[ExI] Fwd: Is Artificial Life Conscious?

[Stuart LaForge]

Quoting Rafal Smigrodzki:


> Message: 6
> Date: Fri, 29 Apr 2022 01:49:04 -0400
> From: Rafal Smigrodzki <rafal.smigrodzki at gmail.com>
> To: ExI chat list <extropy-chat at lists.extropy.org>
> Subject: Re: [ExI] Fwd: Is Artificial Life Conscious?
> Message-ID:
> 	<CAAc1gFh5cvud9WZkRD1fTj3ia9bkcHir+dG3wrpovn26qc+B5g at mail.gmail.com>
> Content-Type: text/plain; charset="utf-8"
>
> On Thu, Apr 28, 2022 at 11:02 PM Stuart LaForge via extropy-chat <
> extropy-chat at lists.extropy.org> wrote:
>
>>
>> The biggest problem that I have with the EM field mediating
>> intelligence or consciousness is that my own studies into the
>> mathematics of neural networks indicate that learning is mediated
>> multiplication of high-dimensional tensors divided into "layers", and
>> the more layers, the deeper the AI. EM fields are governed by quantum
>> mechanics and therefore linear and subject to an unlimited additive
>> property. That is to say that any number of wave functions may be
>> added together to make a new wave function.
>>
>> On the other hand, the most important element of the neural networks
>> are neurons that are characterized by non-linear activation functions.
>> The necessity of the non-linearity of the activation function is that
>> if neurons have a linear activation function, they cannot be organized
>> into separate "layers" and instead all act as one giant layer and a
>> single layer is quite stupid no matter how large. Therefore the sum of
>> quantum fields or other waves do seem mathematically able to exhibit
>> observable intelligence, whereas the sum of non-linear neurons do.
>>
>
> ### Yes, absolutely. (I think you meant "the sum of
> quantum fields or other waves do NOT seem mathematically able to exhibit
> observable intelligence").
>  ----------------

Yes, precisely. Thanks for catching that, Rafal. Waves are linear  
systems and obey the superposition principle of homogeneity and  
additivity. The most complex waves are precisely the sum of their  
parts, nothing more.

Non-linear systems, however, are more than the sum of their parts due  
to non-linearity of those parts resulting in emergent phenomena such  
as intelligence, consciousness, or even chaos.

> I think we can both agree on the following:
>
> Synaptic organization and transmission via ions and neurotransmitters that
> involve sparse synaptic connections between non-linearly activated neurons
> is how the brain implements complex recursive mathematical function on
> tensor-space, which enables intelligence, and which, one might hypothesize,
> has something to do with consciousness.
>
> You and I just pointed to different levels of this integrated process.

I agree our viewpoints are compatible. But I am suggesting that  
intelligence, and presumably consciousness, is an abstraction  
separable from the physical substrate. It is a somewhat dualist  
approach in that regard.

Stuart LaForge