A rarely acknowledged but deep conceptual continuity runs from the earliest associative memories in theoretical neuroscience to the core mechanisms of modern artificial intelligence. This paper argues that associative memory is not an obsolete concept but the central computational primitive underlying both biological and artificial intelligence. We trace a continuous line connecting (i) early formal models of associative memories in the 1960s and 70s; (ii) the role of monosynaptic reflexes and episodic binding in the evolution of animal intelligence; (iii) the emergence of Kernel Machines, radial basis function (RBF) and hyperBF networks as powerful nonlinear associative memories; (iv) hippocampal indexing and scaffolding models; and (v) the attention mechanism in modern transformers, which can be interpreted mathematically as a high-dimensional associative memory (a form of learnable, metric-based hyper-RBF network). This unified perspective suggests that associative memory is not merely a component of intelligence but its origin and structural foundation.