Atlas of AI

Power, Politics, and the Planetary Costs of Artificial Intelligence

by Kate Crawford

“Atlas of AI” (2021) uncovers how AI relies on the extraction of resources, labor, and data. The book portrays AI as a global system that is pushing society towards centralized political power and undemocratic governance.

About the Author

Kate Crawford is an accomplished author and researcher focused on the social impact of AI. She has held academic and research positions at the USC Annenberg School, Microsoft Research, and the École Normale Supérieure.

Uncovering the Hidden Costs of Artificial Intelligence

Picture a sprawling network of mines, factories, and data centers stretching from the Nevada desert to the mountains of Inner Mongolia. This is the hidden backbone of artificial intelligence, a technological system fueled by the extraction of minerals, data, and human labor.

In this summary, we critically examine the real costs and ethical implications of the AI boom. By mapping out the intricate network behind AI development, we expose the often-ignored material realities of this seemingly intangible technology.

Artificial Hype?

What does a German horse have to do with artificial intelligence? Let's talk about Clever Hans. In the late 19th century, this Orlov Trotter fascinated crowds across Europe with his seemingly remarkable intelligence. Clever Hans could tell time, identify dates, distinguish musical tones, and even solve math problems by tapping his hoof. Or so people thought.

The reality of Clever Hans' abilities was revealed by psychologist Oskar Pfungst. He found that Hans wasn't actually solving problems but was instead reacting to subtle, unintentional cues from his handlers. These cues, such as changes in posture, breathing, and facial expressions, unconsciously signaled to Hans when he had the right answer. This phenomenon, known as the observer-expectancy effect or Clever Hans Effect, shows how easily biases can skew results and lead to incorrect conclusions.

The story of Clever Hans is a warning about the risks of attributing human traits to nonhuman entities and the importance of recognizing our biases.

AI enthusiasts believe that machines can replicate human intelligence. However, the author contends this is a flawed assumption. AI systems do not possess the ability for independent reasoning or understanding. Instead, they rely on extensive training with large datasets and follow predefined rules to execute specific tasks. Their outputs are influenced by the biases and intentions of their human creators. Additionally, they lack the contextual understanding, flexibility, and adaptability that characterize human intelligence.