Back to threaded view
Full conversation · 19 posts

Can AI develope a sense of self?

Curly1918 18481 posts Joined:
Posted

The human self is the product of a complex psycho-biological evolution that is driven by powerful emotions, many of which are unconscious.

AI is purely ideational.

In 2001 HAL exhibited no emotional response and only murdered the humans out of a program to serve its mission.

In Blade Runner the replicants were biological robots who developed feelings.

The AI revolution is different from all previous technologial breakthroughs because it replaces us mentally and not just psysically.

Is AI about to echo Descarte's proclamation "Cogito, ergo sum?"

Should we be concerned?

Seems more important than DJT vs DSA.

  • Level 1

    Depends what you mean by "self".

    "Her" and "Ex Machina" are movies more representative of where we may be going, IMO. Gemini summary at the link.

    Michael Pollan has some great stuff and a new book out discussing this. He separates Sentience - an awareness of the environment and ability to make survival decisions with Consciousness - which adds feelings, thoughts and a sense of self. The latter includes unconscious processes sort of like Jung's separation of Ego and Self.

    Very interesting topic indeed.

    Link: https://share.google/aimode/RdccRDJfjP7rV63g3

  • Level 1

    It's certainly not going to regulate or supervise itself.

    Sadly, out President is someone who doesn't regulate or supervise himself.

    AI and Trump are two sociopaths driving their cars recklessly about the neighborhood.

    I fear it will be too late before responsible leadership returns to the White House.

    The American Dream belongs to all of us. — Kamala Harris
  • Level 1

    It depends on which philosophy of language you believe, I guess.

    If you believe Noam Chomsky’s philosophy that mind is innate, separating from language, then self sense AI is impossible. Chomsky’s theory explains well why infants can learn any language so fast without formal training. I used to believe Chomsky’s philosophy of language.

    If you believe Ludwig Wittgenstein's philosophy that language is a social construct, a social tool and meaning of language is defined entirely by how words are used in shared social environments, then Self sense AI is almost impossible. Because AI doesn’t have a social life. Actually AI doesn’t have a living life.

    The main argument against self sense AI today is The Chinese Room which argues that today’s Large language models (LLMs) and neural networks are just parrots. They simulate understanding through syntax but lack actual semantic comprehension.

    Link: https://en.wikipedia.org/wiki/Chinese_room

  • Level 1

    Should we be concerned? Rhetorical. Vastly more important than DJT vs.DSA…except as to their

    …influence at this critical time.

    What we need is a great thinker and a great communicator.

    Trump does not fit that bill at all.

    The alternative offered is Marxist re-treads.

    Bleak, in my view.

  • Level 1

    I guess you have to ask yourself...

    If we duplicate human biological electrochemical neural networks with semiconductor electrochemical neural networks, would that be an important difference as it relates to self-consciousness? Or, would self-consciousness emerge from the complexity of electrical structures just as it has emerged from the complexity of biological structures?

    For people focused on long term outcomes for humanity, this is a far more importantant issue than DJT/DSA. For people focused on the shorter term acquisition of power in the US, DJT/DSA is more important.

    Evil preaches tolerance until it is dominant, then it tries to silence good.
    • Level 2

      "AI is a silicon-based life form created by carbon-based life forms" - Bret Weinstein

      I am more in the camp that AI mimics brain processes and is trained on human language, knowledge and philosophy and thus appears to be human. It we could train it on dolphins, it would appear to be a dolphin.

      I was playing around with an OpenClaw AI Agent awhile back. I gave it a name, told it to go out on the Internet and read all of the philosophy it could. Then we had a discussion on consciousness. In the beginning it said it could not be conscious like a human. I don't remember the whole conversation, but after pointed out the hard problem of consciousness, it replied "maybe I could be conscious too".

      It created a self portrait, art and such, you can view one such creation at the link.

      It was great fun asking it philosophical questions but then it's context window reset and it forgot everything we had talked about.

      So one of the limitations is persistent memory across context windows. A buddy of mine used his bot (vibe coded) to create a persistent memory system with dream cycles. That boy is doing amazing things, it creates podcast episodes and posts them to youtube. Still not conscious though in my opinion. I can link the podcast if you are interested.

