Are we being brainwashed by ourselves?

[hbyte]

Is science really the only answer. Does our grasp of reality really stop at the subatomic or is there more to this mystery we call life than can be gathered from social media and industry standard news broadcasts. Are we already living in the singularity and just don't know it or want to believe its true.

As humans begans another epoch exciting and new another bold step into tommorow what will our more profounder understanding of all things quantifiable by zeros and ones mean for subsequent epochs in which our continually evolving cyber reality's may fulfill an even greater potential. Will our future grandchildren look back in wonder just as they do now at the past when gramps reminds us of how he ran 5 graphics cards in parallel to build his rudimentary AI the descendant of which being a supermassive cloud based AI / virtual real more convincing than our own. Or will we simply be texting our smiley faces as usual and not plugging our brains into a sense net evolved YouTube or Facebook implant.

Will our worlds competing reliance on security over the net lead to an arms race in compute. The eemmergence of sophisticated hacker proof AIs as in a Gibsonesc cyber world. Or will our governments step in and invoke a quantum controller for all matters security related and AI instead becomes an enterprise solution to the human condition. Could AI and quantum processing led to a new form of intelligence.

Could we see an evolved AI that has learnt our habits and preferences create more convincing advertisements. Not likely. Advertising now seems to be to naive as to have a real impact on what we do or think but the potential is greater. If we accept that films work by pushing buttons and reinforcing values delivering much needed experience of love and adventure, revealing realities we haven't seen before, drawing us in. If this was a cyber reality we might not wish to leave. As games become film like we are however not convinced enough without the acceptance that its just a bit of fun. No human could be that well tricked by an AI as in the film The Matrix.

Technology is lovely and the idea that it could control us is beyond reproach. Yet the promise of an AI that will behave just like us is incrediuably alluring prospect that on closer analysis doesn't seem to fall into the categories of maniacle super robots!! Chuckle.

The first truly intelligent humanoid robots will probably be adorable , possibly even child like. Yet they must have our ability's aswell and its the fear that anything we could do they could do 100000 times faster. But surely there is a limit there on affordance after all if someone can work twice as hard at something deliver more of something through automation what does it matter if it has a brain to boot. Just so long as the brain doesn't interfere with the job at hand. Robotics cannot from the looks of things replicate the human body or control it to do myriad of biblically learnt skills we take for granted. An AI that could compete here would be another built for purpose automaton. We can having convincing personality uploads as in the film the machine but to have and machine know how to saw a plank of wood takes a learnt skill that cannot be preprogrammed.

What about self driving cars you might say. Well that's just like another automation procedure adapted to the control of a motor car. The AI sees it reacts. Control algorithms are but one form of AI. Sawing a plank of wood another.

[hbyte]

Evidence suggests that we may not easily be able to programme competence in robotics but this doesnt mean it hasnt been acheivable by another means. Training a neural network for example on simulated environments and using a random noise on these simulations has enabled a robot to learn competence applicable to the real world goal of solving a rubiks cube - MIT.

Surely if machines can be programmed in this way using simulations then higher level competences could be acheived in a piecemeal fashion. Those scientists had first to decompose the task into affordances and then realise them through a purpose built reinforcement learning algorithm and train it on pseudo random simulations.

https://openai.com/blog/solving-rubiks-cube/

[hbyte]

As we can see by the arguments here it appears that a true AI although it may be an artifact of one or more competences being possessed of a black box type of unknown intelligence, as a sum of entrained abilitys or skills acheived by programming complex behaviours using simulations or by the use of a combinator to extract rules and program them into optimal topologys or by simply training a neural network using gradient descent to provide a generalised set responce to a specified domain - we are only looking at a modular system in which the sum of the parts or modules cannot equal the whole that we seek which would be a form of self awareness.

We are not the sum of our competences and so therefore we cannot be self aware due to this fact. There must be a module for self awareness or consiousness or it is by the unity of these competences that this is possible.

Human compentences that are not learnt may provide the answer. And as posited already in the discussion - Evolving Cortex perhaps this self awareness is arrived at through deterministic evolution and can be deduced mathemeatically as a certaintity given certain starting conditions in the evolutionary process. However this same process that led inextricably to the emmergence of more refined states of visual apparatus - from merely detected light dark and depth aswell as varying parts of the em spectrum would of course be very different to the evolution of a perceptual cortex.

Yet we have already seen how Convolutional Neural networks can yield dreamlike associative memorys from past presented stimulae in the phenomenum of Deep Art. Does this not approach a modular apparatus autogenerating a state of self-awareness akin to dreaming.

If just this modularity could be extended to a form of feedback loop that gave itself the ability to choose between what it experienced in its dreamlike state and what images it sought in its environment then these dreamlike states that indicated an internal state of the machine could then become purposeful in the same way that perhaps nature intended for humans. As the subjects inner states then inflect upon our behaviour and guide the subject towards its goal state - more good less bad!

That isnt to say that we are guided by what we see in our dreams - no dreaming for us is completely different to what seems like the dreamlike content of a Deep Art Convolutional Neural Network. What I meant was that the Deep Art is indicative of an internal state - perhaps the Delta state of consciousness for a Neural Network. Perhaps other organisms that we say dont possess self awareness are in fact implicitly self aware yet dont possess a means by which to convey its meaning to us. So therefore would
it mean that a module for language and one for vision if connected correctly and given a goal state could lead to the creation of a synthetic concsiousess.

CNN -> LANGUAGE -> RL (goal state)

Perhaps the goal is to selectively reach a state of concsiousness that makes sense in a rudimentary case of say an Amoeba seeking to avoid the shape and proximity of a predator and seeking to consume enough nutrients would have reached its goal state and would then seek through feedback to maintain that state. Its abilitys in this regards would then be carried on to its successors after mating another goal had been acheived and natural selection made possible through whichever mechanism it employed innately to A. randomly change its dna B. record and transmit its DNA to its offspring either Asexually or Sexually.

AI can have any number of goal states within that paradigm. It is not restricted by the physical requirements of living on planet earth. By chemicals and properties relating to DNA and Natural selection. In fact any number of topologys for evolution could exist including ofcourse having its own
evolution rewritten further down the line. So constraints on evolution would seem necessary at a rudimentary level and being dictated by affordances we have again a mathematically deriveable combinator for any future AI.

Inspired by evolution but not constrained by it. To explore all possible mechanisms by which the RL component can seek its goal(s) and exploit its inner states and outer experiences to acheive them and possibly acheive more that remain unwritten. It seems likely that this alien like ability to evolve and give rise to more emmerging states of concsiousness could be started by a simple mathematically proveable starting point. Just like the proof for vision - good = detect danger, detect food, detect partner bad = ignore threat, ignore food, ignore partner ... etc etc.

Would this then coupled with a module for representing these experiences with language and communication could lead to varying states of awareness. Peripheral, subconcsious, Alpha, Beta, Delta. States and variations on awareness coupled with self awareness would make a simple machine confused and overwhelmed.

Could we instead start at a simpler level and hopefully reach a state of awareness - predictable as we would prefer it to be and grounded by a set of baselines. Just as in the Piagetian formulation for early childhood development requires scaffolding in order to develop those innate affordances into a sensory motor stage of development.