[MUD-Dev] Finding Space
cg at ami-cg.GraySage.Edmonton.AB.CA
Sun Aug 17 08:54:25 New Zealand Standard Time 1997
:> :Does anyone have the answer? :)
:> Well, this is beginning to sound a *lot* like Winograd's "Blocks World".
:> He got a PhD degree for implementing that!
:I dug up a URL that looked kinda interesting. Do you have a pointer to a good
:site/paper for that?
It's pretty old stuff, so might not be available online. I'm not even sure
Winograd is still at MIT, but you might start there. I read all about it
for an AI course in his book "Understanding Natural Language", from
Academic Press in 1972. The capabilities of his system are certainly
something we can aim for, but I don't think we have to be that good for
a game! E.g. the beginning of his sample session:
1. pick up a big red block.
2. grasp the pyramid.
I DON'T UNDERSTAND WHICH PYRAMID YOU MEAN.
3. find a block which is taller than the one you are holding and put it
into the box.
BY "IT", I ASSUME YOU MEAN THE BLOCK WHICH IS TALLER THAN THE ONE I
4. what does the box contain?
THE BLUE PYRAMID AND THE BLUE BLOCK.
5. what is the pyramid supported by?
6. how many blocks are not in the box?
FOUR OF THEM.
etc. (goes up to 44.)
Basically, he decided to approach the issue of natural languuge
understanding by using a micro world (a lot like MUDs do), and providing
detailed semantic understanding in that world. His system modelled a
robotic arm that could move around simple blocks, etc. on a tabletop.
It was able to understand input, plan ways of doing things, etc. I'm
not sure how he handled the problems of 3D positioning, but it was able
to stack things up, and move things out of the way.
Chris Gray cg at ami-cg.GraySage.Edmonton.AB.CA
More information about the MUD-Dev