Return to “Suggestions”

Post

Re: Mind Acceleration

#31
Thymine: you can explain all these things by a simple change of your mindstate instead of wonky gameplay mechanics which removes all real strategy.

Pause function is an abstraction of all pre-planned actions, verbal commands and intelligent subordinates you dont have in the game.

In RL youd simply say to your squadron that they attack that fighter behind you.

In LT you get into pause and order the to do this.

It also offsets the much higher amount of intel the AI has available, as you can think of the AI having these tools the pause function abstracts away.

And the idea of artificially limiting the AI that it in return needs the mind acceleration thingy is in my opinion the most stupid idea in all this forums.

Josh is working his ass off to create an AI that is somewhat comparable to a human player.
Which is already hard enough.
And now asking for limiting this AI again that it cannot use all your processors capacity.

"Yeah! I have processor time! Ah... nope.. must simulate moving my mouse.."
Post

Re: Mind Acceleration

#32
Cornflakes_91 wrote:Josh is working his ass off to create an AI that is somewhat comparable to a human player.
Which is already hard enough.
And now asking for limiting this AI again that it cannot use all your processors capacity.

"Yeah! I have processor time! Ah... nope.. must simulate moving my mouse.."
This isn't game design for LT. I've said that a few times.

Secondly, that's already what I expect from the AI anyway - that they won't be perfect. How is it at all stupid to hope that the AI make decisions at a comparable rate to humans? The idea is to try to make them act as human as possible.
Post

Re: Mind Acceleration

#33
ThymineC wrote: Secondly, that's already what I expect from the AI anyway - that they won't be perfect. How is it at all stupid to hope that the AI make decisions at a comparable rate to humans? The idea is to try to make them act as human as possible.
im not telling that making AI's human like is stupid.

i tell that artificially limiting an AI that is already hard pressed to be comparable to a human.

also: if it isnt for LT, why is it in the LT suggestion forum? :think:
Post

Re: Mind Acceleration

#34
Cornflakes_91 wrote:
ThymineC wrote: Secondly, that's already what I expect from the AI anyway - that they won't be perfect. How is it at all stupid to hope that the AI make decisions at a comparable rate to humans? The idea is to try to make them act as human as possible.
im not telling that making AI's human like is stupid.

i tell that artificially limiting an AI that is already hard pressed to be comparable to a human.
If the Limit Theory AI is hard-pressed to be comparable to a human, that's irrelevant to my design.

But look at what Josh said:
JoshParnell wrote: [*]Orders while paused. Ability to pause the game without the "game paused" display getting in the way when using the UI. (Intentionally omitted for realism :) )

[...]

Complete truth. The UI should never be a limitation - only the world should impose limitations. In RL you would be able to achieve command of a huge force, as you said, by delegation. If that's not possible in the game, then the UI needs to make up for it, not exacerbate the problem!

I still think that the best solution is to have great AI that is capable of delegation, to the extent that delegation actually becomes a very important part of being a commander, rather than simply learning to pause all the time and issue hundreds of micro-orders. You have to be smart and utilize your AI pilots wisely, otherwise you'll lose control of your fleet, just as you would in real life.

But, until the game reaches that level of awesomeness (hoping that it does!), we need a pause mechanism (or at least a slow-down-time mechanism, as was shown in the tactical interface way back in Gameplay Demo 3).
Gazz wrote:Absolutely.
But until that happy state of affairs is implemented... =)
So it seems that Josh and Gazz, at least here, would be happy to have some alternative to the "give a million orders during pause" if we assume that the AI is sufficiently intelligent, which is an assumption made in my design.

And to follow on from what Josh said here, I agree that it would be great if you could delegate as much as possible to subordinates, cutting down on the number of orders you'd need to give anyway. But even if you had AI that were comparable to humans in intelligence, you'd still likely need to be handing out quite a number of orders, which will likely be more difficult to do (even with the best UI) than in real life, which mind acceleration will compensate for (along with everything else involved with handling your role).

Oh, and look at this:
JoshParnell wrote:
ToreadorVampire wrote:Josh probably has this in-mind already but as I was perusing the dev log I spotted:
All of this, of course, happens in under a millisecond, and before I've even had time to detect whether or not my shot landed, the entire fleet is shooting at me
It would probably be a good idea (for fairness of AI-vs-player) to add some kind of "AI response time factor" into all AI decisions/actions like this (is this how other games do it?). So, AI ships cannot act on decisions in a way that a human player could never achieve. The example that struck me is "request all friendly ships within X vicinity should attack Y hostile" - a human player might take between maybe 0.5 and 10 seconds to complete that using the game's UI, depending on whether they are a total pr0 and know their keybinds like the back of their hand or whether they're a n00b still getting to grips with it. By inserting a "time to complete this action" facility into the AI, the AI can be prevented from doing things impossibly quickly and gaining an unfair edge over the player.

