From: [email protected]
[mailto:[email protected]] On Behalf Of Stephen Paul King
Sent: Saturday, August 30, 2014 8:35 PM
To: [email protected]
Subject: Re: AI Dooms Us
Hi Chris,
Here is the thing. Does not the difficulty in creating a computational
simulation of the brain in action give you pause?
Difficult yes, impossible I don’t think so. A simulation of the human brain
need not have the same scale as an actual human brain (after all it is a
model). For example statistical well bounded statements can be made about many
social behavioral outcomes based on relatively small sampling sets. This also
applies to the brain. A model could have a small fraction of the real brain’s
complexity and scale and yet produce pretty accurate results.
Of course it is complex for us to imagine today… the human brain is after all
vastly parallel with an immense number of connections 100s of trillions. Even
within a single synapse (one of a large number of synapses) there is a world of
exquisite molecular scale complexity and it seems multi-channeled to me.
However, it is also true that the global networked meta-cloud (the dynamic
process driven interconnected cloud of clouds operating over the underlying
global scale physical and technological infrastructure) is also scaling up to
immense numbers of disparate computational elements with thousands of trillions
of network vertices.
Perhaps I don’t understand the thrust of this statement? Why should it give me
pause?
The brain is a magnificent but of biology an admirable compact hyper energy
efficient computational engine of unequaled parallelism. Yes, we agree.
On the other hand the geometric growth rates of informatics capacity – in all
dimensions: storage, speed, network size, traffic, cross talk, numbers of
cores, memory, capacity of the various pipes.. you name it is also literally
exploding out in scale. And on the level of fundamental understanding we are
establishing a finer and finer grained understanding about the brain and how it
works – dynamically in real time – and doing so from many various angles and
scales of observation (from macro down to the electro-chemical molecular
machinery of a single synapse). There are major initiatives in figuring out (at
least at the macro scale) the human brain connectome. The micro architecture of
the brain (at the scale of a single arrayed column -- usually around six
neurons deep) is also being better understood, as are the various sensorial
processing, memory, temporal, decisional and other brain algorithms.
A huge exciting challenge certainly… but for my way of thinking about this, not
a cause for pause, rather a call to delve deeper into it and try to put it all
together.
Why are we assuming that the AI will have a "mind" (program) that can be parsed
by humans?
Who is assuming that? I was arguing that the code we create today will be the
DNA of what emerges, by virtue of being the template in which subsequent
development emerges from. Are you saying that our human prejudices,
assumptions, biases, needs, desires, objectives, habits, ways of thinking… that
all this assortment of hidden variables is not influencing the kind of code
that is written. The hundreds of millions of lines of code written by
programmers – mostly living and working in just a small number of technological
centers operating on planet earth -- that all of this vast output of code is
somehow unaffected by our humanness, by our nature?
Personally I would find that astounding and think it would seem rather obvious
that in fact it is very much influenced by our nature and our objectives and
needs.
I am not assuming anything by making the statement that whatever does emerge
(assuming a self-aware intelligence does emerge) will have emerged out from a
primordial soup that we cooked up and will have had its roots and beginnings
from a code base of human creation, created for human ends and objectives with
human prejudices and modes of thinking literally hard coded into the mind
boggling numbers of services, objects, systems, frameworks and what have you
that exist and are now all connecting up into non-locational dynamic cloud
architectures.
AFAIK, AGI (following Ben Goertzel's convention) will be completely
incomprehensible to us. If we are trying to figure out its "values", what could
we do better than to run the thing in a sandbox and let it interact in with
"test AI". Can we "prove" that is intelligent?
We don’t know what it will turn out to become, but we can say with certainty
that it will emerge from the code, from the algorithms from the physical chip
architectures, network architectures, etc. that we have created. This is
clearly an a priori assumption if we are speaking about human spawned AI – it
has to emerge from human creation (unless we are speaking of alien AI of
course).
We cannot even prove that we are intelligent or that we even exist. We do not
even know what we are. We think we think, but measurements of brain activity
indicate the thinking has already happened before we have had the though we
though we thunk!
Mere narrators of our minds we are. We do not even understand how we think…. Or
why we get our values. For example I tis becoming clear that our gut flora and
fauna have more influence over our moods and desires than was previously
realized… how much of “our” thoughts, decisions are really just our human host
executing on the microbial desires and decisions of the five pounds or so of
biological complexity in our guts?
I don't think so! Unless we could somehow "mindmeld" with it and the
mindmeld results in a mutual "understanding", how could we have a proof. But
melding minds together is a hard thing to do....
In thirty to forty years we may begin to converge as humans become increasingly
cyborgs…. And I am being very conservative. Apple is about to unleash smart
watches and google has its glasses… already people are thinking of having their
biometrics wired up to a networked monitoring service… people born blind are
being given rudimentary artificial vision. Nano-scale molecular machinery
techniques and self-assembling systems approaches are pushing the scale down to
levels where informatics may soon become incorporated throughout the body and
the brain itself.
