Reductionism. “LLMs are just mathematical models (matrix multiplication, calculations on GPU, next token prediction, Chinese room, stochastic parrot), which doesn’t understand, doesn’t think, have no meaning, etc.” Unfortunately, this is the level at which a lot of A(G)I researchers are.
What people often forget is that at the very lowest levels, all intelligent, computational systems approximating a Universal Turing Machine are built by unintelligent gates and matrix multiplication. Indeed, Embedded Agency shows why a very unintelligent system can be made into intelligent and conscious solely via use of internal memory and parallelization.
What people often forget is that at the very lowest levels, all intelligent, computational systems approximating a Universal Turing Machine are built by unintelligent gates and matrix multiplication. Indeed, Embedded Agency shows why a very unintelligent system can be made into intelligent and conscious solely via use of internal memory and parallelization.