The missing AI stage between narrow and general intelligence

Current AI systems fall into two categories: narrow (task-specific) and general (hypothetical). This monograph argues for a distinct intermediate regime where meta-learning, AutoML, continual learning, and evolutionary computation converge on a common goal: automating hyperparameter specification. The author defines Artificial Adaptive Intelligence operationally and introduces an adaptivity index to measure progress by combining the fraction of absorbed hyperparameters with performance against task-specialized baselines. Three pathways to removing human parameter specification are analyzed: data-aware configuration, structural morphing, and in-training self-adaptation. Case studies span aerospace design, finance, turbulence modeling, and vision-language systems.