Descartian Dualism of AI - Unifying Declarative & Statistical Realms through Knowledge Graphs and Large Language Models

As I embark on my PhD Research journey focussing on creation, evaluation and application of Large Language Model Augmented Knowledge Graphs, I wondered if these two aspects of AI could be explored through the lens of Descartian Dualism to understand their strengths and limitations from a philosophical angle. Such an approach could then help to assess the approaches for creating and applying hybrid models. Do reach out if you want to comment, collaborate or critique the perspective presented here. My interest, passion and curiosity is to explore how the fields of formal, natural and social sciences could come together to address opportunities and challenges presented through a hybrid model approach. This field is fast moving and I have a lot to learn.

Context

Descartes' dualism, exploring the separation of mind and body, potentially resonates within the field of Artificial Intelligence (AI). The AI landscape reflects a dualism of its own, embodied in the interplay between two well-known representation techniques: Knowledge Graphs and Large Language Models. This dialogue explores the rationalist and empiricist aspects, framing the structured and unstructured realms, and delves into the integration of declarative and statistical inferences to create a unified AI framework.

Descartian Comparison

Key concepts from Knowledge Graphs and Large Language Models when viewed through the Descartian Dualism could be summarised using the below framework

Way Forward

Rationalism-Empiricism with Declarative-Statistical Inferences ?

The Descartian dualism within the field of AI transcends the boundaries of pure rationalism and empiricism to encompass a nuanced integration of declarative and statistical inferences. It's essential to recognize that the distinction between these approaches is not as rigidly binary as it may initially appear. Nevertheless, exploring the symbiosis of Knowledge Graphs (KGs) and Large Language Models (LLMs) holds promise for the development of AI systems that seamlessly blend rationalist depth with empiricist adaptability. Substantial advancements are already underway in both research and engineering spheres, dedicated to the creation and refinement of LLM-KG hybrid models. The potential cognitive synergy emanating from such hybrid models introduces exciting prospects for enhanced explainable and transparent AI models amenable to heightened scrutiny regarding the content and decisions they generate.

Summary

In navigating the nuanced terrain of Descartian AI dualism, the fusion of Knowledge Graphs and Large Language Models not only reconciles rationalist and empiricist principles but also seamlessly integrates both declarative and statistical modes of representation and inference. This synthesis not only represents a potential breakthrough in artificial intelligence but also resonates with the intricate interplay between structured knowledge and the adaptability derived from extensive experiential learning. As we continue on this intellectual journey, the synergy of rationalist and empiricist approaches within AI, transcending the realm of pure science and engineering, and enriched by the integration of declarative and statistical inferences, offers a comprehensive framework for exploring the profound philosophical depth and versatility essential to propel the field forward.