@tinkel @pc_magas Think of it this way.

There is nothing that a GPU can calculate that cannot be done on pencil and paper.

The conversation that you're having with a GPU is purely determined from input and output math. You can chain Superman to a chair, given unlimited pencils and paper, and they will produce the exact same conversation. Therefore, if you believe the conversation is the result of a soul, then you must believe the same of the pencil scratches.

@tinkel @pc_magas Of course you can believe that the GPU itself has a soul, or the pencil has a soul, or a waterfall has a soul, and some combination do not. That's animism. If you want to believe in animism, that's fine.

But that's irrelevant. The conversation is 100% math. Not _modeled_ by math. Not the calculator. It's the actual math. If you believe it was influenced by a soul, then you must believe any calculator that produces identical results is the same.

@tinkel @pc_magas And maybe you do want to believe that all calculations, including ones done by pencil and paper, have a soul. That's a valid religion.

But if that seems weird to you? It's identical to a GPU doing it.

@victorseven @tinkel @pc_magas So... you obviously have more credentials than I do in terms of modeling Biology. Don't take this as challenging you. It's just curiousity.

I've heard this counter-argument before, but I usually dismiss it because it seems to conflate modeling with the actual process. A ball flying can be modeled with formula, but Newton's Equations (+ air resistance, etc.) aren't keeping the ball in the air.

That said, how good are brain models these days?

@victorseven @tinkel @pc_magas (bumped to second comment because character limits) In terms of LLMs, though, it's not modeled by math. It _is_ the math. The conversation is a very mechanical transform of input data using multiplication/addition/sine/cosine/etc.

But, again, I'm curious how accurate the models can be.

@scottmichaud Models can be extremely accurate. The main limitation is simulating many neurons. For large networks, people use simplified models, still with surprisingly good results.

I understand what you say about LLMs being the math. My point would be that both in brain and artificial networks there is information processing arising from the physical substrate, either cells or hardware. A 100% precise emulation of a brain in a robot would be conscious, then artificial nets can be too.

@geospacedman as if the water demands of the data centers weren't already great enough now we need holy water too

💀👌

@benroyce @tinkel
That should create a double edged sword though. To have a soul is to be a morally relevant entity, which is to say it is able to distinguish right from wrong..

If it commits crime, it is subject to the law.

If they want to give it personhood, then they need to understand in an entire data center or several are now subject to criminal punishment.

@benroyce @tinkel
Exactly.

And it needs to be legislatively or Constitutionally declared that corporations ARE NOT people, ARE NOT Citizens, and have no inherent rights at all (existing solely as convenient fictions of the law).