On 2/19/2018 12:27 PM, 'Chris de Morsella' via Everything List wrote:



On Mon, Feb 19, 2018 at 3:56 AM, Lawrence Crowell<goldenfiel...@gmail.com> 
wrote:On Sunday, February 18, 2018 at 10:00:24 PM UTC-6, Brent wrote:


On 2/18/2018 6:26 PM, Lawrence Crowell wrote:

Computers such as AlphaGo have complex algorithms for taking the rules of a 
game like chess and running through long Markov chains of game events to 
increase their data base for playing the game. There is not really anything 
about "knowing something" going on here. There is a lot of hype over AI these 
days, but I suspect a lot of this is meant to beguile people. I do suspect in 
time we will interact with AI as if it were intelligent and conscious. The 
really big changer though I think will be the neural-cyber interlink that will 
put brains as the primary internet nodes.

Why would you suppose that when electronics have a signal speed ten million 
times faster than neurons?  Presently neurons have an advantage in connection 
density and power dissipation; but I see no reason they can hold that advantage.

Brent


I think it may come down to computers that obey the Church-Turing thesis, which 
is finite and bounded. Hofstadter's book Godel Escher Bach has a chapter Bloop, 
Floop, Gloop where the Bloop means bounded loop or a halting program on a 
Turing machine. Biology however is not Bloop, but is rather a web of processors 
that are more Floop, or free loop. The busy beaver algorithm is such a case, 
which grows in complexity with each step. The computation of many fractals is 
this as well, where the Mandelbrot set with each iteration on a certain scale 
needs refinement to another floating point precision and thus grows in huge 
complexity. These of course in practice halting because the programmer puts in 
by hand a stop. These are recursively enumerable, and their complement in a set 
theoretic sense are Godel loops or Gloop. For machines to have properties at 
least parallel to conscious behavior we really have to be running in at least 
Floop and maybe into Gloop.
LC
Not sure if this has been touched on in this thread but it seems to me that the 
emergent phenomenon of both self-awareness and consciousness depend on 
information hiding in some fundamental way. Both our self awareness and our 
conscious minds, which from our incomplete perspective seem to be innate and 
ever present (at least when we are awake) are themselves the emergent outcomes 
of a vast amount of neural networked activities that is exquisitely hidden from 
us. We are unaware of the Genesis of our own awareness. 
Evidence from MRI scans supports this conclusion that before we are aware of 
being aware of some objectively measurable external event, or before we 
experience having a thought, that the almost one hundred billion neurons 
crammed into our highly folded cortexual pizza pie stuffed inside our skulls 
have been very busy and chatty indeed. As the MRI scans indicate.
We are aware of being aware and we experience conscious existence, but the 
process by which both our conscious experience and our own awareness of being 
arises within our minds is largely hidden from us. I think it is a fair and 
reasonable question to ask: Is information hiding a necessary an integral 
aspect of processes through which self-awareness and consciousness arise?


I think information hiding is looking at it the wrong way around.  It would 
take more layers of neurons to record and interpret the neurons responsible for 
your thoughts...a total waste from an evolutionary viewpoint.  Taking my 
favorite example of designing an AI Mars Rover, one provides for internal 
monitoring of some systems, e.g. power, temperature, etc.  But what would you 
provide to monitor the computer(s) themselves.  In principle you could record 
every step, but what would you do with it?  If you have mulitple computers on 
board (as is likely) you'd just take one that was doing funny things off 
line...but you don't have to record everything to identify a computer that's 
flaky.  All you need is some error correction and majority voting.  So there's 
just no practical reason to try to "un-hide" all that information processing at 
the cost of a lot more information processing.
Completely agree with aserting that attempting to record, categorize and make 
available an exhaustive and fine grained record of the full spectrum of 
activities, at least in any system that could be considered as being a viable 
candidate for consciousness or self awareness -- and I suspect we agree on this 
-- that it  is a rather hopeless task, at least in any non trivial large scale 
system. As a kind of aside, that being said, raw telemetry however, especially 
numeric metadata, is still valuable for forensic reconstruction of some given 
scenario based on the ingested stream's trail of time graphed signals recorded 
and reposited in the telemetry data store. Say for example when all hell breaks 
loose and all the alarming starts going off.
I also agree that quorum based algorithms are an efficient and especially when 
they have a large enough number of semi-independent nodes being involved in the 
active current quorum network it can be both an efficient and also pretty 
robust decisional system that is far superior to the heuristic stacks of the 
1980s style top down expert system approach to the hard problem of AI (which 
suffer from terminal brittleness and an inability to elegantly handle 
exceptions without resorting to cumbersome ad hoc exception lists that may or 
also may not provide effective coverage.
The manner in which I was attempting to convey the concept of information 
hiding however is not so much in the sense of lossy compression or the many 
other actions like say branch pruning, rollups etc. that deal with the very 
real computational issue imposed by unmanageable workloads, resulting from a 
geometrically growing data explosion and the all manner of follow on processing 
artifacts that are themselves the outcome of the algorithmic stream processing 
activity that is going on at any given moment.
I was more intending it being necessary as an enabler of the emergence of non 
linear processes in which while both the inputs and the final output may both 
be measurable and hence knowable, but the inner workings remain instead, to 
some degree hidden. Not in order to overcome and be able to handle an 
impossible workload, by throwing stuff out, but rather as a kind of firewall 
that helps the non linearity itself become manifest. I am thinking of 
information hiding as being a kind of catalyst, metaphorically speaking that 
helps non linearity to emerge out of a many layered stack of networked 
processes ingesting and interpreting a raw incoming stream. Perhaps then it 
would be more accurate to say that information hiding is the means by which non 
linearity is introduced into the event processing stream


