This paper introduces AI Implicit, a foundational paradigm that reconceptualizes intelligence as the capacity to extract, compress, and transfer tacit knowledge rather than optimize task-specific performance. It argues that current AI systems are fundamentally limited by the optimization paradigm, which prioritizes correlation-based accuracy while failing in transfer, causal understanding, and epistemic awareness. The proposed framework is built on four core principles: knowledge density, tacit knowledge extraction, cross-domain transfer, and calibrated epistemic confidence. It further establishes a novel evaluation methodology centered on knowledge compression metrics, including compression ratio, extraction rate, and transfer efficiency, aligned with human learning dynamics. Overall, the work positions AI Implicit as a comprehensive research direction offering both a measurable definition of intelligence and a principled pathway toward artificial general intelligence.
Momen Ghazouani (Mon,) studied this question.