Programmable Matter (PM) is composed of materials that can be programmed to modify their physical characteristics, such as alterations in shape and other properties. This paper proposes the implementation of an Artificial Neural Network (ANN) on a modular robot and PM system in order to enhance its computational and self-reconfiguration capabilities. The idea is to imbue these systems with computational intelligence, enabling it to adapt itself dynamically according to its internal constraints and/or its surrounding environment. Indeed, programmable matter is a distributed system that suffers from several challenges, like high number of modules, limited energy and computational capabilities, storage and communication. Deploying Artificial Intelligence (AI) techniques directly on modules will increase the ability of the systems to autonomously and dynamically optimize the use of the available resources. However, most of existing AI solutions need high computational capabilities and rely on centralized servers which are not suitable for PM systems with limited resources. In this paper, we study the feasibility of fusing AI and PM by implementing a distributed artificial neural network model directly on PM systems. The objective of this model is to let the system dynamically and in a distributed manner to determine its current shape. Our approach is demonstrated both by simulations and by implementation on a real Blinky Blocks platform. The conducted experimentation shows that the results are exactly the same as an ANN executed in a centralized manner.