Thread: Ai

    #1
  1. No Profile Picture
    rebel with a cause
    Devshed God 1st Plane (5500 - 5999 posts)

    Join Date
    May 2004
    Location
    The Batsh!t Crazy State.
    Posts
    5,817
    Rep Power
    3462

    Ai


    Has anyone ever worked with any artificial intelligences. I've done some work with AIML but I was looking for other ideas. What experiences have you guys had in the area?

    I'm asking because I'm thinking about starting up a new project involving a learning AI that I can do a wide range of tasks with little intervention from the user. Kind of a helper instead of a tool. It's a continuation of some research I was working on as an undergraduate.
    Dear God. What is it like in your funny little brains? It must be so boring.
  2. #2
  3. Type Cast Exception
    Devshed Supreme Being (6500+ posts)

    Join Date
    Apr 2004
    Location
    OAKLAND CA | Adam's Point (Fairyland)
    Posts
    14,954
    Rep Power
    8617
    Unfortunately the "AI" class I had in college consisted of a guy who would assign "read chapters 4 and 5 in the book" then the next week would, literally, read those chapters in front of the class. Maybe once or twice during the course of the night he'd actually think to write some nonsense on the board and maybe expand upon a paragraph of the text with a personal anecdote, then went right back to the monotonous recital. It's too bad because at the time I was really fascinated by it and wanted to learn, but I got nothing out of that class. We did go on a field trip to some place doing some medical AI research, a sort of diagnostics system. That was pretty cool, but that was about it.

    There are a lot of open source AI applications that would probably be good to study. Chess engines come to mind. Or those sort of chat programs like we were playing with in the lounge "PC answers phone" thread. The latter is kind of like a much expanded "guess the animal" game though.

    I think studying the way chess engines work though would probably be quite revealing. Especially ones that learn by playing. They have to reject almost every possible path, yet still know to investigate paths that on the surface immediately look wrong but actually aren't (e.g. sacrifice moves that, some ten moves later, ultimately lead to check mate).

    I think the concepts are more important than any language specific implentation. Deductive logic is a good thing to review, probably, as well.
    medialint.com

    “Today you are You, that is truer than true. There is no one alive who is Youer than You.” - Dr. Seuss
  4. #3
  5. No Profile Picture
    Contributing User
    Devshed Newbie (0 - 499 posts)

    Join Date
    Jun 2005
    Posts
    156
    Rep Power
    10

    Theres a lot out there


    Theres a good bit of info out there on the web. One cool thing i have read about is mutating programs. Programs (asm) which rewrite machine opcodes (and other machine functions) to fit what they are doing. Try a google, im sure some stuff will turn up, ive found some.

    Also, look into game ai, it would be a great place to start (like chess as was stated above). Maybe you could find the code for some chess, or even more complex game ai engines, you could probably learn alot.
  6. #4
  7. No Profile Picture
    rebel with a cause
    Devshed God 1st Plane (5500 - 5999 posts)

    Join Date
    May 2004
    Location
    The Batsh!t Crazy State.
    Posts
    5,817
    Rep Power
    3462
    Yea I spent about a year as an undergrad doing research on vhui (virtual human user interface). I actually built a bot like the ones you guys were playing with. I wanted to do more work on it but I graduated and I didn't have funding for grad school at the time. So hopefully once I get into grad school I can find a professor willing to sponsor that kind of research but until then I'm working on my own.

    I was looking for insight into the various languages out there for this kind of thing. AIML is ok but I'm assuming there are better alternatives.
    Dear God. What is it like in your funny little brains? It must be so boring.
  8. #5
  9. No Profile Picture
    Contributing User
    Devshed Frequenter (2500 - 2999 posts)

    Join Date
    Nov 2004
    Location
    Washington DC
    Posts
    2,756
    Rep Power
    1572
    My school supposidly had one of the better AI programs and they did a lot in Lisp, so i would imagine that would be a good place to start. If you are looking to get a head start call up the department head at your prospective college.

    The other "languages" i heard tossed around were in house things that were basically just logic languages built for the specific problem at hand.

    AI is exteremly interesting stuff, i got a little discouraged with the whole thing, but i might be willing to give it a shot in a few years. Believe it or not, I'd almost be suprised if they have an AI track in a computer science department. It's more a philosphical and logic disapline than computer science.
    Open for extension, closed for modification
  10. #6
  11. No Profile Picture
    rebel with a cause
    Devshed God 1st Plane (5500 - 5999 posts)

    Join Date
    May 2004
    Location
    The Batsh!t Crazy State.
    Posts
    5,817
    Rep Power
    3462
    Thanks. I might do that. Some of the bigger cs programs like MIT have AI concentrations in their masters programs. But you're right. You almost need a degree in behavioral sciences to have some understanding of how to make these things work.
    Dear God. What is it like in your funny little brains? It must be so boring.
  12. #7
  13. 'fie' on me, allege-dly
    Devshed Supreme Being (6500+ posts)

