Dear Vendors
We have decided to undertake a project to understand life and develop programmatic models around it. We hope such models will not only benefit the AI but will be of good commercial value to the world too. The model will essential consist of a "world" and "forms".
Expressing "life" needs a form because a form has constraints. A form without constraints is free energy with no rules. This would essentially leave the team clueless from where to begin. Besides we also believe that pure energy with no constraints is not capable of experiencing or observing, is not capable of "consciousness". Only constraints give a frame of reference. It is something like in real world gravity defines "up" and "down", our body constraints define "left" and "right", a society defines "right" and "wrong". One can argue that for a zero-gravity place there would be no up or down, for an amoeba there can be no left or right and what is right in one society may as well be wrong in the other. However, we observe, that constraints limit a system and make it easier to program. Constraints should limit the perception but they will also make our models practical, developable and sustainable.
Another pragmatic reason for introducing constraints is that data-processing for life like systems(the forms) becomes extremely difficult and slow for the processing system if it is not ruled by constraints. The main system gets bogged down by too many details. This brings us to the second point of the proposal : probable design for forms. The different forms in the model may be considered separate systems made of smaller subsystems. But the functioning of the subsystems should be essentially decoupled from the main processing system. This means that the working of the subsystems should be essentially invisible to the main system until there is an event worth the time of the main system (like an exception thrown..). We assert that the "Observable" pattern may be pretty useful here. Every task and subsystem may hence have an independent algorithm with which the main system cannot affect to large extent. The data-exchange between them is also restricted. This should have significant advantages in terms of data-handling. One disadvantage that may arise out of it is that the form, despite of all the information, may be essentially unaware of itself.
Since the forms will need to interact, there should be a set of rules for that too, the implementation of which is left to the team. This will define "senses" of the form. Functions like death should be implemented if forms can replicate. However we do not want loss of experience and would like it to persist in some way in the forms replacing the "dead" form. May be we can recycle the energy as the concept of "re-birth".
The forms interact with each other and the "world" by the rules of the model. We propose two models: Evolution Model and God Model. In Evolution model we introduce very simple forms, the rules and just watch the emergent behavior. The forms may combine, learn, change, "evolve". In God model we introduce advanced forms as and when necessary and interfere if system starts becoming highly chaotic. We also drive the forms to a "purpose" for better understanding of directing emergent behavior of bots.
We expect you will start by creating simple worlds and complicate them as you gain experience.
Your Truly
Orion Inc.
PS: Last Minute Request: Once we determine what behavior is going to generate revenue, we would like you to implant bots with a "super-evolution" algorithm. Essentially after crossing a threshold we expect the bots to increase in processing power and observation. So to say the interaction between the subsystems and main system may become more powerful after increase in processing power. If the bot starts showing signs of intelligence and self-awareness, we would like to observe the results.
