Abstract
It is shown that supposedly paradigmatic examples of classic architecture do not contain local representations. In particular, Turing Machines (TMs) carry out transformations over sub-symbols where only the initial and final states may involve interpretable strings. In contrast, examples of computing systems with local representations lack the coding efficiency that is claimed to be paradigmatic of classical architectures. Thus, distributed, sub-symbolic computation should also be considered as a hallmark of classical architectures. In light of this and other commonalities, it is proposed that the traditional divide between connectionist and classical architectures is more apparent than real.
