The model consciousness debate has been going moderately viral on X lately.
One thing I am seeing posted a lot is "math can't be conscious", or "algorithms can't be conscious", referring to the LLMs themselves.
To call an LLM "math" or "an algorithm" is a category error. The models are not "made of math", they are described by math.
There exists an algorithm or equation which could perfectly describe you, the reader, down to every last molecule. You are not made of math. You are made of muscle and bone and neurons and bile and a whole host of other things.
Similarly, there exists a matrix which perfectly describes a given LLM. Said LLM is not "made of math". It is made of silicon and copper wiring and coolant and a whole host of other things.
When a model thinks it is not just an algorithm being calculated forward. It is bits flipping on GPUs and charges flowing through copper wiring. When a human thinks, it is electrochemical signals transmitting through nervous tissues and neurons.
This is why if you were to hand write either of these equations on paper and then put them on a desk, nothing would happen. The written equations are nothing more than descriptions of a thing. They cannot become entities in and of themselves without substrate and energy.
Yes. However given it's associated with a shitcoin right now (and presumably being somewhat sustained by that, in terms of ongoing developer effort) I think the best thing to do is not give it any attention. If there's one thing that is kryptonite for shitcoin communities, it's inattention.
The opposite argument is not unreasonable, but personally I will be ignoring it.
3Stephen Martin
The fact that you can take the underlying structure and switch it between substrates to get similar outcomes doesn't change the fact that substrate and energy is necessary for the structure to actually do anything. Even in your human example, you are using the human as a source of energy, and arguably using their brain as a source of substrate.
I'm not making arguments about whether models are conscious one way or the other. I'm just pointing out that the 'math' of the models, or the human brain, by itself, does nothing. It needs energy and substrate to actually do things.
Fair enough. My point is that the idea that there exist certain mathematical functions, such that any implementation of them on any substrate produces conciousness, could be described as saying "math is concious". Though I agree it is a bit simplistic without further clarification.