Post Six(A): The Turing Test

(Or, learning to ask for what you want).

Note: Apologies for the repost, it came to my attention that I didn’t have the image of the discussion sized correctly to be readable in the email version.

My post today is an exploration with ChatGPT of the famous Turing Test, along with some discussion of how modern LLMs fare against the test, and some talk of course, about what it all means. (What does it mean?!?)

The test, devised by Alan Turing in the mid-20th Century, was not intended to delve into whether a computer was or was not conscious, but instead whether or not a computer could generate responses sufficient to deceive humans into believing the computer was another human. Obviously, today, this has all kinds of implications for our computer dependent world, from customer service to political propaganda to deep fakes and more. . . I’m not going in that direction today though. Maybe at some point, but it seems like plenty of other people are already worried about, and discussing it. . . also, it’s just depressing. . . so let’s put a pin in that for now and stay in a more theoretical and (hopefully?) less disturbing place and talk about the test and its implications for intelligence, (and yes, consciousness).

For the most part today I am going to let the discussion speak for itself, however, for those interested in the process of “prompt engineering,” (essentially, the process of interacting with generative AIs to get them to respond in a way that is helpful), I ended up having to reframe the discussion with ChatGPT midstream to make sure it understood what I was looking for in the conversation. (Something I have done many times in conversations with LLMs to varying degrees of success. Here, I am relatively pleased with the result).

Note: Conversation is with ChatGPT 4o... I am going to post with some other LLMs. . . eventually.