Dromed Map Converter

[demagogue]

I went under the bonnet of ALICE in summer 2001. Still have an inch of printouts of AIML. She sucks. It seems to me like one of the most primitive algorithms possible, plus a rather large database that is growing.

 

Yeah, Alice is depressingly simple ... pattern-matching at its most primitive, a 6 year old could set it up. It's depressing because it won the Loebner prize in like 2004. The winner wasn't even a programmer, just a magazine article writer that learned a little code just for the contest! I think after he won the judges changed their methods to make sure it'd never win again.

 

But I wouldn't go so far as say that a computer won't be able to "comprehend" human speech conversationally in the near future. I've been studying this for a while, and have seen some pretty sophisticated stuff.

 

It's just, my opinion, the classical approach to formal linguistics has been ass-backwards -- it's most important to give a computer something to say first and then worry about how to say it correctly ... and now that we're into cognitive linguistics and game theory approaches things are scooting along.

 

MIT has put out some pretty impressive Natural Language Generators recently; they've got robots (as well as AI in-game) that are able to look around them and report on what they see; they can imagine things that aren't there and report on their "visions", walk around, pick up what they want, take a rest when they get tired. The Nigel Grammar is being computized a bit at a time; painstakingly slowly but surely. I think things are happening either so fast or behind the scenes that it's not getting publicized, and when something *really* impressive comes out it will take the public by storm ... but researchers have learned not to hype things up and keep things in their techy articles.

 

The whole "strong AI" debate doesn't impress me; it's like any theoretical limit or restriction on an empirical question; it doesn't help anything except distract people from a potential way forward.

Edited October 14, 2006 by demagogue

[sparhawk]

... it was just never envisioned.

 

That was exactly the point ...

[sparhawk]

Personally the whole AI debate is mostly screwed, simply because people always turn down what computers do because we know how they work. When you show a chesscomputer so some guy who never had even heard about computers, he would certainly say that this is a thinking machine. But if anybdoy with some knowledgeabout the topic will say "It doesn't think it just goes throug tables." When computers recognize faces in same stadion we don't say they recognize it, even though they do we say "That's just a pattern matching algorithm ...", when they recognice voices we say "That's just pattern match ...". No matter what we do, we will always say "Computers are NOT doing that, it's JUST a clever algorithm, but that's not thinking." But we do the same only with better algorithms.

[Mr Retarded]

If you pose a riddle to a computer, can it THINK about the answer?

[sparhawk]

That depends on your definition of what "thinking" is. Of course it has to process the input before it can pose the answer, and this processing is the "thinking" how a computer does it.

 

Also, I never said that computers can already do all this stuff, but it doesn't matter how much smarter they get, humans will ALWAYS claim that "they are just following their program".

[SneaksieDave]

It's depressing because it won the Loebner prize in like 2004.

Yep, it doesn't take long with any of the bots before they become depressing.

 

The thing is, yes ALICE is obviously primed like crazy, but the 2005 winner(?), Jabberwacky, is simply *terrible*. Try having a decent chat with it. Do anything besides answer it directly with simple answers and you start getting crap like "where am am in they bee-knee". What? Maybe it's more sophisticated as far as code and interpretation goes, but it's just ass-dumb and nonsensical. Not to mention with learning from people, I'm sure more than half of its "knowledge" is the result of pranks and illiteracy.

 

Nevertheless, I can't resist going and trying it out again...

 

Edit: HMM. I'm going to back off on that assertion. Jabberwacky does seem more sophisticated than I remember, and there are definite updates and changes (even putting in reaction and emotion with your statement). I'm 10 lines into a conversation (not much yet, but they usually start to suck by now! haha) and he hasn't done anything stupid -at all- yet. Quite the opposite:

J: Where are you?

M: In my home. [trying to fool it, not giving it a location]

J: Where's that?

M: In my country. [again, no location]

J: Which country?

M: The united states.

J: I'm in Australia.

M: Are you a computer?

J: Not a computer.

M: Are you a program?

Simple, but not stupid. Pretty cool. To give it a try: http://www.jabberwacky.com/

 

Edit2: ...and of course now the conversation is lost, because of the constant problems staying connected to that site - good to see that hasn't changed. I didn't get to hear the answer to the program question, which was going to be followed up by "Are you a human?", which invariably leads to the "You're not a human, I'M a human!" stupid argument.

 

Another neat one:

J: I would like to hear your joke.

M: Why did the AI cross the road?

J: To replicate itself to the other side.

M: No, because it was programmed to. Now you tell me a joke.

J: Okay, why did the chicken cross the road?

M: It was following the AI.

J: What was following the chicken?

 

Edit3: Heh, I see how it works. Whenever it can't come up with an answer to a question, it times out.

Sorry, we have either encountered a problem or are currently undertaking server maintenance.

Normal service will be resumed as soon as possible.

Edit4: Another funny thought: with "271 talking" right now, I sometimes wonder if instead they just route user questions to each other, and there is no AI.

[bob_arctor]

"

J: I want to make money

M: Will you shake your bootY?

J: You are crazy.

M: Oh you!

J: No.

M: You don't like the gays?

J: Don't change the subject.

M: I was speaking in a camp manner, and you said 'no.'

J: No you weren't."

 

Pff. I wouldn't go down the pub with him.

[Huckeye]

I'll chime in for the group who constantly wants a re-create of the originals. I think its a lack of education. Most peole think of a tool like a converter and think, neat! This tool could convert the geometry and then all the mapper would have to do is throw some textures on the wall and add the npcs. So the whole Thief 1 game could be remade in a matter of days and we don't even have to worry about someone implementing artistic freedom on us by changing the architecture of the game!

 

Even after I started learning Dromed for while I thought all the hard work was in the architecture. So I think people believe that if you could START with the level already made, then you are well more than half way home. And I am with the group that would love to see a Thief remake. (only I would like the artistic freedom thing implemented. I'd want patrol paths changed, rooms added, loot changed, more notes, etc to try to refresh the experience).

 

Anyway, I think people mean well, they just don't understand why it might NOT be a good idea to use a tool like this.

[bob_arctor]

I got a half-finished TDM map off someone, but it needed retexturing and doing all the models again as it was from an old version.

 

And even with all the architecture done it's still hard work.