uncopy2002
Level: Smiter
Rank Points: 431
Registered: 07-28-2014
IP: Logged
|
Re: Can you do it?--a DROD solver (+2)
*Massive Necro*
There's a important point that somehow the whole discussion on DROD solvers missed:
Yes, there's no way to make a usable solver that solves a room in reasonable amount of time.
But no, that doesn't mean it's not possible.
I just don't see why we have to limit a solver to "being sentient and plan strategies for us". We don't have to.
Orb solver and trapdoor dropper proves that we can *at least* solve a small set of particular problems which are feasible and fast to solve.
For example, when I'm optimizing I'd love to know the following:
Given this initial horde of monsters with a certain movement order and environment, what sequence of moves can kill them the fastest while I'll move forward the most amount of space possible?
Or
Given this object and environment, what sequence of moves can push this from A to B in the fastest manner?
Or even just simple problems like
Given this pile of tarstuff, is it cuttable? What is the most efficient way to cut it?
Which a solver can quickly determine the answer, so I can know how many moves I'll end up in advance.
Ultimately, no, we're not designing a solver which plans for us. That's what we, as human, do.
But they can do some tiny problems we determined that we'll need to solve, which then we consult their answers to judge whether it's good or not.
Now that I'm writing this post, I'm increasingly finding the whole discussion like what we talk about AIs. We keep talking about their implications, or the difficulty to implement one, but come on, good programs save our time on doing various tedious jobs, and they aren't AIs at all. And they work by us telling them what to work on.
|
.
.
.
.
