Definition
Commonsense Databases are massive repositories of “everyday” human knowledge and logical relationships (e.g., “water is wet,” “if someone dies, their subscriptions stop”). They are used to provide AI systems with the contextual understanding required for human-level reasoning.
Why It Matters
It provides the ‘common sense’ buffer that prevents AI from making catastrophic, literal-minded errors in human environments.
Core Concepts
- Cyc (Encyclopedia): The largest and longest-running AI project, founded by Douglas Lenat. It contains millions of hand-coded facts and rules expressed in first-order logic.
- Inference Engines: Tools within these databases that allow the AI to draw new conclusions from existing facts (e.g., inferring that a “Miami Dolphin” is a football player, not a mammal).
- NELL (Never-Ending Language Learning): A CMU project that scans the Internet 24/7 to learn new facts by identifying patterns in text across millions of web pages.
- The Context Gap: Synthetic AI often excels at math but fails at “common sense” (the Eurisko problem). These databases act as a “logic bridge” to human reality.
- Natural Language Processing (NLP): These databases are essential for making machines understand the nuances, metaphors, and ambiguities of human speech.