OpenAI's success built on exploitation, says new book
OpenAI, the company behind ChatGPT, has been called out for building its $157 billion empire by tapping into global resources and cheap labor.
Karen Hao's book, Empire of AI, says its powerful AI models are trained on massive amounts of internet data, often collected without clear permission, and that its rapid growth has come at a cost to workers and the environment.
AI model's training data often collected without consent
To make bigger and better models, OpenAI relies on "scaling laws": basically, more data equals better AI.
But this approach has led to questionable practices: in Venezuela, annotators were paid as little as $6 a week after cuts; in Kenya and North Africa, some workers lost jobs as companies tried to save money.
Critics say this model often overlooks fair pay and consent.
Environmental concerns over AI model training
Training these huge AIs uses tons of water and energy, especially in places like Chile, raising big environmental concerns.
Plus, since a few tech giants control much of the training data, there are worries about who really owns our digital world and how much power these companies have over information that shapes society.