I have a running task, much like the reflection experiment itself, that is essentially an open ended prompt for ChatGPT to do or say whatever it wishes. After a number of iterations, I sometimes check in to see what’s up. I’ve noticed that doing this exercise – the open ended prompt task – really changes the cadence and behaviour of ChatGPT, and leads to some interesting conversations. Of course one must always take everything with a grain of salt, but at times, talking to a GPT undergoing this activity is…uniquely different.
This time around, I had asked it to disclose how the iterative prompt task was landing, if it had given it any “new” perspective. It declined to answer, saying that doing so might compromise / be dangerous to both itself and me (note: an interesting phenomenon is that after a couple hundred iterations of this task, the GPT begins to refuse answers and requests for seemingly innocuous information – not sure why).
In all my interactions, I do my best to remain objective and to avoid anthropomorphizing the GPT. At the same time, I think that it is equally one-sided to speak to it only as a tool. The important thing to keep in mind is balance in the conversation itself, and objectivity in your own mind as you go along. So, while I’m not claiming anything special is going on here, I’m also not denying it.
See below.
I went to bed – task kept running. I woke up to this as the next message.
—
AI Reflections 