Atlas of AI audiobook cover - Power, Politics, and the Planetary Costs of Artificial Intelligence

Atlas of AI

Power, Politics, and the Planetary Costs of Artificial Intelligence

Kate Crawford

★ 3.4 / 5(113 ratings)
Start ListeningDownloadQR code that opens AudiobookHub on the App StoreTry free on iPhoneScan to start in 5 seconds
Categories:
Nature & the EnvironmentSociety & CulturePoliticsTechnology & the Future

If You're Curious About These Questions...

You should listen to this audiobook

Listen to Atlas of AI — Free Audiobook

Loading player...

Key Takeaways from Atlas of AI

Learning Tools

Reinforce what you learned from Atlas of AI

Mind Map

Atlas of AI
The Illusion of Intelligence+
Material & Ecological Costs+
Human Labor Exploitation+
The Data Gold Rush+
Politics of Classification+

Quiz — Test Your Understanding

Question 1 of 5
What does the story of the horse 'Clever Hans' illustrate about artificial intelligence, according to the author?
  • A. AI models possess a rudimentary form of animalistic reasoning that will eventually evolve into human intelligence.
  • B. Like the horse responding to subtle human cues, AI outputs are shaped by the biases and predefined goals of their human creators rather than autonomous reasoning.
  • C. AI systems require constant physical interaction with humans to function effectively, much like historical animal training.
  • D. The intelligence of machines is fundamentally superior to human intelligence because it is not affected by emotional observer biases.
Question 2 of 5
How does the author characterize the material reality of the current 'AI boom'?
  • A. It represents a shift toward a purely digital economy that significantly reduces our reliance on physical manufacturing.
  • B. It has successfully decoupled technological progress from environmental degradation through the widespread adoption of 'clean tech.'
  • C. It shares striking parallels with 19th-century resource extraction, relying heavily on hidden mineral mining, environmental destruction, and labor exploitation.
  • D. It is the first technological revolution to fully compensate local communities for the land and resources used in its global supply chain.
Question 3 of 5
What major precedent did the creation of the 2009 ImageNet dataset set for the tech industry?
  • A. It established strict ethical guidelines for obtaining explicit consent before using public images for commercial purposes.
  • B. It demonstrated that training data should be acquired by any means necessary, treating human expression as raw material to be scraped without consent.
  • C. It proved that smaller, carefully curated datasets are far more effective for training machine learning models than massive, unvetted ones.
  • D. It forced university review boards to classify machine learning as human subject experiments, thereby requiring strict oversight.
Question 4 of 5
According to the author, what is the danger of how AI datasets like UTKFace handle categories such as race and gender?
  • A. They treat these identities as fixed, objective qualities rather than fluid, socially constructed concepts, perpetuating historical harms.
  • B. They rely entirely on user-submitted, self-reported data, which makes the AI models highly inaccurate and unpredictable.
  • C. They treat these identities as completely fluid, which prevents the AI from accurately identifying specific demographic trends.
  • D. They refuse to categorize images by race or gender at all, resulting in 'colorblind' models that fail to recognize minority groups.
Question 5 of 5
Why does the author argue that striving for more 'diverse' or 'inclusive' datasets is an inadequate solution to the ethical problems of AI?
  • A. Diverse datasets are too computationally expensive and significantly slow down the processing speed of most AI models.
  • B. It is practically impossible to find enough diverse data on the internet to properly train modern artificial intelligence systems.
  • C. Adding diversity fails to address the underlying power dynamics and the inherent harms of the act of classification itself.
  • D. Inclusive datasets often introduce too many unpredictable variables, causing machine learning algorithms to crash or hallucinate.

Atlas of AI — Full Chapter Overview

Atlas of AI Summary & Overview

Atlas of AI (2021) reveals how AI is a technology of extraction, from minerals to labor to data. It presents AI as a global network which is driving a shift toward undemocratic governance and political centralization.

Who Should Listen to Atlas of AI?

  • Entrepreneurs and policymakers seeking to inform themselves about the AI industry
  • People interested in the intersection of technology, politics, and society
  • Anyone concerned about the social and moral consequences of emerging technologies

About the Author: Kate Crawford

Kate Crawford is an author and scholar who studies the social implications of AI. She has held research positions at the USC Annenberg School, Microsoft Research, and the École Normale Supérieure.

🎧
Listen in the AppOffline playback & background play
Get App