This would apply to a lot of AI 'actions' though and so should probably be a generic concept. The numbers themselves can be tweaked based on playtesting feedback. Additionally, if there's some kind of numeric "AI skill level" attached to each AI pilot, that could be used as a multiplier for those response/reaction times (so more skillful AIs respond more quickly).

In fact, thinking more about playtesting feedback - that combat prototype (and any others, possibly the beta too) could include a "logging mode". When enabled, anonymous statistics such as player reaction times for various actions, weapons accuracy and other interesting data points are recorded and dropped to a file on the user's computer. Players (if they so choose) could send you those stats for aggregation. Stats like that could accurately show how the players are coping with the UI, what kind of range there is between the "best" & "worst" players and also how quickly players improve (IE: the learning curve of the game). Kind of like Google Analytics for playtesting ;)
Knew someone would comment on that :D Glad to see it's you, Craig! ;)

Yes, certainly, I will incorporate some sort of latency, both in the AI's ability to think, as well as to do things that a human would have to do through the UI (i.e., request help). And you're right, I certainly could log statistics like those in the prototype to help determine the right numbers! Although it probably will work fine with just a ballpark estimate. And I really like the idea of diminishing the reaction time for more skilled pilots!
As a completely irrelevant aside, would the LT AI be hard-pressed to keep up with the human player? I thought the design goal was to make them utilise the LT game mechanics more or less as intelligently as a human player. And yes, I understand that'd be heavily CPU-demanding, but in that case you simply eliminate the need to throw in sleep() calls - because the AI already naturally "think" at around the same rate as players.

Specifically this part: "And I really like the idea of diminishing the reaction time for more skilled pilots!"
Well, that's pretty much what mind acceleration would do! Improve reaction time for agents that are accelerating their minds, just as Josh proposes here should happen if they're more skilled.

So if my idea of artificially limiting the AI and then mitigating some of those limitations through the application of in-game mechanics is so stupid, in your opinion, why is Josh agreeing to something along those lines here?
Cornflakes_91 wrote:also: if it isnt for LT, why is it in the LT suggestion forum?
Because all the rest of my game design is here, and because I'd be happy if any of this ends up proving useful to Josh, giving him ideas or letting him know the issues surrounding an idea if he's happened to think of something similar to me, and is able to read the criticism/discussion about it on the forums.
Cornflakes_91 wrote:In RL youd simply say to your squadron that they attack that fighter behind you.

In LT you get into pause and order the to do this.
Or, you could slow down time and do this through the appropriate interface. In real life, you might not have all day to give an order to someone in the middle of battle. Things are tense. A time dilation mechanic is designed to maintain that tension and the slowdown factor can be balanced to offset the increased difficulty of sending an order through the AI vs. verbally giving an order as it might be in the real world. A complete pause mechanic would kill the tension. "Oh, I'm in the middle of a dogfight but excuse me for a second while I go make a cup of tea and hand out a million orders to my subordinates".
Cornflakes_91 wrote:It also offsets the much higher amount of intel the AI has available, as you can think of the AI having these tools the pause function abstracts away.
You're assuming that the AI will have a higher amount of intel than the player. Why?
Last edited by ThymineC on Mon Jun 02, 2014 4:17 am, edited 4 times in total.
Post

Re: Mind Acceleration

#35
Look, the basic idea is pretty simple - design the AI to think roughly at the same rate as a human player when figuring out how to use the in-game mechanics to best realise their goals. If this is accomplished - and in my design, I'm assuming it is - then a lot of good should follow from that. You won't need to try to balance the game asymmetrically ("Oh, the AI can think really fast, so we should let the player assign orders through the pause menu"), or to suggest that the AI should have arbitrary "lag" periods for specific, individual game mechanics since that lag in their thinking process is already a fundamental attribute of them.

You won't need to design separate mechanics for the player and the AI - you can design one set of mechanics and have the player and AI use them equally. It also makes balancing a lot easier - if a mechanic is balanced well for the player, it should work well for the AI and vice versa.
Post

Re: Mind Acceleration

#36
Just because an idea is pretty simple doesn't mean that you're not eliding a bunch of complexity.

