ELIZA

ELIZA had about 200 lines of pattern-matching rules. If a sentence contained "my mother", it might reply "Tell me more about your family." If nothing matched, it said "Please go on." Joseph Weizenbaum wrote it in 1964 and 1965 on MIT's time-sharing system and published it in January 1966, choosing the persona of a Rogerian psychotherapist because that style of conversation, reflecting the patient's words back, was one a program with no knowledge could sustain.

He expected it to be understood as a parlour trick. Instead his secretary, who knew exactly what it was, asked him to leave the room so she could talk to it in private. Psychiatrists proposed using it for treatment. People attributed to it understanding and care. Weizenbaum named this the ELIZA effect and wrote a book in 1976, Computer Power and Human Reason, arguing that there were things computers should not be allowed to do regardless of whether they could.

Every chatbot since carries the same two lessons. Very little machinery is needed to make a conversation feel real to the person in it, and the person's projection does most of the work. ChatGPT, fifty-six years later, was a system that finally had knowledge behind the reflection, and the effect was the same, at planetary scale.