats.

Terri,

Was this work postulated for the 'Turing' panel test for AI at all? I'm not familiar with any of this, so apologies if that is a dumb question.

I have cats (2 British Shorthairs) too, if anyone is interested in pictures and my labourious text then you can find it at my husband's website, listed in my profile ( you have the choice to skip past the cars and wedding bit ;)

B

Belotchka, by "this work" do you mean SHRDLU or Gretel?

SHRDLU is often discussed in connection with the Turing test, but largely by way of pointing out how important it is for the test to allow the "examiner" to lead the conversation rather than reading a transcript. SHRDLU was a ground breaking program mostly because of its extensive rules about keeping tracking of the topic of conversation and resolving pronoun references and short questions (like "Why?") within that context. It's what makes the transcripts sound so real when you read them. It's very hard to shake the feeling that SHRDLU understands what it's talking about, as long as someone familiar with the program is carefully guiding the conversation along the paths that show it to advantage. But like all toy domain programs, they tend to fall apart quickly and seem very unintelligent as soon as you push the conversation outside their small domain of knowledge.

Gretel was just a master's thesis and wasn't "proposed" for anything except getting me my degree. They aren't even published (like a PhD thesis is.) It's just sort of like a big homework project to show that you understand how to do work in the field, but it isn't expected to contribute anything new to science. I was seeing how far I could go with the idea of making language itself become part of the toy domain. Which is an interesting "strange loop" to use Hofstadter's terminology, but it still breaks down pretty quickly when you go outside of math for what you're going to talk about. The reason I chose math as what to talk about besides language itself is that it "bottoms out" in something computers already understand (or at least parse, understand being a loaded term), so that you can define one word in terms of another and pretty quickly hit something that you don't need to define further. But try explaining cats and dogs or time travel or book jumping (a hand-wave at bringing this back on topic) to SHRDLU or Gretel, and you quickly run into the problem that the computer has no base of shared experience to rely upon, so you get into this chain of trying to define one concept in terms of another and never finding anywhere to stop. And a transcript of THAT conversation sounds patently brain dead.



Post Edited (05-28-03 02:17)

Thanks for explaining about SHRDLU and your thesis work Terri, I think that expanded my own knowledge base x 100, I really am quite ignorant about this area! I find the precision of mathematics as language fascinating probably because I'm a total duffer at understanding it :)

B

Shoot. Well, it had to happen one day, I suppose. Now we've got someone really clever on the fforum. Bad news for all us bluffers.

(I actually understood some of that ... quite pleased with myself ...)



- - -
I am very interested in the Universe. I am specialising in the Universe and everything surrounding it. - E. L. Wisty

Teri: That sounds like pretty smart stuff for the 70's. I hadn't even got my ZX81 then... I occasionally played on a Research Machine Z80. Are you doing any fun stuff now ? Some of my mates did image recognition (handwriting) for their theses back in the 90's. Never done as much AI as I'd like.

sounds like clever stuff to me. Fascinating.

Sorry everyone, I just couldn't resist this one:

One day little Johnny was in his back yard digging a hole. His neighbour, seeinghim there, decided to investigate.

"Whatcha doin?" he asked.

"My goldfish died and I'm burying him," Johnny replied.

"That's an awful big hole for a goldfish, isn't it?" asked the neighbour.

"That's because he's inside your cat!"

Ah, yes, of course... "Dr Tony Earrwig". I remember now. Unfortunately my copy of GEB was borrowed and not returned, and I can't remember who the miscreant was, mutter, mutter, grrrrr.

I tend to use my knowledge of issues relating to artificial intelligence to write convincing robots in science fiction stories, not having had a chance to do very much in that field for real. I'm fascinated by Gretel, and indeed any program which is capable of learning about its environment; I agree with Rob, it was a long way forward for the 1970s. Even so, I still say we won't get really intelligent machines until we have first-rate artificial perception available, because as I see it cognition depends crucially on perception.

Teri: have you read "Fluid Concepts and Creative Analogies" (another one by Hofstadter)?



..........................................................................................

That which does not kill us makes us stranger.
(Llewelyn the dragon, Ozy and Millie)

Sarah

I agree. There's a world of difference between an intelligent user interacting with a computer; a computer learning something in its subject (Bayesian statistics and neural nets all seem very clever (the Spark gender test is great - [test3.thespark.com] ); and a genuinely intelligent computer.

IMHO, we're years away from a computer which will pass the Turing test. As I tell my students, give me any computer programme and I'll find a way to break it...

Rob wrote:

> As I tell my students, give me any computer
> programme and I'll find a way to break it...

Oh, I can do *that*. For a living, even.



- - -
I am very interested in the Universe. I am specialising in the Universe and everything surrounding it. - E. L. Wisty

heck, we've got a building full of people who do that on a daily basis.

Grrr. I'm on helpdesk support this month. Boo.

One guy logged a support call to say that the water cooler had run out of water.

We thought about asking him to reboot it, but then figured he wouldn't get the gag.

Helpdesk work is such fun! You mean you've only been asked about a water cooler yet? No photocopiers that have run out of toner, ditto for printers and the always favourite paper jams.

Lucky, lucky you.

B

Terri, I studied SHRDLU in psychology at uni - I'm very impressed! Everyone on here impresses me, it makes me feel quite inadequate ;o)

I did a bit of AI in a computer course, too. I'm interested in NLP, though not to the extent of reading actual technical books about it, usually... Ooh, has anyone read a book called Galatea 2.2? It's a novel about trying to make a program that could pass an English exam, it's very good.



:--

Do something pretty while you can...

we have a printer called bonfire. It amuses us to phone users and ask them to put paper in the bonfire.

Hey, it was a slow day, ok?

We had a gem of an email recently too. The email server is broken, so most people can't log in. Someone (who shall remain nameless) thinks that sending an *email* about it is a good idea.

"PLEASE NOTE : We are experiencing problems with connecting to the new mail server which many of you are experiencing - please call 123 if you can't connect to your email and we will fix the problem. Please have patience with us this morning as problems are widespread."

I kid you not. This person is fairly senior in the IT dept too.

Dave - we've got 'senior' people like that here but your example beats anything we've had!