[ExI] Runaway AI not likely

Jason Resch jasonresch at gmail.com
Fri Apr 7 05:58:29 UTC 2023

On Thu, Apr 6, 2023 at 9:19 PM Stuart LaForge via extropy-chat <
extropy-chat at lists.extropy.org> wrote:

> Quoting Jason Resch via extropy-chat <extropy-chat at lists.extropy.org>:
> > When you say you need more evidence, are you referring to the
> computational
> > theory of mind, or my explanation of the feeling of freewill as a
> > consequence of chaotic unpredictability?
> I was referring specifically to the computational theory of mind as
> there are competing theories of mind that all have non-zero prior
> probabilities.

I'll start with two questions to guide the answer I provide you:

1. Do you think anything a neuron does is not Turing emulable (i.e. an
appropriately programmed computer would not be able to simulate it to a
sufficient degree of accuracy)

2. Do you believe philosophical zombies are possible (or plausible), what's
your degree of confidence in either their possibility or impossibility?

> But now that you mention it, some sort of evidence or
> further characterization of the mechanism by which deterministic chaos
> could give rise to the feeling of freewill would certainly not hurt
> its case. :)
This will be long, but let's start by reviewing some definitions of

A) There is the conventional understanding of free will which says we are
able to decide things freely and independently of anything, that is, we can
make a choice, it could be any possible choice, and nothing within our
physical state of our brain or anything else determines it. Nor is it the
result of a random process outside of one's control. This form of freewill
is incompatible with existence within a purely deterministic universe.

B) On the other hand, there is an understanding of free will in which
agents are free to make choices in accordance with their preferences, and
this can take place within a deterministic universe. Because this
understanding of free will is compatible with a deterministic universe it
is sometimes called the "compatibilist" view of free will.

When we analyze these two perspectives, we find both lacking:

Understanding A, seems to call for a logical impossibility. It seeks to
define free will as something that is neither deterministic, nor (random)
non-deterministic. The most basic logical law, that of non-contradiction,
implies this understanding of free will cannot and does not exist. We
cannot have something simultaneously be neither deterministic, nor

Understanding B, it seems to define free will in a way that is not free at
all. It is entirely subject to the laws and low level forces
operating underneath the agent's mind. In what possible way can we define
this subject's will as free, any more than a swinging pendulum can be said
to be free, when all the while ruled by forces beyond its control.

However, there is a more nuanced view, which recovers many of the aspects
of free will which were found in understanding A, but avoiding its logical
inconsistency and permitting it to exist within a system that is wholly
deterministic. To get there will require a few steps:

1. Understanding the various levels of processing operating within the
brain, each with their own level of causality
2. Understanding Douglas Hofstader's notion of a "Strange Loop", where a
higher or top level of causality "reaches down" to interfere within the
process at a lower level of causality
3. Understanding that chaotic nonlinear processes cannot be solved
via analytic formulae, but rather must be solved via simulation of each
intermediate state in order to determine what state the process P will be
in after T time steps.

Taking these together, we can build a picture of a deterministic system
which makes choices (at a high level), choices which impact and effect
events at every level in the hierarchy of causes, and choices which cannot
be predicted in advance (not by anyone, neither the entity making the
choices, nor any other observer, not even God. In effect, the universe
cannot unfold until giving such an entity the opportunity to make their
choice, or another way of saying this, is the universe must invoke their
conscious mind in order for their choice to be made).

To help understand the levels of causality operating within the brain, I
think it is useful to consider this quote by Roger Sperry:

"I am going to align myself in a counterstand, along with that
approximately 0.1 per cent mentalist minority, in support of a hypothetical
brain model in which consciousness and mental forces generally are given
their due representation as important features in the chain of control.
These appear as active operational forces and dynamic properties that
interact with and upon the physiological machinery. Any model or
description that leaves out conscious forces, according to this view, is
bound to be pretty sadly incomplete and unsatisfactory. The conscious mind
in this scheme, far from being put aside and dispensed with as an
"inconsequential byproduct," "epiphenomenon," or "inner aspect," as is the
customary treatment these days, gets located, instead, front and center,
directly in the midst of the causal interplay of cerebral mechanisms.
Mental forces in this particular scheme are put in the driver's seat, as it
were. They give the orders and they push and haul around the physiology and
physicochemical processes as much as or more than the latter control them.
This is a scheme that puts mind back in its old post, over matter, in a
sense-not under, outside, or beside it. It's a scheme that idealizes ideas
and ideals
over physico-chemical interactions, nerve impulse traffic-or DNA. It's a
brain model in which conscious, mental, psychic forces are recognized to be
the crowning achievement of some five hundred million years or more of
To put it very simply, it becomes a question largely of who pushes whom
around in the population of causal forces that occupy the cranium. There
exists within the human cranium a whole world of diverse causal forces;
what is more, there are forces within forces within forces, as in no other
cubic half-foot of universe that we know. At the lowermost levels in this
system are those local aggregates of subnuclear particles confined within
the neutrons and protons of their respective atomic nuclei. These
individuals, of course, don't have very much to say about what goes on in
the affairs of the brain. Like the atomic nucleus and its associated
electrons, the subnuclear and other atomic elements are "moleculebound" for
the most part, and get hauled and pushed around by the larger spatial and
configurational forces of the whole molecule.
Similarly the molecular elements in the brain are themselves pretty well
bound up, moved, and ordered about by the enveloping properties of the
cells within which they are located. Along with their internal atomic and
subnuclear parts, the brain molecules are obliged to submit to a course of
activity in time and space that is determined very largely by the overall
dynamic and spatial properties of the whole brain cell as an entity. Even
the brain cells, however, with their long fibers and impulse conducting
elements, do not have very much to say either about when or in what time
pattern, for example, they are going to fire their messages. The firing
orders come from a higher command.”
-- Roger Sperry in "Mind, Brain, and Humanist Values" (1966)

