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
    Im in. The beauty of freelance science is that unlike microsoft we actually volunteer to do the work

    Please Register or Log in to view the hidden image!

    I have some limited experience with Linux and a setup running it. another setup with xp and a gaming laptop with Vista... im thinking Linux is the optimal choice. I belive this project can go somewhere, we just have to start or figure out where to start. Also probably shouldn't over complicate the thing just yet.
     
  2. Google AdSense Guest Advertisement



    to hide all adverts.
  3. K.FLINT Devil's advocate :D Registered Senior Member

    Messages:
    225
    Ok. so is my "fast open" type program kinda what you had in mind? if not could you spell things out clearly please.
     
  4. Google AdSense Guest Advertisement



    to hide all adverts.
  5. kmguru Staff Member

    Messages:
    11,757
    First find a base program that has some tools and structure to work on. Visual Basic is too cumbbersome, C#, C+ etc is the same - that is because you have to write a lot of classes to get anywhere.

    Some university people must have done some work...if we can find some basic program already develped...they can be improved.

    The program must have a rules engine and goals and cases structure.

    National Instruments has a program called LabView. It is a GUI based program. You can drop a module in the work sheet and setup input and output by connecting the modules with other modules. You need something like that to start.

    Then you need a data extration, transformation tool - again it can be small but GUI based. All the tools you need should be able to drag and drop in to the worksheet.

    The modules themselves have to have some basic structures - multiple input, output, timer, hold on condition etc. Say you have a neural model as a module. You can collect the data from the database (similar to human memory) and do a neural calculation and put the data back to a database or another module such as comparision module.

    These intermediate calculations will be like active data filters etc.

    The final steps would be action outputs. We will start out with Computer display as the output. Later one can add synthesized voice.

    The key is to collect/build all the modules, input, output, calculation, data manupulation etc and then we can work on each module to improve and add more modules etc.

    We need to define first a problem set or activity that this program will do. The NASA people, when they wrote for the Deep Space 1, their objective was what the probe will do without external command. But they had most parameters nailed down, so they worked on that problem space and adjusted the logic.

    Long ago, I worked on a Japanese program where you define a process with Start and End, but can have many processes and subprocesses. If the computer finds a "wait for" statement, it will go to next process while waiting for a condition. It was a cool program. That way, the whole process will be multi-tasking. It was so intuitive that, I was able to write a complex chemical batch reactor program straight from memory.

    Keep thinking....
     
  6. Google AdSense Guest Advertisement



    to hide all adverts.
  7. scifitm Registered Member

    Messages:
    46
    KFlint: short answer = yes, long answer = internet
    KMGuru: lead on bro, you seem to have the history down. The modules seem the best route I agree - let's see if we can get it to function with Flints arrays? - I suggest global warming as the topic

    Please Register or Log in to view the hidden image!

    lets vote on the topic (I just threw one out there). Guru direct us to the program if you know where it's lair be. Sourceforge will probably host our collective until Flint can ready his HDs for our foreign diplomatic envoy. Bueno?
     
Thread Status:
Not open for further replies.

Share This Page