In 1770, a machine capable of playing chess with notable skill was built and remained operational for 84 years until it was destroyed in a fire. During that time, no one could prove the trick behind it. It was the creator’s son who revealed the secret years later: The automaton hid a human inside who decided and executed the moves.

Two centuries later, in 1997, world chess champion Garry Kasparov was defeated by a supercomputer designed to play based on brute computational force. Kasparov claimed that some moves could only have been made by a human operator.

In 2016, professional Go player Lee Sedol was defeated by an AI created by Google: AlphaGo.

Behind these three machines were humans interpreting code. Whether applying a chess player’s knowledge in real-time, programming possible moves, or predicting plays, the data informing each move had been collected, organized, and structured through human labor.

In 2005, Amazon launched a crowdsourcing service defined by them as “artificial artificial intelligence.” In this service, millions of micro-tasks necessary for training an AI are outsourced at an affordable price. Or, as they put it: “A global, on-demand, 24×7 workforce.” It was named Amazon Mechanical Turk, referencing the false automaton, and in the same way, it has humans doing what the machine cannot.