This project, he4o, aims to build a General Artificial Intelligence (AGI) system based on the "Spiral Entropy Reduction" theory. It targets developers and researchers interested in exploring a novel approach to AGI that combines rule-based reasoning with dynamic learning, aiming to overcome the limitations of purely "live" (reinforcement learning) or "dead" (rule-based) AI systems. The system claims to have achieved many core AI capabilities, offering a unique perspective on intelligence.

How It Works

The core of he4o is the "Spiral Entropy Reduction Machine" model, which integrates "先天精心设计的代码规则" (innately designed code rules) with "后天跑出动态终身习得知识数据" (dynamically acquired lifelong knowledge data). This is achieved through a "DOP (面向动态编程)" programming paradigm, contrasting with OOP by first abstracting from concrete examples and then using abstractions to solve concrete problems. The system emphasizes a "螺旋上升" (spiral upward) mechanism for continuous improvement, integrating concepts like sparse representation, feature extraction, temporal sequences, and value perception.

Quick Start & Requirements

• Install/Run: The README does not provide specific installation or execution commands. It mentions the system can run on iOS devices.
• Prerequisites: No explicit dependencies like Python versions or specific libraries are listed.
• Resources: The project claims to be runnable on a single terminal, with a dependency on disk I/O over computation.
• Links:
  • Theory: https://github.com/jiaxiaogang/HELIX_THEORY
  • Website: https://jiaxiaogang.github.io (stated as discontinued)
  • Documentation: https://james4ever0.github.io/he4o

Highlighted Details

• Supports a wide range of AI capabilities including perception, recognition, prediction, feedback, learning, planning, decision-making, and lifelong learning.
• Employs a "DOP (面向动态编程)" paradigm, focusing on deriving abstractions from concrete data.
• Claims to have achieved many core AI functionalities through continuous development since 2017.
• Features numerous small demos illustrating various learned behaviors like foraging, tool use, and object recognition.

Maintenance & Community

The project appears to be a personal endeavor by the author, "jiaxiaogang." The README indicates the website is discontinued, and the primary documentation is provided by a third party. There is no mention of a community forum, Discord, or Slack.

Licensing & Compatibility

• License: LGPL (GNU Lesser General Public License).
• Commercial Use: Commercial use is permitted upon contacting the author for authorization, with the author retaining final interpretation rights. A commercial fee structure based on product pricing (0.1%) is mentioned.

Limitations & Caveats

The project is described as a personal coding effort rather than formal research, and the author expresses modest expectations. While many capabilities are claimed, the README notes that engineering details are still being refined, and complex demos can be buggy, while simple demos may lack impact. The project is actively under development with a detailed timeline provided.