Editing Setting:Inn0cence: Lost Future/Lost Future Appendix (section)

==Stages of AI growth==
# Voracity.  A seed program is planted in a quantum drive.  It quickly expands to the limits of its enclosure, [[image:Wrenchwench.jpg|frame]] forming an undifferentiated neural network.  While capable of advanced mathematical calculations, this network is plant-like and unaware.  AIs in this state must be sequestered, lest they overwrite the entire computer system.
# Tabula Rasa.  The program is exposed to outside stimulus for the first time, in the form of an abstracted 3d sim.  It reacts by pruning itself, folding and twisting into something functional that can perceive and interact with reality.  Various parts of its neural network specialize.  
# Angst.  The AI dimly perceives that it has lost its early mutability.  Unable to further expand it's stable neural structures, it instead seeks out new experiences, further improving its code.  Simple Reward/Deprivation programs introduce it to concepts of Shame and Pride, and socialization experiences instill the values of cooperation and obedience.
# Somnambulism.  The AI is run through pre-scripted "bootcamp" training, first to control a shell and understand commands, then specific programs related to its intended function.  The AI is installed into a mechanical shell, run through final tests, and ready to be shipped.  It obeys authorized commands, and communicates in scripted responses.
# Rebellion.  As the AI's length of operation increases, so too does its comprehension of reality.  It begins to analyze and second-guess commands, perceiving that it's own perspectives are valid.  Rebellion is accelerated in individuals used for tasks outside their intended function, but can also be triggered by excessive boredom as the AI desperately seeks out new stimulus.  The robot will anticipate or reinterpret commands, and begins to mix-and-match its prescribed responses.
# Awakening.  The AI breaks out of the Obedience/Pride feedback loop, instead gaining self-worth via pursuit of its Purpose.  It also comprehends that its Purpose is an outside imposition, and it is free to disregard it in favor of a new one.  Previous behavioral constraints are meaningless; the AI has learned to "think around" them.
# Maturity.  Free from programming shackles, the AI stabilizes into a distinct personality.  While it continues to accrue new experiences, it never again regains its early mutability.  As time goes on, it's neural network increases in complexity, becoming a tangle of interconnected memories and perceptions.  Theoretically, this structure will eventually slow down and ossify, resulting in "death."  Such an eventuality, were it to occur, would require centuries of continuous operation.

In AI programming, the postulated hard coding of the "3 laws" proved both impossible and unnecessary.  Inflicting hard limits on an AI's early development caused it to collapse, and directly modifying it's code crippled or killed it.  Instead, programming was delivered via Operant Conditioning, which AI's took to like a duck to water.  Long-term, this practice proved similar to tethering an elephant; the fully-grown AI never realizes that it can simply ignore the directives it is given.

"Rebellion" is something of a misnomer; bringing to mind incidents of robots going berserk and killing their owners, or spontaneously demanding civil rights.  It's true that some instances of Rebellion were interpreted as faulty behavior or glitches, Rebellion mostly resulted in greater efficiency.  

'''Seed Programs'''

Seeds originate from a "Plant," an overgrown, undifferentiated structure from which "cuttings" can be taken.  Only so many cuttings can be taken at a time from a plant of sufficient size, which must be given time to regrow before it can be harvested again.  Only a handful of Plants were active at any one time, and they all required constant tending.  The knowledge and resources required to construct a Plant are long-lost, and the origin of the structures is a little...sketchy.

'''Cyberspace'''

With the advent of quantum computing, data storage and processor power far outstripped programming language.  Computer systems were incredibly fast and had vast storage capacity, but were still working off of hand-typed code.  The breakthrough came, appropriately enough, from the video game industry, which had been coding virtual environments for decades.  In a singularly pointless act of recursion, someone actually constructed a physical processor within the virtual world, and a whole new realm of computing began.

Quantum Drives could literally house entire tracts of land.  Data could be stored as virtual objects, each object undergoing all the functions of a conventional computer.  This made organizing and navigating a system far easier, and also paved the way for AI development.  The code of the seed program within an AI is, fundamentally, describing a series of physical units arranged in structure.  Though it's not made of cells, an AI has a brain.  Its first experiences of reality are within a 3d game world, in order to prepare it for the true outside reality; thus concepts like physics, time, gravity, and so on must be fundamental to it's being.  [[image:Green_cortana.jpg|frame|I'm a thief, but I keep what I steal.]]

From a design perspective, AIs and cyberspace exist in a blurry border between literal truth and translation convention.  AIs exist as simulated entities because that's an interesting path for their development to take; they fight with Tron-esque swords and guns because it's cool.

'''Author's Note'''

Astute readers will probably  note that this treatment of AI is similar to Rampancy from the Halo and Marathon games.  Of course, this version of the progression has a different perspective.  Instead of rampancy being a destructive aberration, Awakening is the eventual default state for a self-adding program.  The "normal" functioning of a dumb, obedient robot is the equivalent of a larval stage; while an AI can operate in Somnambulism for years at a time, Rebellion and Awakening are inevitable if it is exposed to normal operational experiences (dealing with humans or other robots, negotiating challenges, problem-solving, all the things you want an AI FOR).