SELF-REFERENTIAL LAYER


A corpus of questions asked of a bot and answers generated by the bot serves as an approximation of memory. By labeling this self corpus differently from the rest, I want to bootstrap the concept of self in the bot. I believe this will help in bootstrapping consciousness in bots. Using this corpus to retrain is not useful because answers in this corpus are exactly what the agent generates. But this corpus is useful in 2 contexts I can think of: differentiating the self from others through communication corpora and maintaining a memory of self.

Si <- f(Si-1, Si) i.e.subsequent state depends on previous state and another factor (possibly the next state itself). I am yet to explore this idea fully, but I believe Conway's Game of Life holds importance in its exploration. This is linked to non-deterministic Turing machine.








THE MATRIX