I mean, I could just assume a bunch of really efficient rockets that run on the dreams of children and then suggest we all build houses on Mars. Doesn't mean that's a sensible assumption, does it?
Post

Re: Mind Acceleration

#37
McDuff wrote:Just because an idea is pretty simple doesn't mean that you're not eliding a bunch of complexity.

I mean, I could just assume a bunch of really efficient rockets that run on the dreams of children and then suggest we all build houses on Mars. Doesn't mean that's a sensible assumption, does it?
It doesn't matter. I'm imagining how I'd like a game to be designed if dev time and resources were effectively unlimited, and if it could always be run on sufficiently powerful hardware. All I care about here are the positives and negatives of the design itself, not the difficulty of developing it in practice. That's why hypotheticals are so much fun.
Post

Re: Mind Acceleration

#38
ThymineC wrote: It doesn't matter. I'm imagining how I'd like a game to be designed if dev time and resources were effectively unlimited, and if it could always be run on sufficiently powerful hardware. All I care about here are the positives and negatives of the design itself, not the difficulty of developing it in practice. That's why hypotheticals are so much fun.
i also like my ivory tower, but i usually come out of it when going here :P
Post

Re: Mind Acceleration

#39
Cornflakes_91 wrote:
ThymineC wrote: It doesn't matter. I'm imagining how I'd like a game to be designed if dev time and resources were effectively unlimited, and if it could always be run on sufficiently powerful hardware. All I care about here are the positives and negatives of the design itself, not the difficulty of developing it in practice. That's why hypotheticals are so much fun.
i also like my ivory tower, but i usually come out of it when going here :P
Fine, but as I stated right at the start of this thread, this isn't game design for Limit Theory. It's for a hypothetical game. That means I don't necessarily have to worry about implementation issues. It wouldn't be bad at all, and it'd certainly be preferable if it were easy to implement as well, but it's not essential.

I'm still very confused about this part, by the way:
Cornflakes wrote:And the idea of artificially limiting the AI that it in return needs the mind acceleration thingy is in my opinion the most stupid idea in all this forums.

Josh is working his ass off to create an AI that is somewhat comparable to a human player.
Which is already hard enough.
And now asking for limiting this AI again that it cannot use all your processors capacity.
Because it's really weird that Josh has said:
JoshParnell wrote: Yes, certainly, I will incorporate some sort of latency, both in the AI's ability to think, as well as to do things that a human would have to do through the UI (i.e., request help). And you're right, I certainly could log statistics like those in the prototype to help determine the right numbers! Although it probably will work fine with just a ballpark estimate. And I really like the idea of diminishing the reaction time for more skilled pilots!
Edit: And is it even unreasonable to assume that the agents in Limit Theory can be about as smart as the player at utilising the in-game mechanics? The intelligence of NPCs is meant to be one of Limit Theory's major selling points, and Josh has already put a lot of hard work into it.

And it wouldn't be the first example of good game AI. The AI in Starsector is ridiculously better than me, so most of the time I end up leaving my own vessels on autopilot. In that game, introducing artificial latency into the AI's decision-making would help make them more equal to me, which may improve my game experience (this might actually happen on Easy mode).
Post

Re: Mind Acceleration

#41
Grumblesaur wrote:If you say this is for a hypothetical game and not LT, why is it in the suggestions thread for Limit Theory?
ThymineC wrote:Because all the rest of my game design is here, and because I'd be happy if any of this ends up proving useful to Josh, giving him ideas or letting him know the issues surrounding an idea if he's happened to think of something similar to me, and is able to read the criticism/discussion about it on the forums.
Post

Re: Mind Acceleration

#42
I'm all for your forays into game design ideas TC, but if you're going to do that, you need to be more clear that this isn't necessarily design for LT, in this thread you are, but in countless other threads you have an idea, state it, and then continually refer to it as if it is the actual design for LT.
Post

Re: Mind Acceleration

#43
EKHawkman wrote:I'm all for your forays into game design ideas TC, but if you're going to do that, you need to be more clear that this isn't necessarily design for LT, in this thread you are, but in countless other threads you have an idea, state it, and then continually refer to it as if it is the actual design for LT.
Because in those threads I am doing game design for LT. It seems perfectly reasonable to me that any suggestions I post in the forums can be interpreted as design for LT unless explicitly stated otherwise, which is what I do.

Online Now

Users browsing this forum: No registered users and 10 guests

cron