Neurocompositional Computing: From the Central Paradox of Cognition, to a New Generation of AI Systems
What is the reason for the rapid progress of AI from the 20th century to the 21st century, and what can be done about the limitations that remain? This progress is attributed by the widely accepted narrative to massive increases in computational and data resources that are available to support statistical training in deep artificial neural network. We demonstrate that the development of new types of computation is a crucial additional factor. Neurocompositional Computing adopts two principles which must be respected simultaneously to achieve human-level cognition. These are the principles of Compositionality, and Continuity. They seemed incompatible until a recent mathematical discovery showed that compositionality could be achieved not only by discrete methods for symbolic computing but also novel forms of continuous neuronal computing.
Source:
https://arxiv.org/abs/2205.01128
Leave a Reply