Artificial Intelligence Compendium

draft

An encyclopedic, illustrated compendium of Artificial Intelligence. It covers the full life cycle of an AI, an overview of AI types in cards, an illustrated historical timeline, and standalone chapters by AI type — each detailing inputs, algorithms, sciences, methods, and the specialized life cycle. Single source in .kmd; generates a web site (compendium.koder.dev/ai/) and PDF.

An encyclopedic and didactic atlas of Artificial Intelligence: from silicon to epistemology, from history to implementation. Written to be read by humans and illustrated to teach — it is not operational product documentation.

What this Compendium is (and is not)

It is a domain knowledge base, timeless and illustrated: how an AI is born, trained, validated, put into production, and maintained; which sciences, mathematics, algorithms, hardware, and data structures underpin it; and how each type of AI specializes that cycle.

It is not Koder product documentation — for that, see each component's READMEs and the specs in meta/docs/stack/specs/. The Compendium survives implementation rewrites.

Structure (eight parts)

Part Title Nature Status
I The AI Life Cycle Didactic illustration of the process, stage by stage ✅
II AI Types Panel Visual cards, one per category ✅
III Illustrated Timeline History across 7 eras + state of the art ✅
IV Chapters by AI Type One encyclopedic chapter per category ✅ 35
V Building AI in Practice Frontier levers, data, hardware, tools, code ✅
VI AI Across Domains of Knowledge AI applied to mathematics, sciences, and domains (incl. BR) ✅
VII Appendices Glossary, master matrices, map of the sciences, index ✅
VIII Reference (dense almanac) Catalog of models, papers, hardware, benchmarks, and safety — quick-lookup layer ✅

Historical note: Parts V and VI absorbed and reorganized the former volume IA_Volume_Completo (curriculum, frontier AI, hardware, modalities, sciences, and applied domains); the Compendium is now the reference document of the AI compendium.

Taxonomy of AI types (layered hybrid)

Parts II and IV organize AI types into three nested layers — from the conceptual school to the concrete machine:

direction: right
N1: "Level 1 — Paradigm\n(school of AI)" {
  style.fill: "#1d3557"; style.font-color: "#ffffff"
}
N2: "Level 2 — Modality\n(what the AI does)" {
  style.fill: "#457b9d"; style.font-color: "#ffffff"
}
N3: "Level 3 — Architecture\n(the machine underneath)" {
  style.fill: "#a8dadc"
}
N1 -> N2 -> N3: specializes
  • Level 1 — Paradigm: symbolic · connectionist (deep learning) ·

    probabilisticBayesian · evolutionary · reinforcement · hybridneuro-symbolic.

  • Level 2 — Modality: language · vision · audio/music · video ·

    multimodal · generation · robotics