Common basic knowledge for artificial intelligence algorithms
Completed on 05-Jul-2015 (19 days)

Project 6 in full-screenProject 6 in PDF
Basic ideas >
Defining principles

The proposed system might be considered as a huge multi-layered classification of AI information, where each input is defined on account of its adequacy with respect to a set of categorised features. Thus, the best way to adequately describe the Defining principles of this system is analysing these (recursive) classification processes.

For clarity purposes, I will define two different classification types:
  • Multi-layer Classification
    . Each element is individually defined on account of its association with different categories. This definition is performed in a layered fashion, where Layers are understood as ranked higher-level categories which are similar to enumerations/enums in a programming language. Such a ranking is expected to be based upon rules like "
    Layer 0
    will always be
    weight_0-1
    times more important than
    Layer 1
    " (equivalently to what is expected to also happen with all the other rankings, which will be analysed in the next section). For example: a car can be classified as
    vehicle
    (within the
    Layer Machines
    ),
    mobile
    (within the
    Layer Staticity
    ),
    contaminant
    (within the
    Layer Environment
    ), etc.
  • Inter-relationship Classification
    . Additionally to the aforementioned individual definition of each single element, relationships among different elements have also to be considered. That is: the previous classification can be associated with defining a word, and the current one with defining a concept. For example: according to this classification, a car can also be defined as
    "faster than a bicycle"
    and
    "slower than a plane"
    .
    As far as this second classification cannot be properly understood without knowing the information storage format, I will continue this analysis in the next section.

As already explained, a detailed description of the proposed system is outside the scope of this project. On the other hand, more specific indications seem required in order to adequately transmit the intended ideas; that's why I am including my preliminary thoughts about the highest-level Layers below these lines.
  • Functionality Layer (L0)
    . The information is classified on account of its functionality. Logically, a comprehensive enough list of major functionalities will have to be created (equivalently to what is expected to happen while defining all the other Layers). Sample functionalities:
    understanding
    ,
    movement
    or
    vision
    .
  • Speciality Layer (L1)
    . This second division is based upon the application scope of the given information; that is: useful for any kind of algorithm (e.g., differences between men and women) or only for certain ones (e.g., ability to recognise a very specific material). This Layer is expected to be formed by a common category and
    n
    additional specialised ones. Note that this common sub-Layer will get a special treatment on quite a few fronts, like security-related issues.
  • Security Layer (L2)
    . It accounts for the impact of (the eventual variation of) the given piece of information on other elements in the system. For example: the
    masculine
    /
    feminine
    differentiation has a much more relevant impact (i.e., a higher number of additional elements will be affected) than the
    bulldog
    /
    boxer
    one.