Method

This poster is an educational summary. Its goal is to be accurate without pretending to be comprehensive. The method below describes how claims were chosen, how uncertainty was handled, and what constraints were applied.

1) Source selection

2) Claim shaping

Claims were written to avoid false precision. When a quantity varies widely across model size, hardware, or location, the language reflects that variability (e.g., “can consume significant electricity” rather than “always consumes X”).

Where possible, the poster uses broad, widely cited estimates (for example, the share of global electricity used by data centers) and avoids values that depend on proprietary model details unless the uncertainty is explicit.

3) What “rigorous” means here

4) Limitations

AI infrastructure changes quickly. Electricity mixes change, hardware efficiency improves, and reported values depend on definitions (what counts as “AI workload,” what counts as “data center energy,” what boundaries are used for “water footprint,” and so on).

This site is designed to be updateable. If sources change or better measurements become available, the Changelog should record updates.

5) How to verify

  1. Identify a claim on the poster.
  2. Find its bracketed reference number(s), e.g., [1][3].
  3. Open the matching entry in Sources.
  4. Cross-check the claim against the primary document.

Archival Transcript

This site was developed through an extended dialogue between a human steward and an artificial intelligence system. The conversation itself is part of the method. It documents how claims were evaluated, corrected, and made traceable.

The full archival transcript is available here:

View the complete HTML transcript of the dialogue

This transcript preserves the structural development of the poster, the verification chain, and the creation of this site. It exists as a permanent record of methodological transparency.