= 27d567bc2d48d9f40e9ccc20ea370b15)
Baughn comments on Open Thread: May 2010 - Less Wrong
You are viewing a comment permalink. View the original post to see all comments and the full post content.
= 783df68a0f980790206b9ea87794c5b6)
Loading…= b715d02e85f7b3b4ae65ca3d3e450868)
Subscribe to RSS Feed
Powered by Reddit
= f037147d6e6c911a85753b9abdedda8d)
You are viewing a comment permalink. View the original post to see all comments and the full post content.
Comments (543)
I'm working on one as part of a game, where I'm knocking off just about every concept I've run into - goal systems, eurisko-type self-modifying code, AIXI, etc. I'll claim it's nontrivial because the game is, and I very much intend to make it unusually smart by game standards.
But that's not really true AI. It's for fun, as much as anything else. I'm not going to claim it works very well, if at all; it's just interesting to see what kind of code is involved.
(I have, nevertheless, considered FAI. There's no room to implement it, which was an interesting thing to discover in itself. Clearly my design is insufficiently advanced.)
I happened to see this today, which you might find interesting. He's using genetic algorithms to make the creatures that the player has to kill evolve. At one point they evolved to exploit bugs in the game.
As a programmer, I'm curious exactly how there is no room to implement it. (I understand the “no room” concept, but want details.)
The most obvious problem, which I suspect most games would have in common, is that it has no notion that it's a game. As far as the AI is concerned, the game world is all there is.
It wants to win a war,and it has no idea that there's a player on the other side. Building up its understanding to the point where that is not the case would be, well, both way too much work and probably beyond my abilities.
May I ask what game?
You can ask, but at the moment it's more of a design document plus some proof-of-concept algorithms. 99% incomplete, in other words, and I don't really want to get people excited over something that might never come to fruition.
I can't really describe the game, because that'd be mostly wishful thinking, but perhaps some design criteria will satisfy your curiosity. So, some highlights I guess:
4X space-based RTS. Realism is important: I want this to look somewhat like reality, with the rule of fun applied only where it has to be, not as an attempt to justify bad physics.
Therefore, using non-iterative equations where possible (and some places they really shouldn't be used) to allow for drastic changes in simulation speed - smoothly going from slower than realtime during battles to a million times faster for slower-than-light interstellar travel. That is to say, using equations that do a constant amount of work to return the state at time T, instead of doing work proportional to the amount of in-game time that has passed.
Therefore, having a lot of systems able to work (and translate between) multiple levels of abstraction. Things that require an iterative simulation to look good when inspected in real-time may be unnoticably off as a cheaper abstraction if time moves a thousand times faster.
To support this, I'm using an explicit cause-effect dependency graph, which lead me to..
Full support for general relativity. Obviously that makes it a single-player game, but the time compression pretty much requires that already.
Causality, FTL, Relativity - pick any two. I'm dropping causality. The cause-effect graph makes it relatively (ha ha, it is to laugh - theoretically it's just looking for cycles, but the details are many) simple to detect paradoxes. What happens if there are paradoxes, though.. that, I don't know yet. Anything from gribbly lovecraftian horrors to wormholes spontaneously collapsing will do.
Hopefully, I'll find the time to work on it, because it sounds like an interesting game to play. :P
Ambitious time travel (or anomalous causality) game mechanics are fun.
There's the Achron RTS game which involves some kind of multiplayer time travel. As far as I can tell, they deal with paradoxes by cheating with a "real" time point that progresses normally as the players do stuff. There is only a window of a few minutes around the real time point to do time-travel stuff in, and things past the window get frozen into history. Changes in the past also propagate slowly into the rest of the timeline as the real time point advances. So paradoxes end up as oscillations of timewaves until some essential part moves out of the time travel window and gets frozen in an arbitrary state.
I'm not sure how well something similar could work with a relativistic space game where you end up displaced from time by just moving around instead of using gamedev-controllable magic timetravel juice.
Your concept also kinda reminds me of a very obscure Amiga game called Gravity. Based on the manual it had relativistic space-time, programmable drones, terraforming planets from gas giants and all sorts of hard SF spacegame craziness not really seen games pretty much ever nowadays.
I've been playing Achron, but it's not really an inspiration. How should I put it..
My understanding of physics is weak enough without trying to alter it. If I stick as closely as possible to real-life physics, I know I won't run into any inconsistencies.
Therefore, there will be no time-travel. I might do something cute with paradoxes later, but the immediate solution for those is to blow the offending ship or wormhole up, as real-life wormholes have been theorized to do via virtual particles.
Blow up the paradox-causing FTL? Sounds like that could be weaponized.
I was about to go into detail about the implications of FTL and relativity but realized that my understanding is way too vague for that. Instead, I googled up a "Relativity and FTL travel" FAQ.
I love the idea of strategically manipulating FTL simultaneity landscape for offensive/defensive purposes. How are you planning to decide what breaks and how severely if a paradox is detected?
I think the only possible answer to that is "through play-testing".
As I understand it, real-life wormhole physics gives enormous advantages to a defender. However, this is a wargame, so I will have to limit that somewhat. Exactly how, and to what degree - well, that's something I will be confronting in a year or two.
(And yes, it could be weaponized. Doing so might not be a good idea, depending on the lovecraft parameter, but you can certainly try.)