Kate Crawford: Mapping Empires

Summary

This Long Now talk presents Kate Crawford’s argument that AI is not immaterial intelligence but a deeply material industry organized through data extraction, minerals, energy, water, labor, representation, and waste. Crawford uses Calculating Empires, Renaissance perspective, guano, gutta-percha, lithium, data centers, AI Slop, and Model Collapse to frame AI as a new phase in older imperial patterns of enclosure and externalized cost. The Q&A with Kevin Kelly turns the critique toward Public Interest AI: public accountability, consent-based data, renewable energy, and narrower use where AI demonstrably works.

Key Claims

• Linear perspective was a constructed technology of truth that helped make the world measurable, controllable, and commodifiable.
• AI shifts representational power again by training machine vision and language systems on scraped images, text, and sensor data rather than on a single human observer.
• AI Metabolic Infrastructure makes AI’s material costs central: data, power, water, minerals, land, labor, carbon, and e-waste are part of the system rather than side effects.
• The AI data appetite extends older enclosure patterns from land and resources into culture, bodies, gestures, attention, and everyday physical spaces.
• Commercially biased datasets make models see through online-commerce traces, so generated outputs can inherit platform incentives and market overrepresentation.
• Critical Minerals Geopolitics is also an ecological problem when lithium, cobalt, copper, and rare earths are extracted from places such as Chile for short-lived AI hardware.
• Data-center power and water demand connect AI infrastructure to Data Center Thermal Management, Jevons Paradox In AI, environmental justice, and Data Center Backlash.
• AI Slop and political slopaganda show that low-cost generation is becoming a media economy, not just an aesthetic accident.
• Model Collapse names the risk that repeated training on synthetic outputs flattens diversity, erases outliers, and degrades model quality.
• Better AI futures require Public Interest AI, democratic oversight, renewable energy, consent-based data, public AI literacy, and more precise use of AI in domains where it has evidence of value.

Key Quotes

“There’s no data like more data.” - Robert Mercer phrase Crawford cites for the statistical-AI data appetite.

“works for whom” - Crawford’s framing question for claims that AI benefits society.

“AI slop” - Crawford’s label for low-effort synthetic media and its emerging visual language.

Connections

• Kate Crawford - speaker and source of the historical-material AI critique.
• Rebecca Lendl - Long Now host introducing the talk.
• Kevin Kelly - Q&A interlocutor pressing on public AI, sovereign AI, capitalism, and better uses.
• Long Now - host context; this source adds AI materiality, empire, and infrastructure politics to the Long Now branch.
• Calculating Empires - Crawford’s timeline project mapping communication, computation, classification, and control since 1500.
• AI Metabolic Infrastructure, AI Slop, Model Collapse, and Public Interest AI - main concept cluster added by the source.
• Data Center Backlash, Data Center Thermal Management, Jevons Paradox In AI, Critical Minerals Geopolitics, and AI Compute Continuity - existing AI infrastructure branch extended by the source.
• AI Content Devaluation, AI Content Provenance, AI Recognition Bias, AI Literacy Against Worship, and Human Judgment Under AI - media, bias, literacy, and judgment branch connected by the source.
• xAI and Elon Musk - referenced through Crawford’s South Memphis data-center pollution example.

Contradictions

• No direct contradiction found. The source qualifies the wiki’s AI-infrastructure branch by arguing that energy, water, minerals, and local environmental burdens are not merely scaling constraints but political and ecological costs.

Source Notes

• Ingested from the 02025-crawford-podcast-v3 Markdown export in the podcastatlas episode corpus.