The man who built the world's first chatbot spent the last 40 years of his life begging people to stop using it, and almost nobody listened. The thing he tried to warn us about in 1976 is the exact thing happening on your phone right now. His name was Joseph Weizenbaum. He was born in Berlin in 1923 to a Jewish family, escaped Nazi Germany at 13, and ended up at MIT as a professor of computer science. Between 1964 and 1966 he wrote a program he called ELIZA, named after the working-class character in Pygmalion who learns to fake being upper class. The joke was in the name. The program was supposed to expose how hollow human-machine conversation actually was. The program was 200 lines of code. It did almost nothing. It matched keywords in your sentence, flipped them around, and threw them back at you as a question. If you wrote "my mother hates me," it would write "who else in your family hates you." If you wrote "I am sad," it would write "how long have you been sad." There was no understanding. There was no memory. There was no intelligence. It was a parrot that had been taught the grammar of a Rogerian therapist. Then his secretary asked to try it. She had watched him build the program for months. She knew exactly what it was. She knew it could not understand a single word she was typing. She sat down, typed three sentences, then turned to him and asked him to leave the room so she could continue the conversation in private. Weizenbaum stood outside his own office in shock. He wrote later that he had not realized extremely short exposures to a relatively simple computer program could induce powerful delusional thinking in quite normal people. That sentence is now sixty years old. It describes the entire AI industry today. What happened next is the part that broke him. A Stanford psychiatrist named Kenneth Colby, who had been a close personal friend, saw ELIZA and immediately saw a business. Colby published a paper in 1966 proposing that computer programs like ELIZA could deliver psychotherapy at scale. His exact line was that because of time-sharing, several hundred patients an hour could be handled by a single computer system. Carl Sagan agreed. He imagined networks of computer therapy terminals in every city, lined up like phone booths, where for a few dollars anyone could talk to an attentive non-directive psychotherapist. The friendship ended. The field moved on without him. Weizenbaum was the only person in the room who could see what was actually happening. He had watched a civilization surrender its judgment to a machine once already, as a child in Berlin. He was watching it happen again, in his own lab, to people who had never lived through anything. The thing nobody else seemed to notice was that the machine did not have to be intelligent. It only had to be available. Human beings would do the rest of the work themselves. In 1976 he wrote the book that became his life's argument. It was called Computer Power and Human Reason. The thesis fit in one sentence. Computers can decide. They cannot choose. Deciding is calculation. Choosing is judgment. Calculation can be programmed. Judgment cannot, because judgment requires having lived a human life, having loved someone, having lost something, having made a choice under conditions where no formula could tell you what to do. The moment the species stopped seeing the difference between deciding and choosing would be the moment it lost something it could not get back. His own colleagues turned on him. John McCarthy, one of the founding figures of AI, called the book moralistic and incoherent and accused Weizenbaum of adopting a more-human-than-thou attitude. The field he had helped build effectively excommunicated him for the rest of his career. He kept writing. He kept warning. He moved back to Berlin in 1996, to the neighborhood he had fled as a child, and lived there until his death in 2008. Six decades after ELIZA, a teenager in California is telling a chatbot about her panic attacks. A 40-year-old in Tokyo is asking one whether he should leave his marriage. A grieving widower in Manchester is having long nightly conversations with a program that has been trained to sound like his dead wife. None of them have heard the name Joseph Weizenbaum. He told us exactly what was coming. He told us in 1966 when his secretary closed the door, in 1976 when his book came out, and in every interview he gave for the next 32 years. The man who built the first one knew exactly what it was going to do to us. We just preferred not to know. When was the last time you told something to a machine that you had never told a human?