DriftScript: A Domain-Specific Language for Programming Non-Axiomatic Reasoning Agents
In the realm of artificial intelligence, developing adaptive agents that can function under conditions of insufficient knowledge and resources is a significant challenge. Non-Axiomatic Reasoning Systems (NARS) have emerged as a promising framework for addressing this challenge. However, the language traditionally used to program these systems, known as Narsese, presents some usability issues. A new language, DriftScript, has been introduced to overcome these obstacles, offering a more accessible approach to programming NARS-based agents.
Overview of DriftScript
DriftScript is a Lisp-like domain-specific language designed specifically for compiling to Narsese. It aims to simplify the programming experience by providing a more readable syntax and structured constructs. The key features of DriftScript include:
- Readable Syntax: DriftScript replaces complex symbolic notation with keyword-based S-expressions, making it easier for programmers to write and maintain code.
- Comprehensive Constructs: The language includes constructs for major sentence and term forms required in Non-Axiomatic Logic (NAL) levels 1 through 8, such as inheritance, temporal implication, variable quantification, sequential conjunction, and operation invocation.
- Compiler Architecture: The DriftScript compiler is a zero-dependency, four-stage pipeline implemented in 1,941 lines of C99, ensuring efficiency and ease of integration.
Integration with DriftNARS
When used in conjunction with the DriftNARS engine, DriftScript enables a robust interaction with external systems. This is facilitated through four structured callback types and an HTTP operation registry, allowing agents to perform a sense-reason-act loop autonomously. This integration enhances the capabilities of NARS-based agents, making them more adaptable and responsive to their environments.
Language Design and Evaluation
The design of DriftScript is guided by the principles of accessibility and functionality. Its formal grammar and language features are meticulously crafted to ensure that users can easily transition from traditional Narsese to DriftScript. The evaluation of the compiler has been thorough, involving:
- A 106-case test suite to validate the functionality of the language.
- Equivalence testing against hand-written Narsese to ensure accuracy and reliability.
- A NAL coverage analysis to confirm that all necessary logical constructs are supported.
- Structural readability metrics to assess the ease of understanding and maintaining DriftScript programs.
- Compilation benchmarks to evaluate performance and efficiency.
Conclusion
DriftScript represents a significant advancement in the usability of programming languages for Non-Axiomatic Reasoning Systems. By simplifying the syntax and enhancing the programming experience, DriftScript opens the door for a broader audience to engage with NARS and develop more effective adaptive agents. The source code for DriftScript is publicly available at https://github.com/seamus-brady/DriftNARS, inviting collaboration and further development in this exciting area of artificial intelligence.