Understanding AI Infrastructure:
A Map of Inquiry
Eight domains of inquiry, organized so that each conversation thread can stay focused without losing sight of the larger picture. Items marked โ Covered have been addressed in earlier sessions.
How to use this guide
Use this as a checklist before starting a new conversation thread. Pick a domain, identify where you left off, and begin there. Ask for a summary at any point to capture what's been established before the thread grows too long.
Domain Key
Physical / Infrastructure
Political / Regulatory
Technical / How It Works
Societal / Human Impact
Economic / Financial
01
Data Centers: Physical Reality
Physical
โWhere are data centers being built โ Pacific Northwest, California, nationally? โ Covered
โWhat are data centers actually used for โ AI, cloud, surveillance, entertainment? โ Covered
โWater cooling: the problem and the alternatives โ Covered
โPower consumption: how much electricity does the AI buildout require?
โLand use: how large are these facilities, and what do they displace?
โWho owns what โ Microsoft, Google, Amazon, Meta footprints?
โUndersea and modular data centers โ experimental or viable?
โWaste heat reuse: capturing and redirecting thermal output
02
Energy & Environment
Physical
โWhat energy sources power data centers โ coal, nuclear, hydro, renewables?
โHow does AI power demand interact with regional power grids?
โCarbon footprint of training vs. running AI models
โNuclear revival: are data centers driving new reactor investment?
โHydropower dependency in the Pacific Northwest โ vulnerabilities?
โWhat happens to power grids when AI demand spikes?
โWater table and watershed impacts of large-scale cooling operations
03
How AI Actually Works
Technical
โWhat is a large language model and how is it trained?
โTraining vs. inference โ what's the difference in resource terms?
โWhat are GPUs and why does AI require so many of them?
โWhat does it mean for a model to "hallucinate"?
โOpen-source vs. proprietary AI โ what's at stake?
โAI agents: what are they and how do they differ from chatbots?
โHow do AI companies improve their models over time?
04
Surveillance, Privacy & Data
Societal
โWhat personal data is collected and stored โ and by whom?
โHow is behavioral data used for advertising and profiling?
โGovernment access to data center contents โ legal frameworks
โFacial recognition and AI-enabled surveillance infrastructure
โWhat protections exist โ GDPR, CCPA, federal gaps?
โChina's AI surveillance model vs. Western approaches
โBiometric data: what's being collected without clear consent?
05
Politics, Policy & Regulation
Political
โWhich government entities govern data center approval โ local to federal? โ Covered
โWhat questions should voters ask candidates at each level of government? โ Covered
โFederal AI legislation โ what exists, what's proposed?
โState-level AI laws โ California, Washington, Oregon leading?
โTax incentives for data center construction โ who benefits?
โThe EU AI Act โ does it affect U.S. companies?
โAI and antitrust โ are a few companies too dominant?
โWater rights law as applied to data centers
06
Labor, Economy & Displacement
Economic
โWhich jobs are most at risk from AI automation โ near term?
โWhich jobs and industries are being created by AI?
โData center construction: local employment vs. long-term local jobs
โWho profits from the AI buildout โ concentration of wealth?
โSmall businesses and AI: tool or threat?
โHow much does it cost to train a frontier AI model?
โAI and healthcare costs โ productivity vs. access
07
AI in Daily Life
Societal
โAI in healthcare: diagnostics, drug discovery, patient records
โAI and education: tutoring, cheating, institutional response
โAI in creative fields: art, music, writing โ tool or replacement?
โAI and accessibility โ benefits for people with disabilities
โAI and misinformation โ deepfakes, synthetic media
โSocial media algorithms as a form of AI โ how they shape behavior
โAI companions and mental health โ benefits and concerns
08
The Bigger Picture: Risk & Future
Societal
โWhat is AGI (Artificial General Intelligence) and how close is it?
โAI safety โ what are researchers actually worried about?
โConcentration of AI power: a few companies vs. global access
โAI and democracy โ election interference, propaganda at scale
โInternational AI race โ U.S. vs. China, strategic implications
โWhat does a world with abundant AI look like in 10โ20 years?
โWho decides how AI is governed โ corporations, governments, or publics?
On Thread Strategy
Physical and environmental topics (Sections 01โ02) work well together in one thread. Political and regulatory topics (05) can be woven in when they directly arise from a physical question, or kept separate for focused policy discussions. Sections 03 and 08 are good starting points for any conversation about the technology itself. At any point, ask for a summary to capture what's been established before starting a new session.