    Join Date
    Mar 2003
    Location
    in da kitchen ...
    Posts
    12,894
    Rep Power
    6444
    I hate to say it, but Prolog (spit) has some serious applications in AI because of it's predicate logic, and that it can learn as it's going along. IIRC there was a product called Kappa (1992), so unless it's current, I'd discount that (probably)
    --Ax
    without exception, there is no rule ...
    Handmade Irish Jewellery
    Targeted Advertising Cookie Optout (TACO) extension for Firefox
    The great thing about Object Oriented code is that it can make small, simple problems look like large, complex ones


    09 F9 11 02
    9D 74 E3 5B
    D8 41 56 C5
    63 56 88 C0
    Some people, when confronted with a problem, think "I know, I'll use regular expressions." Now they have two problems.
    -- Jamie Zawinski
    Detavil - the devil is in the detail, allegedly, and I use the term advisedly, allegedly ... oh, no, wait I did ...
    BIT COINS ANYONE
  14. #8
  15. No Profile Picture
    Registered User
    Devshed Newbie (0 - 499 posts)

    Join Date
    Dec 2005
    Location
    Australia
    Posts
    10
    Rep Power
    0
    I did my degree a long time ago. I did software developemnt and AI. It was a double degree. While it was a lot of lot of philosphical discussion, we did actually build some projects.
    We built a lotto number generator using Miranda
    We built a medical diagnosis app using Prolog
    We also examined another medical app which was built using Eiffel
    We also did some programming in Pop-11.
    Unfortunately while I was very interested in this field, there was very little call for this in Australia.

    Ken
  16. #9
  17. 'fie' on me, allege-dly
    Devshed Supreme Being (6500+ posts)

    Join Date
    Mar 2003
    Location
    in da kitchen ...
    Posts
    12,894
    Rep Power
    6444
    It's a big bad world now ... anything you'd like to share with your fellow sheddites. We have mareketing capability, you know, did you like the curtains, no, the dado rails, no

    I for one am lost as to how to appeal to the international business community, in any other way than saving them sh!tloads of money, but that's a 'me' thing kay.

    Have you heard of this sponsorship deal ... sorry, doorbell, BRB
    --Ax
    without exception, there is no rule ...
    Handmade Irish Jewellery
    Targeted Advertising Cookie Optout (TACO) extension for Firefox
    The great thing about Object Oriented code is that it can make small, simple problems look like large, complex ones


    09 F9 11 02
    9D 74 E3 5B
    D8 41 56 C5
    63 56 88 C0
    Some people, when confronted with a problem, think "I know, I'll use regular expressions." Now they have two problems.
    -- Jamie Zawinski
    Detavil - the devil is in the detail, allegedly, and I use the term advisedly, allegedly ... oh, no, wait I did ...
    BIT COINS ANYONE
  18. #10
  19. No Profile Picture
    Registered User
    Devshed Newbie (0 - 499 posts)

    Join Date
    Mar 2006
    Posts
    2
    Rep Power
    0
    I'd suggest doing a lot of reading around the subject ... looking closely at neurofuzzy and genetic algorithms ... Take a look at what Creative Virtual do in the realm of NLP (since the abstraction approach is probably similar to what you're hoping to achieve, even if the implementation serves a different purpose in a different way)
  20. #11
  21. Hello World :)
    Devshed Frequenter (2500 - 2999 posts)

    Join Date
    Mar 2003
    Location
    Hull, UK
    Posts
    2,537
    Rep Power
    69
    I'm can't really comment on AI because I've only read a little about it but I would recommend Lisp as your implementation language; I prefer Scheme myself but both lend themselves to this kind of app much better than more "traditional" languages.

    Good luck,

    Mark.
    programming language development: www.netytan.com Hula

  22. #12
  23. fork while true;
    Devshed God 1st Plane (5500 - 5999 posts)

    Join Date
    May 2005
    Location
    England, UK
    Posts
    5,538
    Rep Power
    1051
    This is definitely lisp's traditional domain. As for flavours, you'll want a nice featureful lisp such as Gnu Common Lisp...
  24. #13
  25. Hello World :)
    Devshed Frequenter (2500 - 2999 posts)

    Join Date
    Mar 2003
    Location
    Hull, UK
    Posts
    2,537
    Rep Power
    69
    There are a lot of good Lisps, you should browser around if you decide to take this rout because they vary in quality and feel. It also depends quite a lot on the platform your targeting also so it's worth taking your time to choose a good one .

    There tend to be IMO better compilers for Scheme and the language is cleaner so if you want to end up with an exe this is a good rout to take.

    Other than that just try as many as you can to get a feel for it.

    Mark.
    programming language development: www.netytan.com Hula


IMN logo majestic logo threadwatch logo seochat tools logo