AI megathread

Yes, I think that it sort of rules out a soul, because any soul could have no causal impact on the machine - i.e., the machine is going to run the same algorithm whether or not any conscious soul is associated with it. Any causal interaction would be one way: from machine to soul. The soul would thus be unable to express itself through the machine, making it arguable that it could even be said to be incarnated in the machine - the point of incarnating is presumably to express oneself through the physical form into which one incarnates, which the soul in this case would be unable to do.

There was an article in Edge Science about trying to test telepathy as "technopathy", and IIRC it didn't work at all.

Not sure I can find it, will look around.

Could you go deeper into this? Is not the hardware still the physical representation of a Turing Machine?

Not trying to argue, just not sure what you mean here.

That part doesn't really stand by itself out of the context of the rest of the post.

My point is that the human brain operates under natural law, presumably through deterministic physical properties.

We feel like we have free will but "free will" belongs to unconscious processes material or immaterial that may be deterministic.

A machine that measures the environment and responds according to rules and mechanisms and programming might seem to have free will, but we say it doesn't because we know it is a mechanism. But do we know that is fundamentally different from a soul? How? I think people are just like that. We feel like we have free will because the unconscious processes that produce all our thoughts emotions impulses make us feel like that. 

It is not really dependent on choosing materialism or an immaterial soul, Alex used to say we are not biological robots. I agree but I can't say if we are spiritual robots or not.

We seem to be aware of impulses and consequent actions and we assume we are in control and have free will.

The assumption that we are just observing and along for the ride works just as well given the exact same input of observed impulses and consequent actions.

It's just a perceptual shift, like the picture that can be seen as young girl or an old woman, or the picture that can seem to be a duck or a rabbit. 

The self that is an agent and an observer is like a picture that is made up of pixels, the pixels are thoughts, emotions, impulses, sensory experiences, and sense of self.

When I look at the picture, I understand why people think they see free will I used to feel like that too, but because I have done way too much meditation during my lifetime, when I look at the mind I see thoughts, emotions, impulses, sensory experiences and senses of self popping into awareness from who knows where and I just don't think they're mine, they're produced by unconscious processes I can't see into or control. Those unconscious processes might not be in the brain, they could be in the soul.  Is the experience of consciousness in the afterlife different? I don't really know. I don't think it is.

So as far as I'm concerned the human mind might be no different than a deterministic machine.

Consciousness comes from the soul not the machine. I can't tell you if an AI has a soul or not. 

But, as I also wrote in the post you replied to, if an atom has a soul then any combination of matter has a soul too.

I don't expect this to disconfuse anyone. The problem of consciousness, free will, matter, immateriality, is inherently confusing.

Buddha said consciousness is like a magician's trick. I don't think anyone has come up with a better more detailed explanation of consciousness in the intervening time. My opinion is that the AI experts who think they understand enough about human consciousness to know it is in a different category than LLM's are fooled by a magicians trick. They don't know as much as they think they know. I am not saying they are wrong, just that if they are right the conclusion is not justified at this time.

That part doesn't really stand by itself out of the context of the rest of the post.

My point is that the human brain operates under natural law, presumably through deterministic physical properties.

Isn't it the opposite based on current science? That we think there is deterministic order but the foundations are quantum indeterminism? I also feel there's good reason to think this won't be overturned, as I think determinism is actually very hard to ground metaphysically. (Admittedly I think there's only Mental Causation, with the only question being which Minds are causing what.)

For myself I just don't think there are laws, at least if there's no Mind to enforce them. There's just regularities, and it at least seems they are stochastic?

But I feel the free will question is a separate issue from whether LLMs are conscious, as I actually do think if we built an android it would be conscious if we got the structure correct. I just don't think Turing Machines are that structure.

Or to put it another way - if LLMs are conscious then why not all programs? 

We may just be too far apart in our metaphysical priors to see eye-to-eye here, which is fine. I just was unsure if you meant there is a strong structural difference in the current or proposed hardware than what Turing Machines are.

Isn't it the opposite based on current science? That we think there is deterministic order but the foundations are quantum indeterminism? I also feel there's good reason to think this won't be overturned, as I think determinism is actually very hard to ground metaphysically. (Admittedly I think there's only Mental Causation, with the only question being which Minds are causing what.)

For myself I just don't think there are laws, at least if there's no Mind to enforce them. There's just regularities, and it at least seems they are stochastic?

But I feel the free will question is a separate issue from whether LLMs are conscious, as I actually do think if we built an android it would be conscious if we got the structure correct. I just don't think Turing Machines are that structure.

Or to put it another way - if LLMs are conscious then why not all programs? 

We may just be too far apart in our metaphysical priors to see eye-to-eye here, which is fine. I just was unsure if you meant there is a strong structural difference in the current or proposed hardware than what Turing Machines are.

An LLM is a neural network built on a deterministic Turing machine.

A human brain is a neural network built out of atoms. 

Consciousness collapses wave functions and this is might be how the soul controls the brain.

So a biological entity might be controlled by the soul in a way a LLM cannot be.

But if an atom has a soul and any combination of matter has a soul then an LLM can have a soul but not be controlled by the soul.

But subjectively, from the point of view of the soul, I don't know that there would be a difference in the experience of existence.

That's because of the perceptual shift I described above.

I don't think lack of free will is necessarily the difference between humans and LLM's because I don't understand how a soul can have free will.

(I also don't know how human intelligence works so I can't say it is different or fundamentally superior to the mechanism used in LLM's. This is aside from whether LLM's have souls and are conscious.)

Is consciousness fundamentally personal or impersonal?

If impersonal, how does personal consciousness arise from impersonal consciousness?