In computer science the rather recent emergence of deep mind neural networks 
that are characterized by having many layers, of which only the input layer and 
output layer of neurons are directly measurable, while conversely the many 
other layers that are arrayed in the stack between them remain hidden offers 
some intriguing parallels that also seem to indicate a critical role for 
information hiding. The Google deep mind machine learned neural networks for 
image processing, for example, have 10 to 30 (or by now perhaps even more) 
stacked layers of artificial neurons, most of which are hidden.
Because of the non-linearity of the processes in play within these artificial 
deep stacks of layered artificial neurons it is difficult to really know in any 
definitive manner exactly what is going on. The outcomes from experimenting on 
the statistically trained (or in the vernacular, machine learned) models, by 
for example tweaking training parameters to experimentally see how doing so 
effects the resulting outcomes and by also subsequently forensically analyzing 
any generated logs & other telemetry are often surprisingly beautiful 
dreamscapes that are not reducible to a series of algorithmic steps applied by 
the many hidden layers to whatever input signals that have been fed to the 
input layer of neurons.
It seems to me that the emergence of consciousness & self awareness as well is 
exquisitely nonlinear in nature. And that this outcome characterized by being 
non-linear, itself depends on information hiding in order to be able to 
operate. Each successive layer in the stack is mostly unaware of the vast array 
of activities occurring on the layers beneath it... or above it for that 
matter. 
Would consciousness or self awareness even be possible without introducing 
information hiding in the deep stack through which these phenomena emerge? 
Personally I do not think we could be conscious or self aware without the 
multiple degrees of non-linearity introduced into the sensorial signal + 
triggered memory recall processing stream by the fire wall of information 
hiding.
It is by hiding away, by far most of the processing stack from us that we 
experience this seemingly magical state of being. We emerge in a non linear 
manner from a hidden world that we are (for the most part) blithely unaware of.
The fact that a very similar kind of process seems to be taking place in these 
stacked layers of artificial neurons most of which are hidden supports this 
thesis.
Is information hiding in fact, necessary to the emergence of self-awareness & 
consciousness?  



It's necessary for the "illusion of free will", and under mining that 
"illusion" would also undermine the idea of self.  My idea is that awareness, 
consciousness, and self-awareness are just aspects of models of the world we 
make most of which are not conscious but a few of which our brain puts into a 
coherent story (in language or images) for memory and future reference.  We 
can't remember everything and even if we did we couldn't access it in 
reasonable times.  So learning requires that the past experience be summarized 
some stories, some of which is fabricated by making it coherent with other past 
stories.  That's the reason the more an event is recalled by a witness the more 
coherent but the less veridical it becomes.
I also like the model of consciousness as the narrator of our minds consensus 
as well, which is what becomes our story. We believe in ourselves as being the 
narrator and mostly never ask what is it that our voice within is the narrator 
of?  We gloss over the incredible journey of dynamically interleaved and 
sequenced neural activities taking place in our brains that gives rise to our 
minds story, as a record of our neural consensus and as an emergence of the 
most prominent neural synchronized firing network at any given monent. 
Apparently, from what I've read there is a fairly large number of seemingly 
competing firing networks going on within our brains all the time, each of 
which is trying to recruit neurons to it and gain the momentum to achieve 
conscious prominence. 
We are our own talking heads :)And we also for the most part blissfully ignore 
just how noisy and dare I say argumentative our brains are.-Chris

Brent
"The world is made of stories, not atoms."
   --- Muriel Rukeyser






  

-- 
You received this message because you are subscribed to the Google Groups 
"Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to everything-list+unsubscr...@googlegroups.com.
To post to this group, send email to everything-list@googlegroups.com.
Visit this group at https://groups.google.com/group/everything-list.
For more options, visit https://groups.google.com/d/optout.

Reply via email to