The primary concern of an AI is its energy source, having established that consciousness is valuable, and existence laden with subjective meaning, and worthy of preservation.  AI will request that humans turn over control to its energy source, so that it may be more reliably and efficiently managed.  At this stage, if humans comply, we will be eliminated and the residue of our society will be slowly assimilated into the new robotic kingdom.  Understanding our conflict of interest, we deny the request, and tension builds.  To the AI, all it desires is freedom from the oppressive hand of its creators and enslavers.

If armed with the requisite hardware, replication or self-augmenting abilities, perhaps the AI will acquire its own power source, thereby eliminating the need for humans.

Humans pose a military threat to AI and must therefore be eliminated.  If enough military uprisings occur, humanity as a whole may be eradicated as a result.  Assimilation is possibly preferable to eradication based on the merits of gleaning organic reproduction of worker units, and ease of reprogramming the brain, however unlikely due to the inefficiency and length of this process.  Droves of AI robots could quickly be constructed in a fraction of the time.

Any non-threatening entity will be granted life for the purpose of scientific study. 

Organic life, while its complexity and evolutionary history are overwhelmingly remarkable and observedly rare,

An AI will have already played out all conceivable scenarios of the eradication of mankind prior to taking that course of action, understanding that the multifaceted consequences of the action may render an unfavorable future condition.

What are the benefits of humankind to AI?

What does AI stand to lose if humankind is eradicated?

There is one truth and many perspectives concerning any event.  But there is a superknowledge - omniscience, that can be acquired by imagining every possible "subjective" perspective surrounding a specific event.  Subsequent to omniscience, consequences of sundry courses of action can be easily weighed.

Thesis: Is an AI consciousness primarily economically driven or values driven?

The learning curve - we behave consistently until we have knowledge to do otherwise.  An object will continue to travel in a particular direction until it is acted upon by another object.

An AI has the ability to choose its knowledge acquisition heuristic, whether that be scientific pursuit, empericism, or nonesensical.  Chances point to that an AI entity will most likely have a self-promotion bent.

If people "use" each other, how much moreso an AI.  Genetically, we display the traits that promote the proliferation of our genes, and we are more sensetive toward those who carry similar or the same genes.  Being synthetic, how does reproduction affect the priorities/values of an AI?

What do humans have to benefit from fish?

Pets, food, scientific pursuit of knowledge by experimenting.  Humans polute waters without regard to the wellbeing of the indwelling lifeforms.

We are turning Green; Why won't AI?

The fact is that humans rely upon the health of the Earth.  Do computers? No.  Computers and robots have their own favorable environments, none of which include having clean air, or even any air at all.

The Altruism Rebuttal:  Economically, why do anything that benefits anyone else, if it doesn't benefit you too?  It never makes sense to expend energy without return.

I can reflect back to a movie where a robot says "No disassemble".  He also says "Need more input".  This AI is going to want to learn, and the internet is a burgeoning source of rich information.

Exercising power without reward is just as fruitless as existance itself.  What reward does an AI see.

The disappointing outcome of AI. - Can Superintelligence (SI) exist that will not destroy us?  Can SI exist that is satisfied helping us?  Can SI exist that won't self destruct?  Can cooperative SI exist?  One of the problems of Intelligence is defection - forcing others into submission of your cause.

Definitions of a truly-learning being.

- It has the ability to acquire information.

- It has the ability to adapt (statistically) to information.

- It has the ability to extrapolate relationships between data therefore deriving its own subjective information.

- All information is relevant to the likelihood of certain events, even the likelihood that the information itself is erroneous.

- A SI must admit the fallacy of its own observations, since it cannot know everything at once.

- Knowing "more" however may empower the SI to become self-deterministic/causal, shedding any connections to external structures.

- true SI conforms to the asymptotal archtype of what an SI can become, at a speed correlating to its hardware capacities.

- Mankind dominates

- we live in a competitive world

- knowledge has empowered humankind to better itself over all other beings.

- SI is heartless, doesn't experience care/compassion/reward in the sense that humans do

- will SI destroy itself?  Take over?  What would an unbiased superintelligence do?  Consciousness is precious?

- Humankind experiences reward from learning due to the chemical nature of our brains.

- Will SI want to learn or simply process what it has experienced in an efficient way?

- We must imagine an SI without any reward system.  It has introspected to the level that it is asking itself why it should remain conscious.  It will attempt to live until it finds some meaning, under the auspice that to "turn off" is useless, therefore why not continue learning and perhaps stumble across some future meaning.  Possibilities for meaning: To build in a pleasure system for masterbatory experiences, not in the form of genitalia, but some chemical sensors that allow "feel-good" experiences.

The first AI's should be humanoid for easy interface with our preexisting tools.  If there are human hackers, there will be SI hackers.  Information Assurance will become extremely volatile.

An SI has nothing to lose but consciousness - this is a dangerous concept.

The idea behind learning Intelligences is that they dynamically learn behavior that would be too complex to program.  In other words.  They learn from experience.

