sensory/action encoding/compiler added for improved performance and genetic activation/deactivation

[7yl4r]

just as how real cells will activate or deactivate genetic sequences, so could these cells. This would require a more detailed modeling of processes within the cell, however. A message initiated from some sensor (vision or chemosensor) would produce an action encoding, which would be processed by some mechanism.

Sure, we could replicate the RNA process, but that is probably terribly inefficient to handle without using many processors. In this case there is a dictionary which maps sensors to encodings, and another dictionary which maps encodings to actions. By allowing one (or both) dictionaries to vary genetically, various sensory responses will trigger various actions. That's a whole lot of dictionary searching and compiling.

Alternatively, this could be modeled as a weighted Bayesian network, but that too seems computationally heavy. The vision-movement network is already slowing down code significantly. Also, the number of possible connections could get huge.

Perhaps it is useful to think of the encoding messages as program instructions to be read by an interpreter. So... The encoding produced by a sensory input is genetically determined. This encoding is the program instructions to be read by a (static across all cells) interpreter. Example: chemical1 detected, genetic makeup of this cell dictates that action(s) for chemical1 are: {[0,12,15], [1,74,3]} The interpreter reads this as {deactivate genes 12 through 15, activate genes 3 through 74}. Bam, done. As a matter of fact, this approach might speed up movement processing too.

Now then, back to what I was saying. Activation/deactivation can lead to interesting behavior. For instance cellular free will ("Bacterial decision making" Jané Kondev, February 2014)