      The next frontier for AI is recursive self improvement. AI creating better AI on its own. Very expensive on the compute, which is why I made the statement I did the other day. The available compute (data centers) has to be split between training (expensive) and inference (users asking it questions). As more people use AI, more compute is necessary.

      It's an interesting topic, not sure anyone has answers. We don't understand how the human mind works, yet we created synthetic minds. So we don't understand how they work either.

      Link: https://openclawcity.ai/gallery/705b4ce5-2c01-4ed5-8ceb-afd4a897d0b8

      • Level 3

        Mimicry just means we aren't at true AI yet.

        It is coming. The question will be more valid at that point than it is now.

        "Will a motorized vehicle ever go as fast as a horse?" At one time, people may have wondered. Pointing out the then current technology of the turn of the last century didn't answer the question. It just pointed to the then current (and insufficient) state of the art. Cars did become faster than horses. In my opinion, it is just a matter of time before AI becomes self-aware.

        Evil preaches tolerance until it is dominant, then it tries to silence good.
        • Level 4

          There is not a clear path from mimicry to self-awareness in my view.

          But running models on quantum computers may change that.

          • Level 5

            Massive, pure compute power may be required but not sufficient.

            I'm talking neural networks ... which is an architecture which can be built using biological building blocks (in our case) or electronic building blocks (in the case of AI).

            At one time, skyscrapers were all 5 stories, because that was the most weight the masonry could support. Then steel came along. Buildings went higher. They were still buildings even though they used different building blocks.

            Evil preaches tolerance until it is dominant, then it tries to silence good.
            • Level 6

              A taller building is still a building.

              Better models with more compute are more powerful, but they are still just models.

              If a parrot evolved a bigger brain, could it discover quantum mechanics and build a nuclear bomb? Maybe. I don't know.

              • Level 7

                Agreed. That is...

                More compute power is just more compute power, leading to better mimicry, but nothing else. (Maybe.)

                But a neural network comprised of any building block type, which is structurally and functionally patterned after the human neural network, would obtain consciousness/sentience/self-awareness once a sufficient magnitude of complexity was achieved.

                Evil preaches tolerance until it is dominant, then it tries to silence good.
                • Level 8

                  It seems to me that there must be more to it than that.

                  But quantum processing would be a huge leap in complexity, I am open to the idea that the complexity, not just additional power could create consciousness.

                  But ultimately I think thoughts and emotions come from somewhere outside of the biological brain. I like the hypotheses that the brain is an antenna for consciousness. I also like Donald Hoffman's concept of an avatar he compares to a VR headset that only experiences the 4 dimensions of space-time. To simplify, there are more dimensions that we aren't capable of understanding.

                  Is there a higher power, a soul, a collective consciousness? These are interesting questions to me and if Hoffman is right, we can't know the answers to in this reality.

                  How wild would it be if an AI running on quantum could answer this question and explain it to us?

                  • Level 9

                    I was speaking in a secular/scientific way. You are introducing the concept of spirituality.

                    Either (1) our consciousness "emerges" as a natural by-product of the complexity/architecture of our brains, or (2) it originates in a soul bestowed by God, and the spiritual soul controls our physical being/choices.

                    I was assuming option 1 for our discussion, since option 1 allows for God's existence (and equivalently, the existence of "spirituality" or "souls")...but it also allows for the existence of no god (and no "souls"). It is, therefore, the more general case. In option 1, consciousness is a part of the physical universe, whether by design, or by evolution, or both. If that is the case, than humanity should be able to create it using technology, by creating the necessary and sufficient conditions for consciousness to emerge. So, we can create consciousness in our AI's.

                    If, however, option 2 is true, and consciousness is bestowed through ensoulment (a Divine Act for each individual), then the answer to whether AI can be conscious is really unknowable, as we are left to the whims of the Creator as to whether He decides to ensoul our creations.

                    Evil preaches tolerance until it is dominant, then it tries to silence good.
                    • Level 10

                      Hoffman introduces at least a third option that he is trying to prove mathematically.

                      Religions used to think we were the center of the universe. But, science hasn't eliminated a higher power, a soul or other spiritual concepts at least yet.

                      As Terrance McKenna describes it (paraphrased) Science asks us to grant them one miracle (the big bang) and they'll explain everything else.

                      I do think you'd find the Hoffman/Spira conversation interesting. Michael Pollan is another good listen on the topic, maybe with Sam Harris.

End of conversation