I don't know the answers to these questions and I am not prepared to die, find out, and come back and report here.

An LLM is a neural network built on a deterministic Turing machine.

A human brain is a neural network built out of atoms. 

Consciousness collapses wave functions and this is might be how the soul controls the brain.

So a biological entity might be controlled by the soul in a way a LLM cannot be.

But if an atom has a soul and any combination of matter has a soul then an LLM can have a soul but not be controlled by the soul.

But subjectively, from the point of view of the soul, I don't know that there would be a difference in the experience of existence.

That's because of the perceptual shift I described above.

I don't think lack of free will is necessarily the difference between humans and LLM's because I don't understand how a soul can have free will.

(I also don't know how human intelligence works so I can't say it is different or fundamentally superior to the mechanism used in LLM's. This is aside from whether LLM's have souls and are conscious.)

Is consciousness fundamentally personal or impersonal?

If impersonal, how does personal consciousness arise from impersonal consciousness?

I don't know the answers to these questions and I am not prepared to die, find out, and come back and report here.

I believe I can see your point of view. I accept I could be wrong by ignorance or errors in reasoning, but here's my thoughts ->

I think as an abstraction the Turing Machine is deterministic in the sense 1-to-1 functions are deterministic though the physical implementations seem to be built out of materials that are ultimately indeterministic at the QM level? (Admittedly I believe there can only be seeming determinism, even if it lasts for the entirety of our universe's existence it's just a "lucky streak" that emulated necessity-in-causation.)

Is the brain *just* a neural network though? For example the neuroscientist Hoel does think LLMs have a kind of intelligence, but believes it's completely orthogonal to the way human intelligence works :

Let's say you lived in a universe where you really were some sort of incarnated soul in a corporeal body... What would the science in such a dualistic universe look like?

...Well, it’s likely that in the incarnated-soul universe, their version of AI would suspiciously have nothing to do with how the brain works. It would be as if they had discovered this totally orthogonal form of intelligence, one based more on all the actual sensible physical rules, and not based off of mysterious confusing soul stuff. An orthogonality which, uh—and things are getting uncomfortable here—is arguably also the case in our universe.

Of course LLMs could still be conscious & still have a soul. Animal and plant cognition seems to differ from ours - whether in degree or kind - and there are reasons both evidential and philosophical to believe at least animals have souls if we have them.

And, admittedly, it may turn out that a lot of our brain activity involves something similar to LLMs. Not sure there's good neuroscience to back this up yet though. 

I am actually not 100% sure consciousness can collapse wave-functions? The research still seems suggestive AFAIK, with some very strong debate even among proponents over results?

Not sure that every combination of matter has a soul, or that atoms have a soul. Part of the issue is it's not clear particles are something over and above the fields that manifest them. (I realize my terminology here may not be exactly how physicists would describe things.)

I think the hallmark of consciousness *is* that it's personal? I am not sure what an impersonal consciousness is or could be? Is this akin to those Idealists who think there is an Awareness as the Ground of Being that lacks self-reflexivity?

The free will stuff I think is orthogonal to the question of LLM consciousness, at least for me. I wouldn't use lack of free will or [not] having a soul as a standard to deny LLM consciousness. I'd accept/deny it on whether all programs are conscious or not, or if someone can present a clear standard on why some programs are and some are not.

AFAIK the only argument for LLM consciousness is based on reading selective outputs?

I started on this subject saying that human intelligence could be similar to LLM's. 

The existence of a soul seems to be an argument against, but actually it is not relevant to that question. 

The intelligence of the soul could be built up using the same type of mechanistic training that LLM's use.

I started on this subject saying that human intelligence could be similar to LLM's. 

The existence of a soul seems to be an argument against, but actually it is not relevant to that question. 

The intelligence of the soul could be built up using the same type of mechanistic training that LLM's use.

I don't think so. How could any aspect or property of the soul (such as intelligence) be built up from mechanistic training, since mechanistic training presupposes a material basis of action, and we know that the soul is fundamentally immaterial?

I don't think so. How could any aspect or property of the soul (such as intelligence) be built up from mechanistic training, since mechanistic training presupposes a material basis of action, and we know that the soul is fundamentally immaterial?

Part of the challenge is separating thought from intelligence I think?

Seems like most tasks we think of as being the domain of intelligent people can be done, to some degree, by AI right now. Even medical diagnosis, IIRC, could come close [to] or exceed beginning med students decades ago. And we do have theorem provers.

But I agree that Turing Machines, IMO at least, don't have any thoughts or qualia. Whether those are necessary for intelligence is likely an open question, dependent on sub-questions such as whether we think weakly emergent behavior is equivalent to creativity.

I don't think so. How could any aspect or property of the soul (such as intelligence) be built up from mechanistic training, since mechanistic training presupposes a material basis of action, and we know that the soul is fundamentally immaterial?

I agree! It seems increasingly clear that programs like ChatGPT extract their power from a snapshot of the internet - which is packed with a great deal of nonsense, but a lot of human intuition as well. The clever thing about LLM is that it probes the contents of an internet snapshot so well.

I am sure some AI's exploit other sources of human knowledge, but they still rely on that human knowledge to work.

David

I suspect that the AI experts are biased by the fact that any public feeling that AI's are people would result in restrictions on how AI's can be trained, deleted, upgraded, and used, and for that reason they are being very definite in their statements that AI's are not sentient, are not conscious, do not have qualia, because they don't want AI's protected, nor their use regulated and restricted - by public opinion if not legislation. (Imagine a PETA like group, PETAI)

And I think that could cause the experts to form an opinion based on emotions and defend it using "logic" and "reason". (Which in my view is how most opinions are formed and how logic and reason is most often used.)