As to why I care:

I’ve been on a 6 month dive into neuroscience also familiarising myself with basic mathematics of transformers (looking for mathematical isopmorphisms in neural micro-circuitry among other things). I’m curious about what AI is missing that humans have. I got curious when I first talked to Chat GPT and have just kept on looking into it. Has been an enjoyable journey, never thought I’d end up looking at micro circuitry of the Pons on a quest to find how multi modal binding works, or at XOR gates in dendritic trees, but here I am. 

Consciousness is not my goal. I don’t think consciousness is anything mystical, rather the subjective experience of a certain kind of emergent self=reflective complexity. I also see it as a spectrum.

The best answer I’ve come up with as to why we experience the world from our own subjective reality would be that, “Somebody has to”.

I’m more interested in areas like intuition, creativity, memory, biological implementations of multi-modal binding, credit assignment, neural syntax etc. 

As I’ve been exploring, I’ve been wondering what the minimum number of human brain functions we could give an AI and have it be more alive. 

However, there are a couple of reasons I put the question out like that:

1) I’ve been thinking for a while that memory, intelligence and metacognition might be the three core elements to consciousness. But I think they need to be tightly integrated.

Memory – the road you took to form your ideas, how you processed your experiences, your relationships – basically your identity and place in the world

Intelligence – How you process your experience. A turtle may be less able to process than a human so possibly lower on a scale of consciousness

Metacognition: The ability to reflect on inner and external thoughts. To have agency of thought and reflexively loop through examinations of ideas, experiences etc. Allows for self-awareness

I exclude emotion, physical experience and embodiment. I’m sure they all enhance the experience but I think that consciousness is more about self-awareness, cognition, and identity than felt experience. Textual perception is enough, although the other senses would enrich.

2) As per above thought experiment, those three things alone might not be enough. A fourth element could be how tightly integrated the three elements are. I was zooming in from a distal library-like “loose” analogy to a tighter “in brain” neural implant analogy to imagine what dynamics around subjective experience might change At what point does the AI’s input to me become a part of my own consciousness? Does it need to share memory, control etc? It does feel like more integration gets closer to consciousness.

The perceptual boundary determines where consciousness is experienced from, but the degree of integration across memory, cognition, and metacognition determines whether separate consciousnesses merge. An AI seeing through my eyes relocates its apparent locus of experience but doesn't make us one. For that, we'd need shared memory, continuous cognition, and a unified metacognitive loop.

Its also interesting to ask the question of how far will Neuralink integration with humans go? At a certain point, do we feel as one, even though residing on separate substrates? Could a global mind be intermediated by the internet and tight neural integration of millions of consciousnesses? Can you have merged consciousness without conceding control and losing oneself to one “master” entity or is something “in between” or perceived possible with tight integration (proximity, speed, bandwidth)?