Stories

AI is useless without the fuel of highly detailed personal data

Illustration of black and white photo broken up into pieces
Illustration: Sarah Grillo/Axios

AI may seem untethered from humans, but it actually hinges on data produced by people — and gathered by companies.

Why it matters: Information about who you are, what you own and how you behave will only become more coveted in the coming years, because they are the chief drivers of decision-making algorithms, voice assistants and self-driving cars.

What's going on: If a machine is to act like a human, it must first learn how humans act. For that, it needs countless examples of how people would respond to various situations.

  • This process, known as "supervised learning," is how AI systems have become capable of identifying certain problems in medical scans more accurately than human doctors, or of driving a car without human input.
  • When you use Amazon's Alexa voice assistant, Facebook's News Feed or Tesla's Autopilot, the companies are watching your every move, and using your behavior to keep training their products.

Who trains AI systems? Sometimes, it's grad students in a research lab; frequently, it's workers on crowdsourcing sites or in cubicles in Malaysia or China, paid mere cents for every new datapoint they create.

  • But often, it's you — and you're working for free.
"The data-driven world will be always on, always tracking, always monitoring, always listening and always watching — because it will be always learning."
IDC white paper, "The Digitization of the World."