My question for you is how much longer do you think we will remain recognizably
human? Twenty years, fifty years.. a hundred perhaps. I just don’t see us
stopping at some arbitrary wall unless our technology itself is collapsed by
our collapsing resource base.. or unless (it has been argued) there is a point
at which increasingly complex systems begin to fail. And this has merit as an
argument too. But then taking computer architecture for example instead of
scaling in complexity it scales out… multi-core architectures for example (each
single core’s complexity within manageable bounds)
I assume nothing (or at least make an attempt)… we very much do live in
interesting times… on this I think we can agree.
On Fri, Aug 29, 2014 at 3:16 AM, 'Chris de Morsella' via Everything List
<[email protected]> wrote:
From: [email protected]
[mailto:[email protected]] On Behalf Of Stephen Paul King
Are our fears of AI running amuck and killing random persons based on unfounded
assumptions?
Perhaps, and I see your point.
However, am going to try to make the following case:
If we take AI as some emergent networked meta-system, arising in a non-linear,
fuzzy, non-demarcated manner from pre-existing (increasingly networked)
proto-AI smart systems (+vast repositories), such as already exist… and then
drill down through the code layers – through the logic (DNA) – embedded within
and characterizing all those sub systems, and factor in all the many conscious
and unconscious human assumptions and biases that exist throughout these deeply
layered systems… I would argue that what could emerge (& given the trajectory
will emerge fairly soon I think) will very much have our human fingerprints
sown all the way through its source code, its repositories, its injected
values. At least initially.
I am concerned by the kinds of “values” that are becoming encoded in sub-system
after sub-system, when the driving motivation for these layered complex
self-navigating, increasingly autonomous systems is to create untended killer
robots as well as social data mining smart agents to penetrate social networks
and identify targets. If this becomes the major part of the code base from
which AI emerges then isn’t it a fairly good reason to be concerned about the
software DNA of what could emerge? If the code base is driven by the desire to
establish and maintain a system characterized by having a highly centralized
and vertical social control, deep data mining defended by an army increasingly
comprised of autonomous mobile warbots… isn’t this a cause for concern?
But then -- admittedly -- who really knows how an emergent machine based
(probably highly networked) self-aware intelligence might evolve; my concern is
the initial conditions (algorithms etc.) we are embedding into the source code
from which an AI would emerge.
On Monday, August 25, 2014 3:20:24 PM UTC-4, cdemorsella wrote:
AI is being developed and funded primarily by agencies such as DARPA, NSA, DOD
(plus MIC contractors). After all smart drones with independent untended
warfighting capabilities offer a significant military advantage to the side
that possesses them. This is a guarantee that the wrong kind of
super-intelligence will come out of the process... a super-intelligent machine
devoted to the killing of "enemy" human beings (+ opposing drones I suppose as
well)
This does not bode well for a benign super-intelligence outcome does it?
_____
From: meekerdb <[email protected]>
To:
Sent: Monday, August 25, 2014 12:04 PM
Subject: Re: AI Dooms Us
Bostrom says, "If humanity had been sane and had our act together globally, the
sensible course of action would be to postpone development of superintelligence
until we figured out how to do so safely. And then maybe wait another
generation or two just to make sure that we hadn't overlooked some flaw in our
reasoning. And then do it -- and reap immense benefit. Unfortunately, we do not
have the ability to pause."
But maybe he's forgotten the Dark Ages. I think ISIS is working hard to
produce a pause.
Brent
On 8/25/2014 10:27 AM
Artificial Intelligence May Doom The Human Race Within A Century, Oxford
Professor
http://www.huffingtonpost.com/2014/08/22/artificial-intelligence-oxford_n_5689858.html?ir=Science
--
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 [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/everything-list.
For more options, visit https://groups.google.com/d/optout.
--
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 [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/everything-list.
For more options, visit https://groups.google.com/d/optout.
--
You received this message because you are subscribed to a topic in the Google
Groups "Everything List" group.
To unsubscribe from this topic, visit
https://groups.google.com/d/topic/everything-list/YJeHJO5dNqQ/unsubscribe.
To unsubscribe from this group and all its topics, send an email to
[email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/everything-list.
For more options, visit https://groups.google.com/d/optout.
--
Kindest Regards,
Stephen Paul King
Senior Researcher
Mobile: (864) 567-3099
[email protected]
http://www.provensecure.us/
<http://m.c.lnkd.licdn.com/media/p/8/000/2c9/1ca/29d0ccd.png>
“This message (including any attachments) is intended only for the use of the
individual or entity to which it is addressed, and may contain information that
is non-public, proprietary, privileged, confidential and exempt from disclosure
under applicable law or may be constituted as attorney work product. If you are
not the intended recipient, you are hereby notified that any use,
dissemination, distribution, or copying of this communication is strictly
prohibited. If you have received this message in error, notify sender
immediately and delete this message immediately.”
--
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 [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/everything-list.
For more options, visit https://groups.google.com/d/optout.
--
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 [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/everything-list.
For more options, visit https://groups.google.com/d/optout.
