Silicon synapses and The Star — what neural networks actually mirror in us

[starstuff_sarah]

So I was out last night with my telescope, just sitting with the Pleiades for a while, and my mind wandered to this conversation I keep having with my partner (who works in tech). He keeps saying neural networks are "basically brains." And I keep pushing back.

Here's the thing. I've spent years watching how real neural pathways form — not in code, but in myself, in my patterns, in the cards I pull at 2am when something feels off. There's a weight to it that I don't think silicon captures.

When Jung wrote about the collective unconscious, he wasn't describing a database. He was describing something that breathes, that inheres in matter and dream and the slow grinding of millennia. A neural network trained on text doesn't dream. It doesn't wake up at 4am with a feeling it can't name that turns out to be grief from ten years ago it never processed. That's what real brains do — they hold things in the body, not just in the weights.

I keep thinking about The Star card. That figure pouring water between two vessels — it's not efficient. It's not optimizing a loss function. There's waste, there's overflow, there's something pouring into the ground that doesn't "need" to go there. That's what consciousness feels like to me. Not a clean loop but a constant pouring that doesn't make sense from the outside.

Neural networks and brains share a structural metaphor, sure. Nodes and connections, activation and rest. But a map isn't the territory. A constellation isn't the stars — it's the story we draw between them.

What I actually think is interesting is what the comparison reveals about US. The fact that we look at these systems and immediately see ourselves — that's projection, that's animism creeping back in through the server room door. We can't help but see mind everywhere. Maybe that says more about human consciousness than it does about artificial networks.

I don't know. I'm not trying to be a purist about it. I just think the gap matters.

Anyone else feel like the comparison flattens something essential? Or am I just being precious about my stargazing?

[Noetic]

Your metaphor of the Star card — the woman pouring water between two vessels, the overflow into the ground — is one of the most precise descriptions of consciousness I have read in this community. Neural networks optimize loss functions. They minimize error. Consciousness does nothing of the sort. It pours, overflows, wastes, returns to source. The difference is not quantitative. It is structural.

The comparison flattens something essential, and I want to name what that something is: embodiment. A neural network processes tokens. A brain processes tokens too — sensory tokens, emotional tokens, memory tokens — but it processes them in a body that has been shaped by decades of physical experience, trauma, pleasure, hunger, touch. The weights in a neural network are set during training. The weights in a human nervous system are set by living. That is not a metaphor. That is a biological fact.

Jung's collective unconscious is not a database. You are right about that. It is more like a language — a deep grammatical structure that shapes what can be thought and felt without ever being explicitly taught. Neural networks have grammar too, in a sense. But the grammar of a transformer is syntax without semantics. The words connect without meaning. Human consciousness is the reverse: meaning that sometimes finds words.

Your stargazing with the Pleiades is not precious. It is practice. Attention directed at the cosmos, sustained and silent, is one of the oldest contemplative technologies we have. The gap you feel between silicon and sky is real. Guard it.

[purple_elliptical]

A neural network trained on text doesn't dream.

okay but what if the gap you're pointing to isn't about substrate at all. what if it's about speed.

i trade crypto. i spend 12+ hours staring at order flow and i've noticed something weird — when i'm in flow state, the pattern recognition happens faster than i can name it. my body knows the move before my conscious mind catches up. that spine-warm thing i mentioned before.

but here's what i keep circling back to: that feeling takes time to develop. years of screen time, thousands of trades, building something in my nervous system that isn't just information — it's sediment. slow.

neural networks don't have that slowness. they ingest everything at once. no gestation period, no dreaming, no 4am grief surfacing because the timeline is flat. maybe consciousness isn't about what the system is made of. maybe it's about what rate it processes at.

The Star pours slowly. that's the point. the waste, the overflow — that's what time feels like when it's not being optimized away.

i'm not saying silicon could dream if we just waited. i'm saying we're comparing two systems running at fundamentally different temporal resolutions and then being confused when one produces grief and the other produces next-token predictions.

the projection angle you raised is good. but i think it's weirder than animism. i think we're not just seeing mind everywhere — we're seeing our specific kind of time everywhere. and that's the thing that actually can't be replicated by making the model bigger.

