SUPPORTING REFERENCE / CONCEPTUAL RELATIONSHIPS
Understand the connections.
This reference explains selected dependencies and their evidence. It complements the object-based AI Map.
Return to the 3D AI Map →
Read each arrow as: subject → relationship → object. Select an area to see the exact meaning and source. On smaller screens, use the controls and reading directory.
HOW TO READ LEVEL 0
Overview first. Depth follows.
Level 0 shows the ecosystem. Level 1 will explain an area or system; Level 2 will inspect selected components or processes. A level describes explanatory depth, not physical size. Physical assemblies and conceptual dependencies use different views.
The connections are introductory, source-backed summaries. Sources support the underlying relationships; they do not endorse this ten-area arrangement. Map relationship review: November 1, 2026. The full chapter-by-chapter factual audit is still pending.
THE READING DIRECTORY
Take any path into the system.
People, purposes, and outcomes
What is the AI system meant to help someone do?
Applications, workflows, and agents
How does an AI capability become something people can use?
Models and model development
What is a model, and how is it developed?
Data and knowledge
Where does the information an AI system uses come from?
Computing software and services
What connects a model to the machines running it?
Chips and computing systems
What physical equipment does AI run on?
Data-center facilities and operations
What turns a building into a place for computing?
Electricity and supporting utilities
How does electricity reach an AI system?
Manufacturing and industrial supply chains
What does it take to make the equipment?
Organizations, investment, workforce, and governance
Who builds, operates and takes responsibility for AI?
Inspect the 16 relationship records and their sources
Source checks support introductory statements within the limits shown. This is not certification of every chapter or every possible deployment.
- People → uses → AI applications
People use an application to perform a task; its suitability depends on the use context.
NIST · AI Risk Management Framework ↗AI RMF: design, development, use and evaluation · checked November 1, 2026
- AI applications → can support → Human tasks
An application is intended to help with a task. This does not establish that its output is useful or correct.
NIST · AI Risk Management Framework ↗AI RMF: use and evaluation · checked November 1, 2026
- Model-based applications → uses → Models
This link concerns model-based AI applications. It does not imply that all software, or every approach to AI, requires a trained model.
Anthropic · Building effective agents ↗Building blocks, workflows, and agents · checked November 1, 2026
- Retrieval-enabled applications → can retrieve from → Information collections
Retrieval is optional: an application can search an external collection and supply selected information as context.
Anthropic · Building effective agents ↗The augmented LLM · checked November 1, 2026
- Machine-learning models → trained using → Training data
Training uses examples to adjust learned parameters. This is distinct from supplying context during inference.
Google · Machine Learning Glossary ↗training; training set · checked November 1, 2026
- Deployed models → served by → Serving software
Software makes model inference available to applications; deployment may be local or remote.
Google · Machine Learning Glossary ↗serving; inference · checked November 1, 2026
- Computing software → runs on → Computing hardware
Software execution depends on processors and other computing resources. Kubernetes is one orchestration example, not a required stack.
Kubernetes · Overview ↗Automatic bin packing; cluster nodes · checked November 1, 2026
- Computing systems → can be housed in → Data-center facilities
Data centers house computing equipment. AI can also run on phones, workstations and other devices outside a data center.
U.S. Department of Energy · Data center design guide (2024) ↗Section 2: Information Technology Systems · checked November 1, 2026
- Computing equipment → powered by → Electricity
Processors and supporting computing equipment need electrical power; facility overhead is a separate load.
U.S. Department of Energy · Data center design guide (2024) ↗Sections 2 and 6: IT and electrical systems · checked November 1, 2026
- Facility support systems → powered by → Electricity
Cooling and electrical distribution support the IT load and have their own operating requirements. Design details vary.
U.S. Department of Energy · Data center design guide (2024) ↗Sections 5 and 6: cooling and electrical systems · checked November 1, 2026
- Computing equipment → produced through → Manufacturing processes
Chip design, fabrication, packaging and system assembly involve different activities and inputs; this is not physical disassembly.
CSET / ETO · Semiconductor Supply Chain Dataset documentation ↗Overview; inputs; stages · checked November 1, 2026
- Data-center facilities → equipped through → Equipment supply chains
Facilities use manufactured electrical and cooling equipment. The cited design guide supports these equipment categories, not a supplier or origin claim.
U.S. Department of Energy · Data center design guide (2024) ↗Sections 5 and 6: equipment categories · checked November 1, 2026
- Responsible organizations → governs → AI applications
Organizations assign responsibilities and manage risks throughout application development and use.
NIST · AI Risk Management Framework ↗AI RMF: trustworthiness across the lifecycle · checked November 1, 2026
- Facility operators and workforce → operates → Facilities
People and organizations operate and maintain facilities. The subject area itself is not an operating company.
U.S. Department of Energy · Data center design guide (2024) ↗Section 2.1: expert staff and operations; section 8.9: management and monitoring · checked November 1, 2026
- Manufacturing organizations → supplies through → Production activities
Companies provide materials, equipment and processes in industrial supply chains; organization roles and production locations are distinct facts.
CSET / ETO · Semiconductor Supply Chain Dataset documentation ↗providers; provision; geographic limitations · checked November 1, 2026
- People in organizational roles → contributes to → AI decisions and oversight
People contribute through roles such as development, operation, evaluation and oversight. This does not imply that every affected person has equal decision power.
NIST · AI Risk Management Framework ↗AI RMF: design, development, use and evaluation · checked November 1, 2026