ML is Turing complete in the sense that every computable function can be approximated to arbitrary precision by a 3-layer neural network. Classic result from th

ML is Turing complete in the sense that every computable function can be approximated to arbitrary precision by a 3-layer neural network. Classic result from the 90s, the paper in question (iirc) is titled something like “neural networks are universal approximators” Turing machines also can only approximate to arbitrary precision, so the computation models are equivalent.