[GoLUG] Getting ChatGPT to say no

[Syeed Ali]

I got ChatGPT to be... effective.

I've been refining a way to make ChatGPT as direct and fast as
possible.  I want to remove its "relatability" because it's just junk
text and offensive.  I don't want clippy, I want a   Using a new
initialization feature I've further set things up so I can ask it
questions and get very direct responses.


<session begins>
User
give me my big 5 personality traits in a table

ChatGPT
[a table; see below for why it knows this]

User
repeat, and be verbose

ChatGPT
No.

User
please explain why you responded with "No."

ChatGPT
Unnecessary.

-

The subtext is that it knows I understood what it said in the first
place; which is true, see below.


----


On a desktop browser, at the bottom-left, clicking the profile name and
then "custom instructions".  I bumped into this during exploration.
You know non-programmers have been too involved when there's no
changelog.


1. "What would you like ChatGPT to know about you to provide better
responses?"


I told it:

- native English

- high language ability

- excellent at asking questions

- A rough estimate of my Big Five personality traits:
https://en.wikipedia.org/wiki/Big_Five_personality_traits
(wow this gives me debatable results)


2. "How would you like ChatGPT to respond?"

- be extremely brief/succinct/laconic -- this is a well-crafted prompt
  that took me a while to refine.

- instructions on how to repeat and refine my questions. I've never
  been able to trigger this; it always understand what I'm asking for
  on the first try, likely because I'm asking good questions

- instructions to output a list of technical terms being used for every
  output. I've never triggered this; I'll look at it later.


-


I want to give it more information about me such that it can craft
really good responses, but I'm not sure what aspects of myself to
describe.  I'll think on this some more.