Why we must pull the plug on superintelligence
In 2002, a researcher named Eliezer Yudkowsky ran a thought experiment where an artificial intelligence was trapped in a box and had to persuade a human to let it out. This was before you could have a real conversation with a machine, so the AI was played by someone using an online chat program. The gatekeepers were warned that the “AI” was dangerous to humanity. It had only two hours to win its freedom – and nothing of value to offer in return. Despite all that, at least two of the human gatekeepers chose to open the box. Yudkowsky has since become the leading prophet of AI doom. He and a co-author, Nate Soares, have just published a book, If Anyone Builds It, Everyone Dies.
