Hey MagnetMan, All,
I think you jumped a few steps in the chain to self-awareness MM - Afterall - How can there be ‘selfishness’ with no ‘self’…? That would be like the rabbit pulling itself out of the hat, would it not…? Where’s the magician…? Where’s the evolutionary-driver…?
I’ve been thinking that a kernal of self-awareness arises automatically within any bottom-up learning system, biological or otherwise, to fine tune a neural or computational network. Self-awareness as a delcate tuning fork, opposed to the sledgehammer blows of evolution and its henchman - Death.
Evolution drives top-down (inflexible/concrete) neural nets by penalizing failure with death, or loss of procreative options - which amounts to the same thing.
Let’s exemplify with recognition systems - discerning one ‘face’ from another. Offspring that cannot easily fix and find their mothers’ faces are offspring not long for this world. Equally a mother who cannot differentiate between young may well feed the offspring of another to the neglect of her own. As long as there is a great population, the evolution of a network for recognition could be driven by the simple deaths of those who fail, leaving successive generations increasingly adept at recognition as they unknowingly ‘compete’.
This kind of neural net evolution requires no conscious awareness on the part of the evolvee. Why…?
Behaviour “recog” →
a:(Quality)good–> (Outcome)probably die/discontinue neural-net geneline.
b:(Quality)bad–>(Outcome) probably live/continue neural-net geneiline.
Why…? Because the qualification scale and outcome is reasonably black and white. You pass and live, fuck up and die. The quality of achievement falls within the range for the process of evolution to imediately ‘notice’ and penalize. The magician who tweaks the net, is external to the evolvee. The magician is starvation, predation, disease, temperature - whatever enviromental factors apply. ie: in a that (bad) variation of the network gets thrown out completely - whereas in b that (good) version of the net is ‘allowed’ to copy itself into the next generation and undergo another itteration of quality-control.
Learning by dying. Species-wide Learning by the ‘luck of the draw’ at the expense of the individual. Rather a harsh school:
“A+ - you get to live!!! B- BLAM.”
This unconscious ‘learning’ process produces networks that are extremely advanced… More advanced than any we can contstruct ourselves so far. However, and it’s a very big ‘however’… It’s wasteful. It’s slow to the nth degree. Imagine if Sony brought out PlayStation 2 after a 30-year interval… 
X-Box would win… 
Where does self-awareness come in…? Or rather more importantly ‘why…?’ What advantage does it give…?
Self-Awareness allows a neural net to learn, to become more proficient at providing a beneficial response to a given stimuli, without having to die. It allows individual “in-house” learning. This process of individual learning - producing individuals with at least a chance of becoming better, more skilful, faster etc. benefits the species in-toto. There is a secondary plus: An unconcrete net, which will learn, takes less space, allowing you to fit more ‘compressed’ programs into the same space. Important during birth, given the restrictions on birth canal width and cranial diameter. A human baby coming straight out of the chute with its brain fully wired up and fine-tuned would get stuck. Try squeezing the Mekon through Dan-Dare’s ass.
Think of learning as a process of ‘decompression’ for an extant skill/behavioural-programme.
How does awareness help…? By setting the criteria by which the success or failure of any particular attempt - any ‘run’ of the neural net - is judged.
Baby learning to walk#1: Self sets goal “I want to go there.” And designates where ‘there’ is. Reward: Pleasure (chemical change manifesting as percieved ‘positive’ change in emotional brain-state) Penalty: Frustration sorrow (chemical change manifesting as percieved ‘negative’ change in emotional brain-state) ie: penalty is unpleasent but SURVIVABLE.
=> Overall - baby learns to walk in about a year or so.
Baby learning to walk#2: Survival (evolution) sets goal “Get to mama before your brother” having already designated mama. Reward: Food/life (at expense of brother) Penalty: No food, and eventual death. ie: Penalty is NOT SURVIVABLE.
=> Baby learns to walk in 1,000,000 years.
I may be exaggerating.
Anyway, in summary:
- ‘Compressed’ learnable abilities give a lifeform an advantage over those lifeforms without.
- For a species to be able to fine tune an abillity or behaviour quickly, at the individual level, within a generation, would allow it to adapt its behavioural set to a wider set of enviromental factors, more quickly, than a ‘one-size fits all’ hardwired species.
- To allow an individual net to be tweaked, there must be penalties that are survivable, and rewards that are instant (and subjectively/self-referentially pleasurable - ie evolution doesn’t understand the concept of “I wanna sweetie” because evolution doesn’t know what ‘sweet’ means).
- To specify the criteria by which outcomes are judged, and so judge attempts, without ‘waiting’ for the ultimate judgement of external evolutionary/enviromental factors - there must be self-awareness, and awareness of external conditions.
ie: Self-awareness arises from the (evolutionary) advantage of being able to learn quickly, on an individual basis.
ie: Bambi is to an extent… Self-aware.
Tab.