 To understand the idea of a strange loop, here is a quote from Douglas

“My belief is that the explanations of “emergent” phenomena in our
brains–for instance, ideas hopes, images, analogies, and finally
consciousness and free will–are based on a kind of Strange Loop, an
interaction between levels in which the top level reaches back down towards
the bottom level and influences it, while at the same time being itself
determined by the bottom level."
-- Douglas Hofstadter in "Godel Escher Bach" (1979)

This might sound paradoxical, but it's not. Consider that in the brain, the
motions of atoms determine, the molecules and chemistry, the chemistry and
proteins the biology, the biology determines the behavior of neurons, the
behavior of neurons determine the behaviors of neural networks and larger
brain regions, this activity determines what thoughts and ideas come to be.
And yet, there is also a reverse downwards causality at play: one's
thoughts and ideas lead to single nerve cells firing (e.g. when I tell you
to wink your right eye and you get that thought to do so, and this triggers
the nerve of your right eye lid to signal, or a state of fear causes a
release of hormones or adrenaline, triggering changes at the molecular

We also find a certain independence between the levels, a clear analogy
exists within computers, where the unfolding of the computation is defined
by the instructions of the program, as much (or in some ways more than) the
physical state of electric charges and fields within the physical computer.
This is so much so that the programmer can entirely ignore the physical
state of the computer when writing the program, he is operating within a
higher causal structure, one that is independent and freed from the laws
operating below. Tegmark notes this here:

“How can something as physical as a bunch of moving particles possibly feel
as non-physical as our consciousness? Well, I think it’s because our
consciousness is a phenomenon that doesn’t only have properties above and
beyond those of its parts, but also has properties that are rather
independent of its parts, independent of its substrate, independent of the
stuff that it’s made of.
Computation is also rather substrate-independent, because Alan Turing
famously proved that any computation can be performed by any substance as
long as it has a certain minimum set of abilities to compute. So this means
that if you were a self-aware computer game character trapped in your
game-world in some game in a future super-intelligent computer, you will
have no way of knowing whether you are running on Windows, on macOS or on
some other platform, because you would be substrate-independent.
Now I think consciousness is the same way. I think consciousness is a
physical phenomenon that feels non-physical, because it’s just like waves
and computations. More specifically, I think that consciousness is the way
information feels when it’s been processed in certain complex ways. So this
means that it’s substrate-independent, and this also means that it's only
the structure of the information processing that matters, not the structure
of the matter that’s doing the information processing.”
-- Consciousness is a mathematical pattern: Max Tegmark at TEDxCambridge
2014 <https://www.youtube.com/watch?v=GzCvlFRISIM&list=WL&index=4>

To understand the idea of non-predictability of chaotic nonlinear dynamic
systems, consider any of these problems in physics:

   - https://en.wikipedia.org/wiki/Double_pendulum (A swinging pendulum
   with a joint)
   - https://en.wikipedia.org/wiki/Three-body_problem (Three gravitational
   bodies in some kind of motion/orbit)
   - https://en.wikipedia.org/wiki/Hadamard%27s_dynamical_system (Bouncing
   billiard balls)

Knowing the state of these systems at time T, and asking for what the state
will be at time (T+n), cannot be solved in for in a single step. These are
problems for which there are no closed-form solutions and numerical methods
(simulating each time step) is required. This is obviously the case for any
non-trivial program running on a Turing machine, there is no way to get
immediately from the state of the Turing machine at time T, to the state of
the Turing machine at time (T+n). If it were possible, we wouldn't need
supercomputers to run long computations, we could jump straight to the end
and get the final answer.

You can view the brain as a collection of billiard balls, making it a
non-linear chaotic system. Or you could view its neural network as a kind
of circuit or computer, which similarly require numerical solutions to know
their future state.

Given this, a person's mind state at time T, cannot be determined or
predicted, without fully and accurately simulating it. If philosophical
zombies are impossible, this simulation would invoke that person's mind. We
would then not really be predicting that person's behavior, but rather, we
are invoking that person, together with their conscious mind, and watching
them to see what they do.

There is no way around this. Therefore, we can say the person's behavior is
fundamentally non-predictable, because the only person who can make their
choice is that person. We cannot learn what they will do, before they do
it, without their mind having the chance to make that choice. Note that all
this occurs naturally within a fully deterministic frame of physics (or of
the computer running the person's brain simulation). So it is compatibilist
in the sense it works fine with deterministic underlying layers, but like
the first understanding, it yields a kind of free will which is
(non-determinable) by anyone other than the person who makes their
particular choice. And before they make their choice, it is also
non-predictable to them. In this way, they do have a free will, which is
neither (determinable by outside parties, nor fully random or
non-deterministic). We just have to "wait and see" to find out what any
person will do.

To my mind, this recovers all the important aspects normally ascribed to
free will. But if there are others I have missed, I hope others will point
them out.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20230407/65319bb1/attachment.htm>

More information about the extropy-chat mailing list