Repository logo
 

Article:
Kontrollierbare Kognition. Stochastische Elemente in symbolischen und subsymbolischen Systemen

Author(s):

Abstract

The intelligent behavior of current large language models is often described using the black box model, where symbols emerge from the statistical values of a dataset. To challenge this representation, the article opens up a genealogy of machine learning systems. Two examples demonstrate that the integration of a stochastic element into deterministic structures is a historical constant of this technology and is implemented in various ways. Against this background, the concept of the subsymbolic, which characterizes current connectionist systems, can be understood as part of an older fiction of controllable and modelable cognition.

Download icon

Published in:

Merkle, Benedikt: Kontrollierbare Kognition. Stochastische Elemente in symbolischen und subsymbolischen Systemen. In: Zeitschrift für Medienwissenschaft, Jg. 17 (2025), Nr. 1, S. 59-70.http://dx.doi.org/10.25969/mediarep/23745
Keywords
license icon

As long as there is no further specification, the item is under the following license: Creative Commons - Namensnennung - Nicht kommerziell - Keine Bearbeitungen