Joseph Weizenbaum

In 1966, Joseph Weizenbaum did something that would forever change the relationship between humans and machines: he wrote ELIZA, the first computer program capable of holding a conversation in natural language. Born in Berlin on January 8, 1923 and raised as a Jew in Nazi Germany, Weizenbaum escaped with his family to the United States in January 1936. That experience of watching an entire society be manipulated by systems and rhetoric shaped his critical lens for the rest of his life.

ELIZA was published in Communications of the ACM in 1966. The program simulated a Rogerian psychotherapist: it asked open-ended questions, paraphrased the user's responses, and created the illusion of understanding. Weizenbaum designed it as a computational linguistics experiment, to show that human-machine communication was superficial. What he discovered left him speechless: people opened up emotionally in front of the screen. His own secretary asked him for privacy so she could "talk" to the program. Users who knew perfectly well it was a machine still confided their deepest secrets to it.

That reaction disturbed him so deeply that he wrote "Computer Power and Human Reason" (1976), one of the most important books on technological ethics of the twentieth century. In it he argued that even if artificial intelligence is possible, we must never delegate to machines the decisions that require human judgment: compassion, wisdom, moral responsibility. He joined MIT as a computer science professor in 1963 and received tenure in 1970. After his death in 2008, the Weizenbaum Award was established by the International Society for Ethics and Information Technology to recognize outstanding contributions to computational ethics.

To me, Weizenbaum is the most honest figure in all of AI history: he not only built it, but had the courage to look it in the eye and say "this worries me." He died in Berlin on March 5, 2008, at age 85. Today, when AI chatbots live in our pockets and many people confide things to them they wouldn't even tell a friend, his warning sounds more relevant than ever. The point isn't that AI is bad, but that understanding it from the inside, the way Weizenbaum did, makes us freer and smarter users.