Source
- URL: https://x.com/Jonathan_Blow/status/1398461530996240385
- Author: Jonathan Blow (@Jonathan_Blow)
- Posted: 2021-05-29 02:09:45
Thread
1/ @Jonathan_Blow
RW Hamming dropping the pills.
It occurs to me that this is as relevant to game design as it is to technical design (and is consistent with my personal experience).
2/ @TylerFCloutier
@Jonathan_Blow
Being on the surface does not imply that one or more of the parameters are pushed to their extreme. It just maximizes the sum of the square of the parameters, so I’m not sure I follow the point.
3/ @Jonathan_Blow
@TylerFCloutier
Explain how “maximize sum of squares of parameters” is not a synonym for “push them to extremes”.
4/ @TylerFCloutier
@Jonathan_Blow
Well, I mean maximizing the sum of squares could mean equal weighting of parameters in all dimensions, so that no individual parameter is at its maximum allowed value. Maybe I mistook the design advice, but I thought his suggestion was to push certain dimensions to their extremes
5/ @Jonathan_Blow
@TylerFCloutier
What? Why? Why can you, having set up this situation, not then increase the value of one of the parameters? It sounds like you are inventing a constraint that is not part of the posed problem, and acting as if that constraint were mandatory.
6/ @TylerFCloutier
@Jonathan_Blow
He said literally “surface of the feasible design region”. If you kept increasing your parameters you would just go out of that region, so it’s a constraint of the problem. He’s implying that to reach that you have to push individual parameters to their extreme, which is wrong.
7/ @TylerFCloutier
@Jonathan_Blow
If you want to stay within a surface, then at the surface you can only increase one param at the detriment of others. To be honest the whole thing sounds a bit hand-wavy/pseudo-profound to me, but maybe I’m wrong. I, for one, don’t see how it relates to better design tho.
8/ @AndriyTyurnikov
@TylerFCloutier @Jonathan_Blow
Maybe design region is called region, because it is bound … like with constraints, like mass, size, cost of production, structural integrity. If you can push any parameter wherever you want without trade off - you don’t have a problem - you have a dream.
9/ @TylerFCloutier
@AndriyTyurnikov @Jonathan_Blow
That’s more or less my point, but it doesn’t tell you that you should go all-in on one (or several) dimensions. I must have misunderstood the advice or something. Maybe it’s just, hey you should think about exploring the edges of your parameter space, the edges are big.
10/ @AndriyTyurnikov
@TylerFCloutier @Jonathan_Blow
Range 0 to 1. Aim close to 1 on max dimensions, to get results like this:
Branches: 2021-05-29-syranide-its-a-sphere-you-can-get-1-in-one-dimension-and-0