Preface

The remarkable advancements in artificial intelligence (AI), primarily driven by deep neural networks, have significantly impacted various aspects of our lives. However, the current challenges surrounding unsustainable computational trajectories, limited robustness, and a lack of explainability call for the development of next-generation AI systems. Neuro-symbolic AI (NSAI) emerges as a promising paradigm, fusing neural, symbolic, and probabilistic approaches to enhance interpretability, robustness, and trustworthiness while facilitating learning from much less data.

Project MONARK

Our team is leading the development of a groundbreaking AI paradigm named MONARK. In this project, symbolic AI is rethought and pushed beyond its previous limits, and also integrated with neural networks.

It integrates fundamental principles such as uncertainty, causality, and intention, derived from common knowledge. With its modular and object-oriented design, MONARK is engineered to offer significant advantages, including explainability, trackability, collaboration, and revisability, establishing new standards for AI systems in both functionality and transparency.

Metaphor, Meaning

In many ways, MONARK is a metaphor for transformation: the classic symbolic AI caterpillar reborn as a butterfly, embodying a new paradigm of artificial intelligence. This rebirth brings forth a system transformed into an reasoning objects providing capabilities of probabilistic reasoning, explainability, modularity, and belief revision.
MONARK with ending "K" is not just a butterfly, but one with Knowledge at its core.

It actually stands for:

Modular Object-Oriented Neuro-Symbolic Artificial Intelligence Reasoning Engine & Knowledge Base System

Rebirthed AI

Detailed Description

Object-Oriented Knowledge Base

• MONARK's knowledge base is built around object-oriented principles, providing rich, structured knowledge.
• Every entity is a Thing, allowing consistent modeling across diverse domains. This includes living and inanimate objects, events, intentions, and more.
• A hierarchical class structure enables inherited behavior, ensuring scalability and logical organization.
• Enhancements are also achieved through composition, further extending functionality without altering base classes.

Reasoning Engine

• Actions in MONARK are objects, distinct from Things, and are closely tied to their agents. Each action inherently understands constraints and consequences, allowing for intelligent decision-making.
• The action delegates sub-problems to specialized actions (or even action agents) for modular problem-solving.
• It operates in multiple modes, including:
• Backward reasoning: Problem-solving based on desired outcomes.
• Forward reasoning: Investigative evaluations from initial conditions.
• Event scheduling, case-based reasoning, or even neural-network integration for specific problems.
• Temporary facts generated during reasoning are stored in quantum spaces, safeguarding the static knowledge base (the "main universe"). Permanent facts, unless tied to hypothetical situations, are added to the main KB.
• MONARK seamlessly integrates external data sources (e.g., APIs) for dynamic knowledge acquisition, such as movie theater projection schedule, GPS directions and travel times, or statistical data about the likelihood of certain facts or events.

Probabilistic Reasoning

• MONARK excels at decision-making under uncertainty, evaluating potential outcomes in uncertain scenarios (with probabilistic annotations).
• Individual facts and events are assigned probabilities, which are combined to infer conclusions.
• A dedicated framework models probabilities and propagates them through reasoning nodes to ensure coherent and reliable evaluations.

Modularity

• The knowledge base is composed of knowledge modules, allowing flexibility for specific applications. Only the required modules are included, much like NuGet packages in a C# program.
• Future versions aim to support module addition at runtime, enhancing dynamic adaptability.
• Knowledge engineers have produced modules for common knowledge domains like time, space, and relationships, all rigorously tested with unit tests to ensure robustness.
• Teams can create and publish private or public knowledge modules, fostering collaboration. A long-term vision includes a global repository, akin to "WikiKnowledge," for shared AI resources.

Learning and Belief Revision

• MONARK maintains tracking when new facts are learned or deduced, contributing to its explainability and adaptability.
• Belief revision:
  When false facts are detected, deduced facts are revisited and revised accordingly.
  • Knowledge modules are automatically updated and distributed. Human intervention may be required for fundamental revisions.
  • Version upgrades are application-dependent, ensuring controlled and appropriate deployment.
• Rigorous unit testing and integration testing are crucial for ensuring system reliability.

Status

Our flagship project is in the design and proof-of-concept phase at this point.

A detailed presentation including code, diagrams, and a live demo (or video) will be available in Q3-2025 for potential partners. Contact us if you are interested.