Conditional AI

1. The basic for this project is the concept that Intelligence, or rather the decision making part of Intelligence, can be broken down to a simple series of conditions.
2. These conditions are mostly dependant (condition parameters depend on results of previous condition), meaning that long term memory is also required.
3. There are many fallback conditions, which are consulted when encountering new situations.
4. The system must be able to learn new conditions, either by teaching it, or by way of self-learning from the results of the decisions, which requires the system to be able to get feedback about it's decisions.
5. in order for a decision to be made, a set of given parameters must be passed to the system, which will analyze them according to it's conditions DB.
6. The parameters can include actions done to the system(someone talking to it, someone hitting it, etc.) and/or events that happened around it (the light went off, someone hit someone else, etc.). they may include information about the surrounding environment, the object interacted with, and any other information.
7. It is up to the system to take the parameters and search for the most suitable condition that match those parameters, and start following the line of conditions until a decision is made. This is better clarified by an example:
   • Parameters given:
     1. you are in a room
     2. it is dark.
   • The conditions in the DB are as follows (pseudo code) :
     1. If you are in a room, then there should be a door, a window, and at least one light switch.
     2. If it is dark, then you need to make light.
     3. If you need to make light, and you know where a light switch is, then turn it on.
     4. If you need to make light, and you know there SHOULD be a light switch, then look for it.
   • The chain of conditions will be:
     

     STATE:	location=in a room	lighting=dark

     1. Condition 'a' is executed, since the location parameter matches it, and the state changes to reflect the result:

        STATE:	location=in a room	lighting=dark	possible_objects=door,window,light_switch

     2. Condition 'b' is executed, because it is dark:

        STATE:	location=in a room	lighting=dark	possible_objects=door,window,light_switch	desire=need to make light

     3. Condition 'c' cannot be executed yet, but condition 'd' matches the state, so it is executed, and an action is called for (look for...). The system then tells whatever called it, that it has reached a decision, and the decision is that it should look for a light switch. for our purpose, we'll assume the light switch is found, and the system is called again to make a decision, this time, with the parameter stating that the light switch is found:

        STATE:	location=in a room	lighting=dark	possible_objects=door,window	desire= need to make light	known_objects=light_switch

     4. Now, that we know where the light switch is, we can execute condition 'c', and that creates a decision to turn on the light switch. Viola! We have created light!!!