Featured in Ripley's Believe It or Not!
You’ve probably heard of ChatGPT, a computer program that is trained to follow your instruction and provide a variety of wide ranging responses. As someone that has spent some time actually using the AI, I have to say, the results it produces can be eerily human.
But did you know that computer scientists have been working alongside chatbots as early as the 1960s? Today, we are rewinding back and looking at the revolutionary chatbot of the ‘60s known formally as ELIZA. We’re going to be taking a closer look at the curious world of chatbots and examining the existence of ELIZA and the influences it had on ChatGPT as we know it.
It was the late 1960’s and Joseph Weizenbaum, an MIT computer scientist, had just completed work on his revolutionary chatbot ELIZA. Weizenbaum was born in Berlin, Germany, in 1923, and fled the country with his family in 1935 to escape the political turmoil. Weizenbaum came to the United States where his road to computer science would ultimately begin.
After time spent in the Air Force, Weizenbaum would go on to study as a computer scientist and eventually work in the industry. You have to remember, computers at that time were not portable devices that could fit in our pockets. In fact, they often barely fit into a room!
As an associate Professor at MIT, Weizenbaum became obsessed with the way computers could directly interact with humans through language. It was this early through line between computers and human language that would work to lay the foundation for his own chatbot and eventually lay the groundwork for the AI development of programs such as ChatGPT, Siri and Alexa.
Weizenbaum believed that the most effective way to go about creating a connection between human language and computers would be through mimicking the structure of a therapy session. The general idea would be to have ELIZA fixate on key words and continually ask open ended questions through a reformatting of the initial remark.
Eliza was completed in 1966 and Weizenbaum offered MIT students the opportunity to interact with the chatbot. This process consisted of messages typed into the computer by students and responses would then be provided by ELIZA and were routed to an electronic typewriter and printer.
Eliza was programmed to fixate on key words such as “girlfriend,” “depressed,” “what,” “mother,” and “father.” In the event that ELIZA had trouble interpreting what the user was asking, the program could fall back on “Please go on,” “That’s very interesting,” or “I see.” Although non-specific, in many ways the phrases helped to create the illusion that one was interacting with a human while simultaneously working to keep a flow in conversation.
Believe It or Not!, throughout the process, students were told not to use question marks when asking questions as ELIZA interpreted them as a line delete request!
Weizenbaum was initially happy with the response that was garnered from users’ experience with ELIZA but there was one thing he did begin to notice that he viewed as considerably concerning. Overtime, Weizenbaum made note of users starting to divulge deep personal information, looking for help similar to that of a therapy session.
This observation ended up pushing Weizenbaum to advocate for caution when relying too heavily on computers for human thought.
Seeing the mixture of responses users had towards ELIZA and how quickly some were able to conflate a computer program to human interaction made Weizenbaum weary about the idea of computers being capable of human thought and emotion. Eliza’s one dimensional interaction with PARRY further solidified Weizenbaum’s concerns.
Coming back to 2023, I can’t help but think about what Weizenbaum would’ve thought regarding ChatGPT. I want to leave you with a quote Weizenbaum made to The New York Times regarding computers replacing human intelligence:
“There are aspects to human life that a computer cannot understand—cannot. It’s necessary to be a human being. Love and loneliness have to do with the deepest consequences of our biological constitution. That kind of understanding is in principle impossible for the computer.”
In the age of advanced AI with ChatGPT, I think Weizenbaum would stand by his early warnings and continue to advocate caution as we barge full steam ahead into AI.
So there you have it – from ELIZA, to ChatGPT, humans have continued to fixate on advancing computer technology to eerie levels. Have you had any experience using Chatgpt? Let us know in the comments below!
Source: Meet ELIZA: The 60s Chatbot That Set The Groundwork For ChatGPT
