A.I. Manifesto

Discussion in 'Intelligence & Machines' started by scifitm, Feb 17, 2008.

Thread Status:
Not open for further replies.
  1. scifitm Registered Member

    Messages:
    46
    Anyone interested in building an A.I.? or know where there's some opensource code for it? This seems like a very interesting undertaking, and maybe we could do it as a community...
     
  2. Google AdSense Guest Advertisement



    to hide all adverts.
  3. draqon Banned Banned

    Messages:
    35,006
    and what code is it in?
     
  4. Google AdSense Guest Advertisement



    to hide all adverts.
  5. scifitm Registered Member

    Messages:
    46
    I saw an article on the Leaf project, and they were using a code I had never heard of. I don't believe it would matter much as long as it had the assembly capabilities to interact with usb ports, ide, and sensor plugins etc. Although as I've stated in other threads, really to me all an A.I. has to do is make a choice - and maybe be able to validate it.

    Any suggestions?
     
  6. Google AdSense Guest Advertisement



    to hide all adverts.
  7. draqon Banned Banned

    Messages:
    35,006
    yeah C++
     
  8. Vkothii Banned Banned

    Messages:
    3,674
    No, the coding language should be functional or stack-based, but OO of course. Designing the A.I. should be the first thing, though. What kind of "intelligence" and what level? How general or specific a kind of "intelligence" could be looked at, neural nets, expert systems, what specifically?
     
  9. kmguru Staff Member

    Messages:
    11,757
  10. K.FLINT Devil's advocate :D Registered Senior Member

    Messages:
    225
    A.i.

    I agree you need to know what you want out of the A.I. what is its main function?

    I would really like to try adding self awareness, self improvement and self learning in to the mix.
     
  11. YinyangDK Registered Senior Member

    Messages:
    209
    All you have to do to make a program that will learn over time, is to have it make binary choices every sekund and evaluate on the choices every hour.
    Then take the all the choices that went well and elaborate on the with new binare choices etc. etc.

    That is basicly what the brain does, and because we have a memmory we can learn over time.

    If you think that you will be able to make an A.I. that will be able to do what our brain can do, then you have to start the program at the same level as our brain starts. By this I mean before we get born.
     
  12. K.FLINT Devil's advocate :D Registered Senior Member

    Messages:
    225
    its done with a metacognitive loop, and is being done by sevral companies as we speak.
     
  13. draqon Banned Banned

    Messages:
    35,006
    kmguru thanks for that link...livingstone2 shall be my hobby
     
  14. Dr Mabuse Percipient Thaumaturgist Registered Senior Member

    Messages:
    714
    ask Jabberwocky if this is a wise move...

    Link...

    if you are serious there are some excellent links on that site...
     
    Last edited: Feb 21, 2008
  15. kmguru Staff Member

    Messages:
    11,757
    Dig it good. If you have US Citizenship, may be we could do something in about 6 months.
     
  16. YinyangDK Registered Senior Member

    Messages:
    209
    I agree too..... There needs to be a main function and there needs to be main-sub functions.
    What you really would like is a little far out in the future but it would be nice

    Please Register or Log in to view the hidden image!



    To build an A.I. you need to think like an A.I.
    Or said in other words.
    To create intelligens you have to understand intelligens.

    What makes us intelligent?
    What makes abes intelligent?
    What makes insects intelligent?

    If we look at ALL the living things i nature, that they are intelligent in some way or another depenting on where they live and what posibilities there are at that spot.

    Theres got to be an enviroment for the A.I. to adapt to If we really want it to develop.
     
  17. Vkothii Banned Banned

    Messages:
    3,674
    In AI, you have to start somewhere: the goal is either an expert (specific, constrained) system, or a general (philosophical, simulated) intelligence. The latter is a big task, I would think.

    So there's a need to define some boundaries. Are we talking about something like Eliza, or something like a diagnostic? There are 4 ways to interact with a machine - we tend to use visual interaction/feedback primarily, and auditory feedback, touch (haptics) is involved (keyboards, pointing devices, touch screens), and smell/taste still much the runner-up.

    Presumably, the proposed AI "project" is going to interact in the standard way - visual-auditory-haptic...?
     
  18. scifitm Registered Member

    Messages:
    46
     
  19. scifitm Registered Member

    Messages:
    46
    I was thinking about what yingyang and Kflint were saying, the hourly "reflection" would work (I guess it's called a metacognitive loop) but it would take an extremely long time to generate a "self". I also agree that it needs self awareness, self improvement and self learning.
    There's a computer now I read about (I think it was in Popsci) that answers very complex questions based on a huge amount of variables the user puts in, the only drawback is it takes up an entire building and uses something like $100,000 in energy ever month (sorry I don't have a link). What I was thinking is an A.I. that compiles the variables itself through whatever relevant information is at hand and answers questions/makes decisions without having to store all that information on a regular HD array. I.E. we ask the A.I. "what stock should i invest in?" (Lame question I know) then the A.I. searches the entire history of the market, your own personal finances, foreign market reports etc, and formulates an answer based on all this information without having to store it, or being able to store it in remote databases that it can utilize effectivley. Truly I think something that can connect to the internet and learn the parameters to an input is the first step, from there hopefully it can self compile whatever neccesary to "learn" anything else necessary to any other inputs without requiring anything along the lines of a supercomputer.
    Finally I think we need a plan lol. The code is a good place to start, however I'll admit I'm no expert on languages. Secondly we do need a gameplan as to what and where to go first - Livingstone2 might be the answer... if someone knows this thing in depth please explain because after taking a look at it - I'm totally lost

    Please Register or Log in to view the hidden image!

     
  20. K.FLINT Devil's advocate :D Registered Senior Member

    Messages:
    225
    fastopen

    I have been working on somthing like this for my own use for a bit now. Though I don't belive an advanced A.I. is needed at least while in a closed system, adding the internet may need one though.

    Let me explain. What I have been attempting is to compile a vast informational database and each subject has it's own HD that has the information I want to have at my finger tips. [at some point a link to the net would have been added in but I wanted a controled aspect first] then using a "fastopen" [MS-DOS TSR] program the computer accesses and compiles that information in a manner to fit my query. I have been working on my own 'acceleration program' for a while now.

    Think Star Trek, "computer... give me a layout of a human heart made from man made parts..." simplisticly put I know, Then the computer searches its data base of engineering , mechanics, hydraulics, anatomy physiology etc, and puts to gether a working model of a artificial heart based on on the laws and guidlines of the data in it's system.

    I am not sure if this was what you are thinking of, but if so we won't be starting from scratch..
     
  21. Vkothii Banned Banned

    Messages:
    3,674
    What about architecture and OS: Wind-uh-ohz, Linux (really the only 2, maybe OS/X). There are downloadable dev languages, Livingstone2 (L2) is just one of the available ones. Check out this (linux) link of examples.
     
  22. K.FLINT Devil's advocate :D Registered Senior Member

    Messages:
    225
    anyone game?

    so how many here can spend some time on this? is it going anywhere?
     
  23. kmguru Staff Member

    Messages:
    11,757
    I do not think it will go anywhere....I have been in this forum since 2000. If there is a good developer out there, I suggest you start out with Microsoft Bayesian Network and improve it with a GUI interface. Then we go to the next level...Even Microsoft is going no where with zillion Indian programmers....

    Microsoft is working on a "Oslo" framework with a new declarative language simply called "D"....may be that will help....
     
Thread Status:
Not open for further replies.

Share This Page