Agential AI for Integrated Continual Learning, Deliberative Behavior, and Comprehensible Models 869

Zeki Doruk Erden, Boi Faltings · 2025

Read Agential AI for Integrated Continual Learning, Deliberative Behavior, and Comprehensible Models 869 by Zeki Doruk Erden, Boi Faltings (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-16922.txt.

Description

Zeki Doruk Erden and Boi Faltings present an early design for a new kind of AI system in a 2025 paper. Modern machine learning is strong at statistics but weak in three ways. It does not plan well, its inner workings are hard to read, and it cannot keep learning over time. Their system, called Agential AI, is meant to sit on top of statistical methods or work alone, and fix those gaps.

At its core is a learning method that models how the world changes over time. It carries guarantees of completeness and minimality, and it learns by varying and selecting small components. A behavior layer plans on the learned model and reuses high level behavior patterns. Early tests in a simple world show the design working as intended. The authors present it as a first sketch of a more deliberate, readable AI.

Catalog record

CREATORZeki Doruk Erden, Boi Faltings
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE78 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-16922

Browse the full library · About the Tower of Babel