AI Compendium · Part II — Panel of AI Types

draft

A visual overview: each type of AI presented in a compact card. The panel is the gateway to the encyclopedic chapters of Part IV — each card points to the chapter that details it.


How the panel is organized

The types are addressed by the triad of the hybrid taxonomy defined in the master index: Paradigm → Modality → Architecture. The panel's primary grouping is by paradigm (Level 1) — the "school" of AI. Within each paradigm, the cards carry their modality and architecture.

direction: right
IA: "Types of AI" { style.fill: "#1d3557"; style.font-color: "#fff" }

sim: "Symbolic"            { style.fill: "#457b9d"; style.font-color: "#fff" }
con: "Connectionist\n(Deep Learning)" { style.fill: "#457b9d"; style.font-color: "#fff" }
prob: "Probabilistic\n/ Bayesian" { style.fill: "#457b9d"; style.font-color: "#fff" }
evo: "Evolutionary"            { style.fill: "#457b9d"; style.font-color: "#fff" }
rl: "Reinforcement"           { style.fill: "#457b9d"; style.font-color: "#fff" }
hib: "Hybrid\n/ Neuro-symbolic" { style.fill: "#457b9d"; style.font-color: "#fff" }

IA -> sim
IA -> con
IA -> prob
IA -> evo
IA -> rl
IA -> hib

con -> "LLM": Transformer
con -> "Diffusion": U-Net/DiT
con -> "Vision (CNN/ViT)"
con -> "Audio/Music"
con -> "Video"
con -> "Multimodal"
sim -> "Expert system"
sim -> "Planner/search"
prob -> "Bayesian network"
evo -> "Genetic alg."
rl -> "Deep RL"
hib -> "Agent / RAG"

🎨 Figure F-II.1Mural of cards. Brief: a "collection gallery" style mural, 6 columns (one per paradigm), each column with stacked cards; each card with an icon, a name, and a color band for its modality. Sticker-collection / visual-encyclopedia style. Compendium palette; each column header in the color #1d3557.


Mural of AI types

Anatomy of a card

Every card follows the same template (rendered as a card on web/PDF):

<icon> Type name Paradigm · Modality · Architecture

  • What it is: one-sentence definition.
  • Examples: recognizable real systems.
  • Algorithmic basis: the core algorithm/architecture.
  • Peak: period of greatest prominence.
  • Capabilities / modes: what it knows how to do.
  • → Part IV: link to the chapter (planned).

The six galleries

Gallery (paradigm) Content Cards
01-symbolic.kmd Expert systems, planning, logic, knowledge graphs 4
02-connectionist.kmd Deep learning: LLMs, diffusion, vision, audio, video, multimodal, GNN, SSM, MoE 14
03-probabilistic.kmd Bayesian networks, HMM, Gaussian processes, classical ML 4
04-evolutionary.kmd Genetic algorithms, neuroevolution, swarm intelligence 3
05-reinforcement.kmd Deep RL, AlphaZero, RLHF, control 4
06-hybrid-neuro-symbolic.kmd Agents, RAG, neuro-symbolic, recommendation, embodied robotics 5

The panel does not exhaust the combinations of the taxonomy — it highlights the most recognizable types. Part IV may open chapters for intersections not covered here (every gap in the panel is a candidate for a chapter).