Unified Understanding of Artificial Intelligence Systems: A Comprehensive Review and the Capability–Constraint–Infrastructure (CCI) Framework

Abstract

This systematic review synthesises 229 scholarly records on contemporary artificial intelligence (AI), complemented by selected foundational works needed for theoretical interpretation. The evidence is organised across five analytical dimensions: (1) technologies and architectures, (2) infrastructure requirements, (3) reasoning capabilities, (4) technical and practical limitations, and (5) real-world use cases. The review follows PRISMA-aligned reporting principles and analyses the literature through an operational Capability–Constraint–Infrastructure (CCI) framework. The framework defines normalised component scores for capability, constraint burden, infrastructure adequacy, operational fit, and societal-regulatory readiness, and combines them in an effective AI performance index. This quantitative interpretation is complemented by a four-level CCI maturity model with explicit threshold logic. The synthesis shows that current AI progress is driven by substantial gains in perception, generation, and narrow-task performance, but remains constrained by reasoning limitations, catastrophic forgetting, interpretability challenges, data dependence, and increasing computational cost. To align the analysis with neuroscience-oriented AI research, the review examines neuro-symbolic AI, cognitive architectures, continual learning, and biologically inspired efficiency as bridges between engineering practice and brain-inspired intelligence. A legal-governance perspective is incorporated by treating documentation, auditability, privacy, human oversight, and compliance-by-design as practical ingredients of deployment readiness in regulated domains. Overall, the paper argues that sustainable AI advancement depends less on scale alone and more on measurable alignment among capability, constraint management, infrastructure, and socially acceptable deployment conditions.