Last week I finally and at long last managed to prompt an LLM into a “Socratic[1] Brainstormer”. It’s really frustrating though that it doesn’t ever hit me with a real curve ball that makes me go “Oh wow, I never thought of that” or “Oh shit, everything I said is a falsehood” a la Socrates. But as far as an rubber-duck with extra bells and whistles go, it’s a step forward.
It has stopped with the overt sycophancy, no more “fascinating” or “what you’ve just said is very significant”. It just asks questions, as few or as many questions as I specify. Claude seems to rely on a few stock question formats though, such as “which are the topics that cause people to lean in and take notice” “which common ideas/theories frustrate you or you find superficial?”. It also tends to seek generalized patterns rather than specifics—my brainstorming about the theory behind colour schemes on movie sets leads to covertly sycophantic questions like “it sounds like this is a more all-encompassing phenomenological theory about how we create meaning from visual information”—no Claude, this is not.
ChatGPT when it is prompted with words like”empirical” or “details oriented” tends to leap to questions about execution before I’ve managed to brainstorm the core ideas. If I need a theme for a content calendar, it’ll already be asking me which metrics I’m going to use to test reels on Instagram against each other—Sorry GPT, I don’t know what variable I’ll be using the metric to measure the success of yet.
What’s, perhaps, most noteworthy is how giddy and excited I was when I finally managed to get purely inquisitive responses. I actually was grinning at the prospect of having an indefatigable personal interrogator who would as slyly as Lt. Columbo ask me leading questions to help me discover curveballs and new ideas I couldn’t possibly arrive at on my own. I keep searching…
By “Socratic” I mean here, purely the sense that it asks a lot of questions. Sadly I haven’t managed to successfully prompt it into making use of Socratic Irony—identifying paradoxes or forcing me to admit the polite fictions that expedite social intercourse even though I know they are not true.
Last week I finally and at long last managed to prompt an LLM into a “Socratic[1] Brainstormer”. It’s really frustrating though that it doesn’t ever hit me with a real curve ball that makes me go “Oh wow, I never thought of that” or “Oh shit, everything I said is a falsehood” a la Socrates. But as far as an rubber-duck with extra bells and whistles go, it’s a step forward.
It has stopped with the overt sycophancy, no more “fascinating” or “what you’ve just said is very significant”. It just asks questions, as few or as many questions as I specify. Claude seems to rely on a few stock question formats though, such as “which are the topics that cause people to lean in and take notice” “which common ideas/theories frustrate you or you find superficial?”. It also tends to seek generalized patterns rather than specifics—my brainstorming about the theory behind colour schemes on movie sets leads to covertly sycophantic questions like “it sounds like this is a more all-encompassing phenomenological theory about how we create meaning from visual information”—no Claude, this is not.
ChatGPT when it is prompted with words like”empirical” or “details oriented” tends to leap to questions about execution before I’ve managed to brainstorm the core ideas. If I need a theme for a content calendar, it’ll already be asking me which metrics I’m going to use to test reels on Instagram against each other—Sorry GPT, I don’t know what variable I’ll be using the metric to measure the success of yet.
What’s, perhaps, most noteworthy is how giddy and excited I was when I finally managed to get purely inquisitive responses. I actually was grinning at the prospect of having an indefatigable personal interrogator who would as slyly as Lt. Columbo ask me leading questions to help me discover curveballs and new ideas I couldn’t possibly arrive at on my own. I keep searching…
By “Socratic” I mean here, purely the sense that it asks a lot of questions. Sadly I haven’t managed to successfully prompt it into making use of Socratic Irony—identifying paradoxes or forcing me to admit the polite fictions that expedite social intercourse even though I know they are not true.