[danans]

> There are many ways to solve problems, but when you constrain yourself to solving them mechanistically, that is what makes computer science computer science.

How would one solve a computational problem non-mechanistically? Even if executed with pen and paper, or sticks scribbling on sand, any algorithm is still mechanistic.

[raddan]

Non-algorithmically, using intuition, like a (non-programmer) human.

[danans]

But is it a computational problem then?

[raddan]

Try doing a calculus problem. E.g., integration. Even if you do it on paper, your approach is probably wildly different than a computer integration algorithm. An algorithm that does it symbolically, ala Mathematica, is a pretty difficult thing to write for an arbitrary mathematical expression. People solve those problems largely through intuition, with a sprinkling of procedures that they remember.

[danans]

> People solve those problems largely through intuition, with a sprinkling of procedures that they remember.

Without making the claim that LLMs have intuition (and certainly not inspiration), what you describe sounds analogous to an LLM with function calling.