The Cartesian Cut in Agentic AI

TL;DR

This paper explores how large language models (LLMs) can be integrated with engineered control systems to create agentic AI, comparing different control architectures and their trade-offs.

Contribution

It introduces the concept of Cartesian agency in LLMs, contrasting it with other control approaches and analyzing their implications for autonomy and robustness.

Findings

Cartesian agency enables modularity and governance in LLM-based systems.

Control split affects system sensitivity and bottlenecks.

Different control architectures trade off autonomy, robustness, and oversight.

Abstract

LLMs gain competence by predicting words in human text, which often reflects how people perform tasks. Consequently, coupling an LLM to an engineered runtime turns prediction into control: outputs trigger interventions that enact goal-oriented behavior. We argue that a central design lever is where control resides in these systems. Brains embed prediction within layered feedback controllers calibrated by the consequences of action. By contrast, LLM agents implement Cartesian agency: a learned core coupled to an engineered runtime via a symbolic interface that externalizes control state and policies. The split enables bootstrapping, modularity, and governance, but can induce sensitivity and bottlenecks. We outline bounded services, Cartesian agents, and integrated agents as contrasting approaches to control that trade off autonomy, robustness, and oversight.