The Inventor of the Chatbot Tried to Warn Us About A.I.

Technological history may not repeat, but it occasionally rhymes. Last September, ChatGPT, the popular generative A.I. program made by OpenAI, released a voice mode that allowed users to talk to the software. ChatGPT then “talks” back. A few days later, OpenAI’s head of safety posted on X about a surprisingly moving interaction she had with the program:

Just had a quite emotional, personal conversation w/ ChatGPT in voice mode, talking about stress, work-life balance. Interestingly I felt heard & warm. Never tried therapy before but this is probably it? Try it especially if you usually just use it as a productivity tool.

The short post was a rich text, ripe for satirical interpretation—the highly paid, emotionally obtuse Silicon Valley engineer who isn’t familiar with therapy but thinks she’s encountered a worthy simulation of it in the notoriously bug-ridden product her employer produces. Where to begin?

Some observers pointed to the example of ELIZA, a famous early chatbot released in 1966 by MIT computer scientist Joseph Weizenbaum. Weizenbaum named his chat program after Pygmalion’s Eliza Doolittle, a working-class woman who passes herself off as an aristocrat after some elocution lessons. Designed to resemble a therapist, the software used a canny set of rules and a therapist’s dialogue script to simulate conversation. The program borrowed a technique from Rogerian psychotherapy by often restating a user’s remark as a question. For example, a user would write that they were unhappy and ELIZA asked why they were unhappy. “This was an elegant way to create the effect of a computer holding its own in a conversation with the user,” observed the writer and game designer Matthew Seiji Burns.

Leave a Comment