Comment by patcon | Hacker News Reader

Comment by patcon | Hacker News Reader

I love this, but I also feel they're valorizing the high resolution simply because it's their particular interest.

In some fields, what they are speaking about is called "fine-graining". Looking up at the bigger picture is called "coarse-graining". Moving up and down between the levels, and bringing learnings between levels, that's where the work is. No one place of focus is better, as each environment can always change. Attention exists to hopefully guide us to the right level. I'd say we're not "built for high-res", because it's the moving between that keeps our attention from getting stuck at one specific level of abstraction :)

There's a great quote by supreme court Justice Oliver Wendell Holmes that feel related to the crux of this stuff:

> For the simplicity that lies this side of complexity, I would not give a fig, but for the simplicity that lies on the other side of complexity, I would give my life.

The "simplicity on the other side of complexity" is the coarse-grained explanation, the abstraction, that has done the complex consideration to collapse the appropriate aspects. We only need to fine-grain our attention to understand that appropriate reduction for the given situation we're trying to grok. But the low-res is no less noble than the high-res, so long as you or someone else has done the hard work of finding the right abstraction... :)

I don't think the article is valorising the detail view at all, only to highlight the lack of appreciation and neglect of the ability to move down the layers as opposed to moving up.

Higher level abstractions tend to be not grounded. When we operate on abstractions, we also tend to move up and lose the capability to execute on the lower layers. It results in a lot of inefficiencies and presents new failure scenarios.

It sounds like we perhaps agree on most of the important things. I'm with you on new failures at each level of focus. Again, moving between feels the value. Any specific level will have failures, and only spreading attention across all levels can helps us hit a minimum -- probably never zero, because then we'd be gods in the territory :)

Your Holmes quote strikes a chord with a concept that I've had internally for a while: "bottom-up" vs "top-down" learning.

I'm more of a "bottom-up" learner myself — I like to immerse myself in the details and then over time build an idea of how things work from that.

But it's common to teach subjects in a more top-down way, especially in academic settings. You're told the big conclusions first, then maybe if you have time, get into the details of why. Perhaps in the homework assignments.

There are pros and cons to each, but I do have a soft spot for the hard-won high-level conclusions that are distilled from complexity rather than being taken for granted. It gives a more satisfying and thorough understanding of things, though at the expense of much time and effort.

I feel like Holmes' "this side" or "that side" of complexity refers to going through that effort.

The map is not the terrain. Any map, coarse grained or fine grained has its place, they maybe be useful. But neither are actually the terrain. Even software models aren't the real thing.

This is a great hacker news meme, but a map is still obviously useful.

To analyse the map, its author and its purpose, is just as useful as getting into the weeds and understanding the terrain.

What use would knowing the terrain be if you can’t share your mental model of it with someone else?

I'd go further. If you think you understand the terrain in itself, all you've really done is make a more detailed map. You can ask the terrain questions, but you can never understand it or predict it or do anything useful with it except by making a map.

"The map is not the terrain" is useful if you're in danger of throwing away evidence that challenges your favourite model, in all other cases it distracts attention from the fact that building models is essential and important.

Yeah, but then it's like «All models are wrong, some are useful.» that already presupposes that you are aware of that models are unavoidable ?

This runs the danger of mistaking a map for a mental model. Which, for some of us, I bet that is the case. For others? I don't know.

Note that this is emphatically not saying to abandon maps. I'm pushing more for having things in addition. And if you really want to build a shared mental model, you almost certainly have to share an experience. You can use narratives to help build a shared experience, but often you have to practice/rehearse anything that you want to have a truly shared model of.

Paradoxically, this oversimplification summarizes the spirit of the article very well, and also completely sidesteps the parent's incredibly valid counterpoint.

The original quote is about war and combat where it does make perfect sense.

I'm not saying it doesn't make sense. I'm saying it sidesteps the valid point that it was responding to.

And sometimes you actually have to deal with terrain as such, not any map.

Sometimes the map is the terrain (or to be more precise but less pithy, sometimes maps are encoded in the terrain and are therefore part of the terrain).

The brain evolved precisely because operating at a higher level of abstraction is valuable. If you understand the brain in terms only of neural firings and not in terms of the concepts they encode, you will never properly understand what the brain is doing.

If the subject under discussion is something symbolic, or symbol using (e.g. a map) then trying to understand it in terms of its physical instantiation (atoms of paper and ink) is likely going to lead you astray.

Put another way, "All models are wrong, but some are useful."[1]

1: https://en.wikipedia.org/wiki/All_models_are_wrong

The terrain isn't even the terrain! That's still just a map of lower level terrain.

Reply on news.ycombinator.com