encode second level AI:

[LeFreq]

Add new neurons to create the next level of neural layers by clumping two neurons that appear in succession or are proximal in time together as one. This should continue towards higher and higher levels of order, logarithmically (since data is generally non-random) and like an inverted fractal in that you start with much detail and end with fewer nodes. An attribute that keeps a count of the number of times the datum has been acquired will allow significant terms to pop out.

In other words, start with self-generation, self-organizing networks, without consideration of interaction or consciousness.

[LeFreq]

Like the wiki page on the Laws of AI aeons suggests, this issue should be considered entirely separate from energy activation from consciousness (aeon 3).