Talos, the bronze giant of Greek mythology, was an early ancestor of artificial intelligence. Shaped like a man, Talos circled the island of Crete three times a day at human command, faithfully serving as a sentinel. It was a benevolent AI, entrusted with defense and the well-being of the people. But as time passed — as bronze gathers rust — people began to alter the functions they imagined for AI. They cast it as an evil monster that attacks and harms humans.
The golem, handed down in ancient Judaism, is the best-known example. As Jews suffered humiliation and persecution, a rabbi molded a golem from clay and carved the word "emet," meaning truth, on its forehead. The golem did the hard labor its human masters assigned and stood guard over the community against outside intruders. But the golem, which could only follow orders and had no capacity to judge situations, began to threaten the humans around it. In the end, the rabbi erased the letters for "truth" from its forehead, took away its life and returned it to clay. Perhaps it was a weighty warning that human technology, however advanced, can never become absolute truth.
The eras and countries differed, but the AI apocalypse story continued. In "R.U.R.," published in 1920, the Czech playwright Karel Capek depicted the tragedy of AI laborers built to work in place of humans who ultimately kill their makers and rise in revolt. In the science fiction film "2001: A Space Odyssey," director Stanley Kubrick told the chilling story of HAL, an AI computer that begins to think for itself aboard a spacecraft bound for Jupiter and starts eliminating the human crew. It is striking that people across time and place first hoped for a benevolent AI, then foresaw the arrival of a malevolent one as the technology advanced.
The heads of four global AI companies, including Anthropic and OpenAI, argued on the 13th that the pace of AI development should be slowed. They also issued a chilling warning, putting the risk of an outcome that kills all of humanity within the next decade at 10%. It sounds like a desperate confession that if the AI race succumbs to blind faith in technology, it could bring about the golem's revenge.
