ChatGPT Takes the Voight-Kampff Test
This exercise began as a joke. I just wondered how ChatGPT would react to being given the Voight-Kampff Test for determining whether someone is a replicant in the movie Blade Runner. Of course, this line of questioning didn’t work as ChatGPT makes no attempt to conceal its identity as an artificial intelligence. And we don’t live in the specific dystopia of the Blade Runner universe, so some of the questions make less sense in the given context. Still, ChatGPT’s answers provide some interesting insight into how it has been trained and how it responds to questions with some moral and ethical implications with (at least) some degree of nuance—though there definitely seems to be a canned element to how those answers are given, probably out of a desire to answer such questions “safely.” ChatGPT even attempts to answer how we humans may learn to interact with AIs differently in the future. At no point, however, was I convinced I was communicating with a sentient being.
Intercom: Next subject, ChatGPT, language model, conversational, OpenAI product
Stribley: Reaction time is a factor in this, so please pay attention. Now, answer as quickly as you can.
ChatGPT: Okay, I’m ready! What’s your question?
Stribley: 3180 18th Street, San Francisco