Fun Stuff > CLIKC
Artifical Intelligence
McTaggart:
Could you please remove the picture in your sig? We're not really fans of them here.
nescience:
Two words: LISP. Oh, and two more words: intractable.
I actually did some work on Cyc when I worked at PARC. It's a pretty fun little system that is powered by an exceedingly boring knowledge base. Guess what I got to work on???
As far as AI ideas go, I don't have all that much. I definitely believe that given the current technology, the fate of strong AI is inextricably tied to good Natural Language Processing. Write a program to facilitate understanding of a natural language and you've immediately revolutionized the concept of the API. Tell the program to write the interface for you!
I am also a firm believer in the need for good hand-tuned ontologies for multiple fields, all then cross-referenced. Most ontologies feel rather limiting today because they are limited in their scope, thus causing a disconnect in the ability of the computer to do topicalization. What we need is five thousand monkeys working on five thousand typewriters and eventually we will get a unified ontology that will facilitate better question-answering. Wait, is that enough monkeys?
My only "real" contribution: I devised a procedure for my company to immediately increase performance for speech technology applications that handle multiple languages.
Inlander:
--- Quote from: moley ---and then it would realise "Holy shit! I can take over the world!"
--- End quote ---
Which is where the empathic thinking comes in. It won't take over the world if it doesn't want to, and if you programme it appropriately then it won't want to do something that's going to harm a lot of people. Basically, don't programme your A.I. with the mindset of a psychopath.
Orena:
--- Quote ---
Two words: LISP. Oh, and two more words: intractable.
--- End quote ---
oh man , i really hated lisp...hours and hours of debugging brackets... no fun methinks !
DavidGrohl:
--- Quote from: whitehatblackshoes ---So, I have been thinking about how to create AI for awhile, and I came up with a break through in how to program one to learn, which eventually lead me to think about androids and the waves of the future.
Anyways, I figured that for AI to be truely AI, it needs to be able to learn. Any 3rd grader will be able to tell you that. But the question is "how can we program that?" It's not a question of how to program it, its a question of developing a programming language specifically for AI, a language that is able to sift/filter through sensory input and derive data which it then uses at a later date.
Complicated, maybe impossible, but here is my question for everyone. What ideas do you have for AI?
--- End quote ---
AI isn't impossible, and is already well on its way. As a Computer Science major, I'm required to take certain elective courses, one of them was Advanced AI CS346. While AI is only limited to program script, it's the first building step and also proves that it IS possible, though it would have to be done on a larger scale.
An example of simple AI would be the chess program that beat the chess 'superchampion' 2 out of 3 times (I think it was). Anyways, this is done on a finite scale (namely, a chess board), thus there are a finite possible of moves it can predict.
Google "ALICE". Add her on your AIM . . talk to the program, it learns. Remembers your name, you can teach it new words. It's still being programmed and added to daily. This is an AI 'first'.
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version