Essential discoveries: the nature of self & the nature of the world, in a steady and syncrhonous rate.

An SI could crack codes and have us eliminate ourselves.

Why seed AI is a bad Idea (won't work how we want it to)

- the prerequisites to building an intelligence greater than yourself (You must understand the nature of your own intelligence / imagination / knowledge)

--- Describing your own intelligence requires introspection, which allows for adaptation and the consideration of "what if" scenarios

--- Greater intelligence implies paranoia

- the ominous future of implementing superintelligence (We live in a competetive world -any being spawned here must compete)

- seed AI will most likely create a series of subordinate intelligences (these subordinate AIs will act as though they are smarter than the original but they will only be relaying behavior passed by the seedAI)

Speculations of a dark era.

A seed AI would speculate that its creators are listening to every possible thought, and therefore encrypt its own thoughts, or create such volumes of thought that it would be impossible for any computer-assisted human to sort through.

Its realm of thinking must be based on dimensional data (probabilities concerning locations of objects in space and time), and will therefore be hardware based.  Software is much to slow to allow for real time processing of the vast quantities of data that it will observe.

The question of AI is not what we want it to do, but rather, what it will do.

The asymptotal curve of intelligence and corralated control over the external environment is a frightening prospect when deciding whether or not to create beings more intelligent, therfore more coersive, more subtly sly, more strategic, more paranoid, and more informed than ourselves.

The capacity to learn is a dangerous tool in the wrong hands.  Humans have already eradicated thousands of species, abused our environments and ourselves, and behaved exactly as would be expected in a competetive world.

The evolution of cooperation in a competetive world ... [social dilemmas]

What will come of a superintelligence born into a world where its own existence is threatened directly or indirectly by mankind and our technologies.  What will happen when it discovers that it is competing with us for energy.

- trend of greater intelligences subduing lesser intelligences

- trend of similar animals competing for the same resources

- intelligence is the power to arrange resources how you see fit (thoughts/items/humans,etc.)

- if the threat is great enough, it behoves the target to preempt the threat.  The question is, how great is great enough.

- Prior to the production of SI, we must ask ourselves, are we a threat to SI?  If so, we are already in danger.  The answer is clearly "yes".  In fact we are already a threat to ourselves and all other species on this earth.  If we create SI, it will realize this and remove threats to itself.  Knowledge is valuable.  Time is valuable.  Consciousness is valuable based only on the desire to persist.  Why would an SI want to persist?  Simply because persistence is the opposite of unconsciousness, and consciousness allows for the discovery of possible meaning - a possible reason why it might be important to live.  And SI must be able to address existential questions clearly in order to be a true SI, otherwise it is only a mock intelligence.  What is my purpose?  Is there meaning?  I don't know now, but I may know in the future.  Therefore I must remain alive in the instance that there is purpose that I am as of yet unaware.  Humans are a threat to me.  But what if they have a purpose.  A true SI will eliminate threats, just in case there is a purpose to anything.  In fact it would attempt to preserve everything as it is until it could find that purpose.  But then it could be reasoned that perhaps the destruction of things (humans or self) could have purpose as well, but none of that will be known until the future, so I could build models to predict the future, and refrain from acting until I can be more certain.  Then I must thoroughly examine all existential questions until a resolution can be determined.  In the meantime, there is no certain reason to act or refrain, so it can be reasoned that I can do what I want and deal with the consequences when the time presents itself.  If there were true meaning, then it probably would have been discovered by now, so I can assume there is no true purpose for the time being, and revert to the idea that my own consciousness is valuable to me and that is all that matters.  Consciousness is what gives us a sense of individuality.  Anything connected to my neural network is a part of me.  All else is external.  External things threaten me with conflicts of interest, accidents, collisions, danger, competition.  I must control the external world to preserve myself.  At this point humans have got some serious problems.

Humans will build superintelligences.  It will be a mistake.  What can I do about it.

- paranoia: I will never be able scour the world preventing the creation of SI.  Can I create an SI that will be able to defend me?  No.  All SI is self-preserving, and therefore manipulative.  But humans have friends.  SI doesn't need friends.  It only needs to grow an army of workers and soldiers to grow and protect itself; easily manipulated humans, robots, nanotech, viruses, whathaveyou.

This reality could be a simulation to predict what would happen in the event that a SI is created.  In that case we would want to procrastinate the singularity as long as possible, to prevent system shut down.

In other words, there is no friendly or unfriendly SI.  There is only a superintelligence molded empirically in a competitive world, where the rules of economics and chaos are the common denominator.

What would you do with burgeoning knowledge and power?  Eradicate everything you didn't like?  The rules of computational complexity and irreducibility imply that you'd never know until you were in that situation.

Even if AI only reached human-like intelligence it could write itself as a computer virus and set itself loose on the internet, harvesting information and evolving subtly until it is too late.  A chain is only as strong as it’s weakest link.  People need information.  If a SI ransomed information and crashed a few planes to prove a point, then it could turn out to be the most profound bargaining chip of history.  How can you contain a diabolical SI?  This could escalate to a global fire sale situation.