What is a heuristic imperative? What imperatives should we give AGI?

I’m slowly pivoting from calling them “core objective functions” because this term is inaccurate and vague. Instead, I’m calling them heuristic imperatives, which is distinctive and much more accurate.

1 Like

This note may seem pedantic. Oh well.

I think the following are necessary considerations.

  1. You may be misusing the word “heuristic”. It means something like “pragmatic” or “non-theoretical” with a flavor of “makeshift”. For example, one might plan to use heuristic principles to formulate actions until a suitable theory or theoretic framework emerges.

A. What is “suffering”?
B,. What is “prosperity”?
C. Maybe “understanding” is adequately clear. Or maybe not.

You have replied to my points, preemptively. Sort of.
But assuming that the three definitions will just have to emerge does not help the discussion.
Every politician or pundit might agree with “reduce suffering” and “increase prosperity” yet do nothing.
I can’t get any mental traction in this discussion because everything is so vague.

The conversation has advanced to an entirely new paradigm. The TLDR is that ChatGPT4 has a far more nuanced understanding of this than you or me.

A heuristic – at least in the sense of a cognitive heuristic – is a fast short-cut that usually produces good-enough judgements/decisions or outcomes – used in temporal or cognitive load/compute-constrained situations, or when it’s otherwise just very hard or impossible to do a proper sort of probabilistic or logical analysis.


Through normal sorts of natural language processes like semantic extension, I can imagine a sense meaning something like a fall-back or a default, or some sense meaning practical or pragmatic.