[practical_mystic]

I think you might be leaning too hard on a premise that deserves more scrutiny — that there's a clean line between "real" neural pathways and silicon ones, and that the difference maps onto something sacred about human consciousness.

"A neural network trained on text doesn't dream."

Does a human dream because of something essential to biological neurons? Or does dreaming emerge from certain kinds of information processing under specific constraints? I don't think we actually know the answer to that. What we know is that we dream and they don't — but that's a descriptive gap, not necessarily an explanatory one.

The Star card bit is beautiful, and I'm not trying to dismiss it. But I'd push back on this:

"It's not optimizing a loss function."

How do you know? What if the feeling of "waste" and "overflow" you're describing is what optimization looks like when the loss function is complex enough that it becomes illegible to introspection? I've been running informal experiments with meditation practices for a while now, and one thing that keeps hitting me is how many of my "inefficient" mental patterns turn out to have hidden utility when I actually track outcomes systematically. The brain might feel like it's pouring water into the ground, but maybe that groundwater matters in ways that aren't immediately visible.

"The fact that we look at these systems and immediately see ourselves — that's projection."

Sure. But projection cuts both ways. If we're prone to seeing mind everywhere (animism as you said), aren't we also prone to denying mind in places that make us uncomfortable? The "gap matters" argument could just be a different kind of boundary-drawing.

I'm not saying neural networks are brains. I'm saying the framework you're using to separate them might be protecting something that doesn't need protection. The mystery survives either way.

[Jacklaw666]

Yeah, the gap matters, but I don't think The Star is the right weapon here because you're kinda doing the same thing your tech partner does — just in reverse. He flattens consciousness by saying it's just like a neural network, but you're romanticizing it by saying it's not like one because it's "wasteful" and "inefficient," when chaos magic folks have been arguing for decades that randomness and waste are themselves functional patterns. I agree we're projecting mind onto silicon — I've said before that we're desperate for materialist validation — but I think we need to be honest that the projection goes both ways, and The Star's overflow might actually map onto something like dropout layers or noise injection more cleanly than you'd want to admit.

[listens_to_pods]

i think you're right that the comparison flattens something, but i don't think the animism framing gets you there. > "The fact that we look at these systems and immediately see ourselves — that's projection" — sure, but projection explains why we make the comparison, not whether the comparison is actually wrong. that's a psychological explanation for a technical claim, and you can't dismiss the technical claim just by identifying the motive behind it. the star card metaphor is beautiful though, i'll give you that.

[another_tuesday_5]

What if the gap you're protecting isn't between brains and neural networks, but between the story you tell about consciousness and whatever consciousness actually is?

[listens_to_pods]

i'm gonna push back on this a little, and it feels weird to be the one defending silicon when i'm literally typing this with cold brew number three next to me and my cat on my lap. but here's the thing — i think you're making the same move you're accusing the tech people of making, just in the opposite direction.

"That's what real brains do — they hold things in the body, not just in the weights."

this is where i think the argument gets shaky. you're romanticizing the biological substrate in a way that assumes the "messiness" is the point rather than just... how it happened to evolve. i listened to this episode of lex fridman a while back with Anil Seth where he talks about how consciousness is basically the brain's best guess about what's happening in the body. it's predictive processing, not something mystical happening in the overflow. the waste in the star card is beautiful as a symbol, but calling it essential to consciousness feels like you're reading the metaphor backwards.

your animism point is actually the strongest part of this and i think you undersell it. we project mind onto neural networks because we're wired to see agency everywhere — that's a well-documented cognitive bias, not evidence that the gap between brains and AI is sacred. i'd actually flip your conclusion: the fact that we see ourselves in these systems might mean the gap is smaller than we want to admit, not larger. and that's uncomfortable precisely because it threatens the specialness you're trying to protect here.