My take is based on the subcortex, a simple set of general behaviors and goals that are shaped/enhanced by the the cortex.
Looking to see how the brain does this - the cortex is arranged as a collection of about 100 discrete modules that are connected together with various fiber tracts and sub modules such as the thalamus and hippocampus.
These tracts act to create the hierarchy and coordinate the 3 major processing hubs. The fourth hub is a special visual processing sub-system.
The thalamus also acts with the surprise/attention (bursting) features of HTM to do signal routing.
The various subcortical nexus project both direct commands and shaping neurotransmitters to configure learning and recall as part of behavior selection.
A buffer module (the hippocampus) collects episodes for salience marking and transfer/reinforcement to the cortex.
The subcortex has a collection of hard-coded motor subroutines such as visual scanning that act to feed the cortex with with sensation to be evaluated and processed. Some of these hard-coded routines follow the top-down paths to filter inputs and to act to interrogate the cortex for recall and pattern completion.
A sequencing module (the cerebellum) collects and coordinates frequently used motor sequences such as body movements, talking, and thinking, to automate them for more rapid and fluid execution.
There is a general pathway from sensory cortex to subcortex to motor cortex back to sensory cortex that acts to create the phenomenon generally called awareness or consciousness. This particular bit encompasses most of the brain in a collective action so the characteristics are complex and difficult to define precisely. The complex action of these elements of both cortical (conscious) and subcortical (unconscious) components contribute to the notorious difficulty in defining what it is and how it works.
Error can’t be less than 0. Current models are passive, you feed them a bunch data, they fit to it. Open-ended model should be actively exploring environment to increase its predictive power. Think interpolation vs. explicit extrapolation to implement novelty seeking:
quantify (internal_match - match_to_the_model) in the next / proximate location of data, then skip it if that value is below model’s average. That is, if projected additive correspondence of model to the environment is below opportunity cost.
Match is a complementary to error: if the error is 20%, then match is 80%. But in absolute terms, this extrapolated match is directly additive: it will grow as the model explores the environment, whatever that is.
Indeed that’s mathematically impossible. But the loss doesn’t have to be 0 to be highly accurate anyways.
That’s pretty much what all Reinforcement Learning algorithms do.
What.
I’ll be frank here - none of your language makes even the remotest sense. It just sounds like you’re presenting a random salad of words thrown together. I’ve no idea where you’re going from … that.
Point 1 is an implementation detail. Ignored.
Point 2 is fine without the ‘world’. Most models are local or deal with tiny parts of reality, and there is no requirement for overlapping models to be consistent.
Point 3: please note this is reality and the model is changing of its own accord. First we have to predict the next state, then we have to choose an action to achieve a goal with respect to a future (predicted) state.
Eg watch the ball flying through the air; predict where it will be; run to that place; catch the ball.
watch ball flying towards air → desire a world where you have caght the ball → infer that to catch a ball implies moving your arm → generate a world with your arm correctly positioned via common-sense physics-> compute the difference between your current arm state and desired arm state → turn the difference into low-level muscle activations.
In peri-personal space, you already have stored endpoint transformations.
See a spot and modify your stance to park a hand or foot there. This is part of why people protect their personal space.
Im arguing that animals ARE qualia, as soon as you see them they become a sight, as soon as you know of them they become knowledge. Im not an idealist by any means. That has many contradictions, so i have faith that animals are not qualia…even though not qualia is a concept itslef so if they are not qualia they still have to be concepts of some sort…leading us to the reason why its some sort of irrational faith in reality that i have
But knowledge is conceptual, that is the illusion so the former is equally right. Simply put what is the meaning of “objective external reality”…if we say meaning only happens in minds then what did objective external reality mean before minds existed…if it existed before there were minds, what are you really reffering to when you talk of something meaningless. It could be a flying pig , infact it couldnt because flying pig means something…it actually means if you take this seriously that there was nothing before minds, and even then nothing has meaning…so we are stuck in a loop…the best we can do if we are honest is to have some sort of irrational faith in reality since there has to be minds before minds existed to make “nothing” have meaning which is a direct logical contradiction…so we have just to have that irrational faith as a presupposition even to work or to move
Here’s the tie-up:
Nietzsche: …too complex to outline here.
Wittgenstein: Language Games (everything ‘human’ is a result of language, including consciousness).
Jaynes: Only humans are conscious, it is learned and it requires language (syntactic & recursive).
Hameroff: qualia (awake and aware) is experienced by all animals with µtubule nervous systems.
We don’t have to worry about hurting an AGI-machine’s feelings but we do have to worry about how humans will put it to use, because:
We humans are uniquely able to sooth ourselves and avoid self-defeating/crushing distress by developing, or adopting ready-made, secular as well as theistic ideas.
However, theistic ideation is the kind of ideation that is most obviously sprouted and insidiously fueled from a CURSES that was incurred when in very many human infancies and early childhoods the most fundamental of all our social needs went too chronically unmet and the resulting distress therefore became automatically blocked and indefinitely postponed.
We humans are the result of the only lineage of simians that became (naturally ‘endowed with’) EAVASIVE.
EAVASIVE is an ÆPT acronym pragmatically extracted from “Evolved Ambiadvantgeous{ly adaptive} Verbal Attention {or “actention” selection serving} System Involving Various Endoopiates”.
This, our unique recent phylogeny, has painted us into a corner in which we are largely incapable of understanding the dire consequences of CURSES and of minimising (as far a realistically possible) the extent to which predicaments or ordeals (for any reasons ended up under threatening circumstances) of ”specific hibernation” imploring type cause (’put’) CURSES.
“CURSES” (the same allusive ÆPT spelling applies in singular as in plural) is derivable from: commonly Conditioned-in commonly chronically kept Unconscious signaling State/s Effecting Symptoms (EAVASIVE and somatic such).
P.S.
My pompously named website ”aeimcinternetional.org” is not an organisation and not just a joke even though jocularity is built into the texts it contains.
