Artificial Intelligence & Perception

Eleven protected courses are available for academic review. Student access remains closed.

Course 01 naming reconciliation: the existing Machine Learning & Data Foundations course is preserved. The proposed pathway label is Data, Probability & Machine Learning Foundations; no rename, new course or added probability curriculum has been performed.

Simplify the language. Do not simplify the knowledge.

Start with a real-life problem, question and prediction; explain plainly, introduce technical terms and a visual model, then mathematics, simulation, observation, evidence, verification, application, limits, reflection and deeper study.

Simulation-first access uses small supplied or synthetic datasets and browser models. Beginning does not require a GPU, robot, camera, LiDAR, paid cloud account or AI API. The virtual entrance is conceptual; no physical laboratory or hardware is provided.

Evidence labels: SIMULATION · CALCULATED RESULT · INTERACTIVE MODEL · SUPPLIED DATASET · SYNTHETIC DATASET. PHYSICAL MEASUREMENT and PHYSICAL EXPERIMENT apply only to actual physical evidence, never to synthetic examples.

Data provenance, uncertainty and human verification

Record source, dataset type, rights/license, date/version, purpose, transformations, known bias and limitations when applicable. Never fabricate provenance. A model may fail, misinterpret context, generate plausible unsupported output or degrade under changed conditions. Ask: what data, what assumptions, how certain, where can it fail, how can we verify, and when must a person intervene?

Evaluation covers train/validation/test separation, generalization, overfitting/underfitting, confusion matrices, accuracy, precision, recall, calibration awareness and distribution changes at the appropriate course level.

Vision examples need authorized image rights and explicit camera permission where applicable. Generative outputs require source verification; introductory learning agents stay in simulation. Sensor fusion requires noise/timing assumptions and physical validation before real use. The model is one part of data → model → service → application → monitoring → evaluation → human oversight.

Human intervention is necessary when uncertainty is high, validated conditions do not apply, or safety, privacy and significant consequences are involved. AI cannot fabricate evidence, reveal protected answers or replace human Academic Review. Curriculum changes require a new approval cycle.