Creating a Powerful AI

[chrisname]

Good idea. Although I would have a few steps between the XML parser and the English language parser!

[Albatross]

Erm... Latin parser?

-Albatross

[helios]

Oh, god, no! Anything but Romance languages! English is ~20 times easier to parse.

[chrisname]

How about a Finnish or Icelandic? Those will be fun.

[firedraco]

Maybe it would be easier working backward...i.e. starting at something like a simple scripting language and expanding it to work with more keywords etc, or trying to generalize it. I dunno, just throwing out random ideas.

[fruitloop2222]

Well if you think that I'm not being realistic then you obviously don't understand my goal. I am working exactly how "firedraco" has specified. I'm starting simple. Even though my dreams are big, you think I'll become discouraged by the dreams themselves. No, this is not true. How can a program become complex without it being simple in the initial process? Duh! I'm going to start with something simple.

[helios]

What exactly is a "universal parser"?

[Albatross]

A universal parser is a parser that takes in all the knowledge of the universe and spits out a distribution with a mean value of 42. Standard deviation: 0.

Where did you get that from, though? Deleted post?

-Albatross

[helios]

I've began research in parsing as well, indeed. So I wanted to create a universal parser system. I'm working on a parser class and once I get home I'll be able to show it to you.

Maybe he means this?

There are three general types of parsers for grammars: universal, top-down,
and bottom-up. Universal parsing methods such as the Cocke-Younger-Kasami
algorithm and Earley's algorithm can parse any grammar (see the bibliographic
notes). These general methods are, however, too inefficient to use in production
compilers.

The Dragon Book doesn't use the word "universal" in a similar context again.

[Albatross]

Then I think he meant just that: a parser that can parse any grammar.

-Albatross

[sargon94]

This thread is getting old, let's kill it. And by the way i love the post from Albatross:
value=42 SD=0
awesome!

[cnoeval]

You can't kill a thread, you can only let them die!

oop! Just revived it again.

[helios]

You wasted a chance to say "you can't kill a thread, you have to let it return on its own".

[m4ster r0shi]

Actually, you can kill a thread. But when you do this, you don't allow it to do any necessary cleanup...

[helios]

I know, but that didn't fit in the context.

[Albatross]

Can't we just kill this thread and forget the cleanup? It's bad form, but I'm not sure how much can be done here...

(Seriously).

-Albatross

[m4ster r0shi]

helios wrote:

but that didn't fit in the context

It was intended not to.

EDIT:

perhaps I should have said:

You wasted a chance to say
"Actually, you can kill a thread. But when you do this, you don't allow it to do any necessary cleanup..."

[exiledAussie]

Wow...

I work in AI, and kid - you sound like a jerk.

That said, I like that you are interested. So...

Language is hard... f**king hard. Some people work on getting computers to understand language and I am impressed by what they do. But since we do not really understand how language works, a better bet is to work on analysis of data (once we understand language, um, hopefully this will help us there too...).
Read up on neural (of all varieties, since it is a pretty varied bunch) and baysian networks.
Learn the math that will allow you to understand the above properly.

Show the bastards by creating the best AI ever :)

[cnoeval]

The threads on the Python forums have automatic garbage collection. 8^p

[sargon94]

humm, a few things:
1. @cnoeval: Ya but python fails
2. @exiledAussie: even though i supported the guy, i haev to admit he's way out of his league here. The day he creates AI i'll eat my hat, sadly for him, I'm against hats and don't own one.
3. C'mon! can't we just let this go?? Viblicent, declare the thread solved or something!