A Reliable Artificial Intelligence System Architecture Based on Knowledge Reasoning

Artificial intelligence (AI) technology is rapidly advancing and gradually sweeping into various domains, bringing profound changes to production and daily life. However, with the opportunities presented by AI technology come inherent risks. Ensuring AI is safe, reliable, and controllable to enhance users’ confidence is crucial for the long-term development of AI. Developing trustworthy AI involves focusing on aspects such as reliability, interpretability, fairness, robustness, and privacy protection. It is of paramount importance to constructing a reliable class of AI systems. Addressing the reliability issues prevalent in current widely deployed AI systems, this paper proposes an architecture for a reliable AI system based on knowledge reasoning. The aim is to achieve efficient, accurate and reliable knowledge processing and reasoning. AI-Math, a knowledge graph-based question-answering system developed according to this architecture, serves as an effective application example of reliable artificial intelligence.