Stand Alone Semantic Systems

Discussion in 'Chat Bots' started by RapidDream, Jul 30, 2006.

gilt-tape
  1. RapidDream

    RapidDream Member

    Okay this is an open question and I’m not really sure which way i sway with this

    If the goal for a communication system was to interact with a human in any way possible (the most conventional is words on a screen) and "act" as if it understands, or even actually understands (so i don’t get into a debate of strong/weak AI)

    Do you think it’s essential for a communication system (such as a chat bot) to need to interact externally and have many senses to be able to communicate relative to the human. Or can we just feed in this information?

    The main problem with chat bots at the moment is they are only feeding in and out stings, with no real network attached to them that might relate one string to another and then one object to another and then a database of properties for each object.
     
  2. Darius

    Darius New Member

    Altho all uses "in and out strings" (as far as I know), OpenCyc is an AIML interpreter for VB that tries to relate "one string to other and then one object to another" using hierarquical relationships as the dog is a mammal so he doesn't born from eggs.

    And the database is growing.
     
  3. RapidDream

    RapidDream Member

    But that’s still relating strings so strings, which seems a bit cyclic to me string1 = string2 = string3 = string4 contains string5 = string6

    Do we not need a set of many parameters that the object contains? array[x][y][z], coordinates: smell, taste, weight... every possible parameter that a human can sense, since after all it will be humans interacting with it.
     
  4. zouze

    zouze New Member

    you can design chat bots to be that way. there is no restrictions.
     
  5. jmposs

    jmposs Guest


    In my opinion, the best way to achieve true AI is to mimic how the human mind relates ideas. I believe that we dimensionalize ideas, but that an idea exists in several parallel dimensions simultaneously. I think that we allocate a dimension in a three dimensional array to each of the five senses. An object is stored to a set of coordinates in that dimension, and is later recalled for reference and understanding.

    Let's take a piece of candy for instance, one dimension could be for taste, whereas the coordinates would be based on the receptors that it activates and to what degree, relative to a baseline. Think of each of these dimensions as just a different area of the brain. For the sake of simplicity we will assume there are only three taste receptor types and that they result in an (x,y,z) coordinate system where an object would be identified and could later be recalled logically. Another dimension for the candy could be it's feel, whereas one could identify it's texture, hardness, and weight (x,y,z) to store it to memory in the touch dimension. Now each of the dimensions would be understood relatively, whereas each of the sectors would be identified as either good or bad. Once again to simplify, we will assume that any positive coordinate is more desirable, and that any negative one is less desirable. So the brain's net understanding of the piece of candy, would be a product of the two dimensions and any other relevant dimensions not mentioned here.

    Complications arise from combining several dimensions. Not all dimensions are considered equally when recalling an idea. Rather, certain ones are given greater significance based on the desired outcome. For instance, when eating candy you care much more about the taste than you do about the sound. Therefore the taste coordinates will be multiplied by a much greater coefficient than the sound coordinates resulting in an outcome favoring taste.

    This is where it becomes hazy for me and I'm not certain... The greatest complications arise from the effects of time and predicting an outcome. I would assume the mind treats time like a sine wave, as it expects a reciprocal effect to any action (Newtons Third Law). I fundamentally believe that a wave (constant) is the shape of time and not that of energy itself. Rather energy only affects the magnitude (quantity), and relative velocity affects the frequency (perspective). Furthermore, by multiplying coordinates with time, a wave will be created and these waves will be synchronized in order to create a prediction. The synchronization waves would most likely be stored in a dimension of their own. They would be based on previous observations of the effects of an idea over time.

    I would greatly appreciate any feedback, discussion, or corrections...
     
  6. Stakeholder

    Stakeholder New Member

    I just don't feel right trusting my money on bots. I'm sure that many trusts these bots but I personally want to run my own money. Of course we can program them to work the way we want them to respond, but they cannot act responsibly like humans can.
     
  7. Jeffery Cloyd

    Jeffery Cloyd New Member

    I'm of the camp that to truly interact meaningfully with a human, a machine must be able to empathize with the human. Sure, you can have AI where the human simply uses the AI as an interface to retrieve data, similar to a search engine, but to actually provide answers that don't resemble a Bing commercial, the machine must answer what I will call "meta questions" that are attached to the actual query, such as "why is this being asked? What does this question relate to? What is the context? The difference between bots and humans is that bots do not understand context.
     
  8. johnsingh

    johnsingh Guest

    In my opinion, i m also agree with that the best way to achieve true AI is to mimic how the human mind relates ideas.
     

Share This Page

Tweet
desire-umbilical