[ExI] Fwd: Is Artificial Life Conscious?

Jason Resch jasonresch at gmail.com
Tue Apr 26 12:13:04 UTC 2022

On Tue, Apr 26, 2022, 1:53 AM Colin Hales <col.hales at gmail.com> wrote:

> On Tue, Apr 26, 2022 at 2:13 PM Jason Resch <jasonresch at gmail.com> wrote:
>> On Mon, Apr 25, 2022 at 11:09 PM Colin Hales <col.hales at gmail.com> wrote:
>>> On Tue, Apr 26, 2022 at 2:01 PM Jason Resch <jasonresch at gmail.com>
>>> wrote:
>>>> On Mon, Apr 25, 2022 at 10:54 PM Colin Hales via extropy-chat <
>>>> extropy-chat at lists.extropy.org> wrote:
>>>>> On Tue, Apr 26, 2022 at 1:02 PM Rafal Smigrodzki via extropy-chat <
>>>>> extropy-chat at lists.extropy.org> wrote:
>>>>>> ### I would be very surprised if the functional capabilities of
>>>>>> brains turned out to be impossible to replicate in digital,
>>>>>> Turing-equivalent computers.
>>>>>> Rafal
>>>>> Wouldn't it be great to actually do some empirical science to find
>>>>> out? Like start acting as if it was true (impossible) and start building
>>>>> artificial inorganic brain  tissue that is NOT a general-purpose computer
>>>>> (that artificial tissue would also have functionally relevant EEG and MEG),
>>>>> and then comparing its behaviour with the general-purpose computer's model
>>>>> of of the same tissue?
>>>> It looks like this work is in the process of being done:
>>>> https://www.youtube.com/watch?v=ldXEuUVkDuw
>>>> Jason
>>> Not even close. Can you see what just happened? There's a general
>>> purpose computer and software involved.  The game ends right there! Did you
>>> not read what I wrote.
>>> To build an artificial version of natural tissue is not to simulate
>>> anything. You build the EM field system literally. The use of computers is
>>> a design tool, not the end product. The chips that do this would be 3D and
>>> have an EEG and MEG like brain tissue. No computers. No software.
>>> The game has changed!
>> What if the computer simulation includes the EM fields?
>> Would that be sufficient to make a  conscious program?
>> If not, do you predict the computer simulation including the EM fields
>> would diverge in behavior from the actual brain?
>> Jason
> *This is exactly the right question!*
> To find out you have to do it. You do not know. I think I know, but I
> can't claim to have proof because nobody has done the experiment yet. My
> experimental work is at the beginning of testing a hypothesis that the real
> EM field dynamics and the simulation's dynamics will not track, and that
> the difference will be the non-computable aspect of brains.

I commend you and your work for challenging base assumptions. Such work is
always needed in science for progress to be made.

The difference, I predict, will be in how the devices relate to the
> external world, which is something that cannot be in any model because it
> is precisely when the external world is unknown (that nobody can program)
> that you are interested in its response (that forms the test context of
> interest). In the end it is about the symbol grounding problem. I have a
> paper in review (2nd round) at the moment, in which I describe it this way:
> ----------------
> The creation of chip materials able to express EM fields structurally
> identical to those produced by neurons can be used to construct
> artificial neurons that replicate neuron signal processing through allowing
> the actual, natural EM fields to naturally interact in the manner they do
> in the brain, thereby replicating the same kind of signalling and signal
> processing (computation). This kind of in-silico empirical approach is
> simply missing from the science. No instances of in-silico-equivalent EM
> field replication can be found. Artificial neurons created this way could
> help in understanding EM field expression by excitable cell tissue. It
> would also facilitate a novel way to test hypotheses in-silico.

What is the easiest way to test this theory of EMs role in consciousness or

Would you consider the creation of an artificial neural network that
exhibits intelligent or novel behavior to be a disproof of this EM theory?

Neuroscience and physics, together, could embark on such a development. It
> would help us reveal the neural dynamics and signal processing that is
> unknowingly not captured by the familiar models that abstract-away EM
> fields and that currently dominate computational neuroscience. *Note that
> the computational exploration of the EM fields (via Maxwell’s equations)
> impressed on space by the novel chip would constitute the design phase of
> the chip. The design would be sent to a foundry to build. What comes back
> from the foundry would express the EM fields themselves. The empirical
> method would be, to neuroscience, what the Wright Brothers construction of
> flying craft did for artificial flight.*
> -----------------
> The flight analogy is a strong one. Simulation of flight physics is not
> flight.

I see this argument a lot but I think it ignores the all important role of
the perspective in question.

For a being in the simulation of flight, it is flight. If we include an
observer in the simulation of a rainstorm, they will get wet.

That our hypothetical simulators see only a computer humming along and no
water leaking out of their computer says nothing of the experiences and
goings-on for the perspective inside the simulation.

As consciousness is all about perspectives and inside views, changing the
view to focus on the outside is potentially misleading. I could
equivalently say, "A person dreaming of a sunrise sees a yellow sun full of
brilliant light, but the room is still pitch dark!" But the darkness of the
room doesn't tell me anything about whatever experiences the dreaming brain
could be having.

I think it's the same with computer simulations. There's an external view
and an internal view. Each tells very little about the other.

I predict that in exactly the same way, in the appropriate context
> (novelty), that a simulation of 'braining' will not be a brain (in a manner
> to be discovered). The reason, I predict, is that the information content
> in the EM field is far larger than anything found in the peripheral
> measurement signals hooked to it. The chip that does the fields, I predict,
> will handle novelty in a way that parts company with the simulation that
> designed the chip. The chip's behaviour (choices) will be different to the
> simulation.

I do think that given the chaotic nature of a large and highly complex
system where small changes can be magnified, anything not modeled in a
brain simulation can lead to divergent behaviors. The question to me is how
important those unsimulated aspects are to the fidelity of the mind. Is it
all important to the extent the mind is inoperable without including it, or
is it something that makes a difference in behavior only after a
significantly long run? (Or is it something in between those two extremes?)

> The grand assumption of equivalence of "brain" and "computed model of
> brain" is and has only ever been an assumption, and the testing that
> determines the equivalence has never been done.

I agree with you that it should be.

You do not find out by assuming the equivalence and never actually testing
> it with a proper control (null hypothesis). Especially when the very thing
> that defines the grand failure of AI is when it encounters novelty ...
> which is exactly what has been happening for 65 years non-stop.

I am not sure I would say AI has failed here.

Take for example, my AI bots. They encountered novelty when I changed their
environment repeatedly, and each time they responded by developing new more
optimum strategies to cope with those changes.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20220426/7830dcce/attachment.htm>

More information about the extropy-chat mailing list