The Cartesian Cut in Agentic AI
TLDR
This paper introduces "Cartesian agency" in LLM-based agents, where a learned core is separated from an engineered runtime, discussing its pros, cons, and alternatives.
Key contributions
- Introduces "Cartesian agency" in LLM agents, separating a learned core from an engineered runtime via a symbolic interface.
- Analyzes the benefits (bootstrapping, modularity, governance) and drawbacks (sensitivity, bottlenecks) of this design.
- Compares Cartesian agents with "bounded services" and "integrated agents" based on control, autonomy, and oversight.
Why it matters
This paper is crucial for understanding the fundamental design choices in modern AI agents. It highlights how the separation of control in LLMs (Cartesian agency) impacts their development, robustness, and governance, guiding future architectural decisions.
Original 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.
📬 Weekly AI Paper Digest
Get the top 10 AI/ML arXiv papers from the week — summarized, scored, and delivered to your inbox every Monday.