Advancing Common Sense in Artificial Intelligence
Brain Simulator Thought is a cognitive architecture designed to model meaning, understanding, and common sense rather than pattern matching alone. It builds on years of development in Brain Simulator III, introducing a refined and more powerful internal model centered on atomic Thoughts, Links, and dynamically constructed ThoughtFields.
At its core, Brain Simulator Thought provides a substrate in which knowledge is not stored as static facts, but emerges from interconnected, competing, and evolving structures—much closer to how biological cognition operates.
Brain Simulator Thought is built around a new formulation of the Universal Knowledge Store (UKS). In this system, Thoughts are the atomic units of meaning, and Links (which are also Thoughts) define how Thoughts interact. Together, they form ThoughtFields—living cognitive spaces that support reasoning, learning, inference, and explanation.
Unlike traditional knowledge graphs, ThoughtFields are not passive data structures. They are actively constructed, searched, pruned, and reinforced as the system learns from sensory input, language, and action.
The system supports modular Agents, which operate over ThoughtFields to perform perception, reasoning, learning, planning, and control. Agents are independent modules and may be written in C# or Python. Brain Simulator Thought runs on Windows and macOS.
The project is supported by the non‑profit Future AI Society, which hosts regular online development meetings. Participation is free, with optional paid memberships to support continued development.
The initial upload of this project is in progress. Documentation for installation, execution, and development may be functional Open Items: Testing, Documentation.
The project is open‑source and distributed under the MIT License.
Brain Simulator Thought leapfrogs conventional AI approaches by addressing a core limitation: the inability to represent and manipulate meaning at the level required for common sense.
The system can:
- Represent multi‑sensory information, allowing sounds, words, images, and actions to coexist in a shared cognitive space.
- Maintain a real‑time mental model of the current environment, analogous to human situational awareness.
- Represent and reason over ambiguous, incomplete, or conflicting information.
- Learn action–outcome relationships, enabling experience‑based behavior.
- Update ThoughtFields continuously for real‑world and robotic applications.
- Incorporate extensible Agent modules to add new cognitive capabilities.
A Thought is the fundamental unit of cognition. Any idea, concept, object, action, or internal state—such as dog, play, happy, or eat—is represented as a Thought.
Individually, Thoughts have no intrinsic meaning. Meaning arises only through their Links to other Thoughts. Internally, a Thought is implemented as a lightweight reference, making large‑scale cognitive structures efficient and scalable.
In keeping with the biological precept that all neurons are functionally similar, a Link is a Thought which connects other Thoughts. Each Link references three Thoughts:
From → LinkType → To
For example:
Fido → is‑a → dog
Links themselves are first‑class entities and may have attributes and properties. Most Links are considered simultaneous and unordered, forming the structural fabric of a ThoughtField.
A ThoughtField is a collection of interconnected Thoughts and Links sufficient to support meaningful reasoning within a context.
ThoughtFields are constructed dynamically during perception, search, and learning. They are not static graphs, but active cognitive regions that evolve over time.
Brain Simulator Thought implements inheritance in a way that mirrors human reasoning efficiency.
If:
dog → has → 4 legs
Fido → is‑a → dog
then querying Fido will automatically include the attribute has 4 legs, even if it was never explicitly stored.
Exceptions are supported naturally:
Tripper → is‑a → dog
Tripper → has → 3 legs
The exception overrides inherited information without duplicating all attributes. This allows the system to remain compact, expressive, and efficient.
A Statement may consist of a single Link or a structured combination of Links. This enables conditional and contextual reasoning.
Example:
[Fido → plays → outside] → IF → [weather → is → sunny]
Statements allow the system to represent knowledge that is not simply true or false, but dependent on conditions.
A Sequence is an ordered set of Links, typically representing spelling, actions, or temporal structure.
An Event represents a single occurrence involving Thoughts and Links. Ordered Events form Traces, which are used for learning, behavior modeling, and provenance.
Before adding new information, Brain Simulator Thought performs Search to determine whether equivalent or competing structures already exist within a ThoughtField.
Search supports tasks such as:
- Finding the best matching Thoughts given a set of attributes.
- Matching ordered sequences.
- Determining the most useful attributes for a given Thought.
Learning follows a biologically inspired cycle:
Capture → Compete → Consolidate → Prune
This process ensures that ThoughtFields grow in capability without uncontrolled expansion.
The Universal Knowledge Store is the underlying data structure that supports ThoughtFields. In Brain Simulator Thought, the UKS has evolved from a simple node‑and‑edge store into a substrate optimized for atomic Thoughts, expressive Links, inheritance with exceptions, and dynamic cognitive construction.
Rather than storing facts, the UKS enables meaning to emerge from interaction. The UKS is a .dll which easily incorporates into projects on manny languages and platforms.
Brain Simulator Thought is developed in the open and supported by the Future AI Society.
- Join free development meetings
- Contribute code, ideas, or research
- Support continued development through membership
Visit: https://futureaisociety.org
Thank you for your interest in Brain Simulator Thought—a step toward machines that don’t just compute, but understand.
Thank you for your interest in Brain Simulator Thought—a step toward machines that don’t just compute, but understand.