Title: The Triumph of Psychotherapy: ELIZA, the AI Chatbot Holding its Ground in the Turing Test
In 1966, renowned sociologist and critic Philip Rieff published The Triumph of the Therapeutic, shedding light on the pervasive influence of psychotherapy on modern Western culture. Coincidentally, that same year, computer scientist Joseph Weizenbaum introduced ELIZA in a research paper titled ELIZA – A Computer Program For the Study of Natural Language Communication Between Man and Machine published in the journal Communications of the Association for Computing Machinery. ELIZA, the earliest version of what we now call a chatbot, is famous for responding to user input in a nonjudgmental, therapist-like manner.
Even in the 1980s, ELIZA continued to garner interest, as seen in a television clip where its narrator remarks on its seemingly understanding replies. Despite users being fully aware that ELIZA had no actual understanding of their words, some interactions with the chatbot became emotionally charged. This led to ELIZA passing a kind of Turing test, a test proposed by computer scientist Alan Turing to determine if a computer can generate output indistinguishable from human communication.
Surprisingly, even six decades after Weizenbaum’s development of ELIZA, the chatbot still holds its own. A recent preprint research paper titled Does GPT-4 Pass the Turing Test? by researchers from UC San Diego compared OpenAI’s GPT-4 AI language model, human participants, GPT-3.5, and ELIZA to determine their success in tricking participants into thinking they were interacting with a human. The study revealed that human participants correctly identified humans in only 63% of the interactions, with ELIZA surpassing the AI model powering the free version of ChatGPT by reflecting users’ input back at them.
While this doesn’t imply that ChatGPT users should revert to Weizenbaum’s simple novelty program, it does suggest the value in revisiting his subsequent thoughts on artificial intelligence. Weizenbaum later condemned the worldview of his colleagues and warned of the dangers posed by their work, viewing artificial intelligence as an index of the insanity of our world. As early as 1967, he argued that no computer could fully understand a human, and he even went further to claim that no human could fully understand another human—a proposition seemingly supported by the extensive history of psychotherapy.
So, while ELIZA maintains its relevancy and success in the Turing test, it is crucial to consider the broader implications and limitations of artificial intelligence. Weizenbaum’s skepticism prompts us to question the potential consequences and complexities that arise when attempting to replicate human understanding and connection through technology.
Related content:
– A New Course Teaches You How to Harness the Powers of ChatGPT for Productivity
– Thanks to Artificial Intelligence, Historic Figures Come to Life in Conversations: Shakespeare, Einstein, Austen, Socrates & More
– Noam Chomsky Critiques ChatGPT: Deeming it High-Tech Plagiarism and a Way of Avoiding Genuine Learning
– What Happens When Crochet Enthusiasts Craft Stuffed Animals with Instructions from ChatGPT
