ADM-201 dump PMP dumps pdf SSCP exam materials CBAP exam sample questions

AI学习重制电脑游戏 – 译学馆
未登录,请登录后再发表信息
最新评论 (0)
播放视频

AI学习重制电脑游戏

AI Learns To Recreate Computer Games | Two Minute Papers #195

各位学霸同学 这里是由Károly Zsolnai- Fehér带来的两分钟论文
Dear Fellow Scholars, this is Two Minute Papers with Károly Zsolnai-Fehér.
对于大多数视频游戏 如果我们看它运行了一段时间
In most video games that we’ve seen running for at least a few moments, we learn to anticipate
我们就会尝试预测下一秒钟会发生什么 甚至如果有足够的耐心和技巧
what is going to happen in the next second, and even more, if given the patience and skills,
我们可以尝试重建游戏本身的一部分
we could attempt to recreate parts of the game itself.
这篇论文甚至可以做到更好
And what you see here in this work is actually even better, because it requires neither the
因为它既不需要耐心也不需要技巧
patience nor the skills to do that.
这就是这个价值百万美元的想法:使用学习算法来观看游戏视频录像
So here is the million dollar idea: let’s have a learning algorithm look at some video
然后让它去重建 这样我们就可以专注于玩游戏了
game footage, and then, ask it to recreate that so we can indulge in playing it.
这个想法已经被用在“超级马里奥”上了 稍后你也会看到
The concept is demonstrated on the Super Mario game and later, you will also see some results
它在千禧一代童年最喜欢的游戏“洛克人”上的表现
with the millennial childhood favorite, Mega Man.
先前有许多深入到游戏源码的做法
There are many previous works that hook into the source code of these games and try to
它们尝试通过读取代码级的指令预测接下来会发生什么
read and predict what happens next by reading the code-level instructions.
但是这篇论文不是这样 它是通过观看视频输出和
But not in this case, because this technique looks at the video output and the learning
在像素层的学习来完成 因此不需要通过
takes place on the level of pixels, therefore, no access to the inner workings of the game
游戏内部
is necessary.
我们为算法的学习过程提供了两种东西 第一种是精灵调色板
The algorithm is given two things for the learning process: one, a sprite pallette that
它包含所有可能出现在游戏中的元素 包括背景砖块
contains all the possible elements that can appear in the game, including landscape tiles,
敌人 金币等
enemies, coins and so on.
第二 我们还提供了游戏的完整通关视频作为输入
And two, we also provide an input video sequence with one playthrough of the game to demonstrate
来演示游戏元素的结构以及它们之间可能发生的交互
the mechanics and possible interactions between these game elements.
视频就是一系列的帧 这种技术可以从其中学习如何从一帧
A video is a series of frames, from which the technique learns how a frame can be advanced
推进到下一帧
to the next one.
学习过足够的训练样本后 它就能够
After it has been exposed to enough training samples, it will be able to do this prediction
在以前从没见过的视频帧上做预测
by itself on unknown frames that it hasn’t seen before.
这就意味着我们可以开始玩这个它试图模仿的游戏了
This pretty much means that we can start playing the game that it tries to mimic.
可供利用的游戏有许多相似之处 把从其他游戏中
And there are similarities across many games that could be exploited, endowing the learning
重复用到的知识赋予学习算法 这样它们甚至可以重建
algorithms with knowledge reused from other games, making them able to recreate even higher
更高质量的视频游戏 即使是某个情节从没在训练视频中
quality computer games, even in cases where a given scenario hasn’t played out in the
出现过也可以
training footage.
上班时间玩游戏曾经是计算机图形学研究者的特权
It used to be the privilege of computer graphics researchers to play video games during work
但是显然 机器学习的科学家也加入到他们的阵营了
hours, but apparently, scientists in machine learning also caught up in this regard.
恭喜
Way to go!
但是它也有不足之处
A word about limitations.
它预测的速度并不快 并且预测是基于从视频序列中学到的事实进行的
As the predictions are not very speedy and are based off of a set of facts learned from
因此有一个问题 将这种技术扩展到更复杂的
the video sequences, it is a question as to how well this technique would generalize to
3D游戏 效果会怎么样呢?
more complex 3d video games.
和几乎所有的研究工作一样 这是一个跳板 但却是很重要的跳板
As almost all research works, this is a stepping stone, but a very important one at that as
因为它证明了一个很棒的想法的可行性
this is a proof of concept for a really cool idea.
你知道该接下来的步骤 会继续出现一些这方面的论文 然后这个想法会有
You know the drill, a couple papers down the line and we’ll see the idea significantly
巨大的改进
improved.
这里的结果显然并不完美 但它很好地展示了一个新概念
The results are clearly not perfect, but it is a nice demonstration of a new concept,
并且从中了解到机器学习研究的发展速度
and knowing the rate of progress in machine learning research, you will very soon see
不久你就可以看到梦幻般的结果
some absolutely unreal results.
甚至 我希望通过生成式对抗网络或生成式潜在优化 在现有的游戏中
What’s even more, I expect that new levels, enemy types and mechanics will soon be synthesized
合成出新的关卡 新的敌方类型以及新结构
to already existing games via generative adversarial networks, or generative latent optimization.
如果你想了解更多相关内容 可以点击
If you would like to hear more about these, as always, the links are available in the
视频说明中的链接
video description.
同样 如果你喜欢本期节目 请在Patreon上支持我们 这样可以让我们
Also, if you enjoyed this episode, please make sure to help us tell these incredible
被让越来越多的人知道
stories to more and more people by supporting us on Patreon.
您的支持一直以来都是我们前进的动力
Your support has always been absolutely amazing.
细节部分已经放在视频说明中了
Details are available in the video description.
感谢您的观看和大力支持 我们下期再见
Thanks for watching and for your generous support, and I’ll see you next time!

发表评论

译制信息
视频概述

本视频介绍了利用机器学习玩游戏的最新进展,以及利用生成式对抗网络重建游戏的发展。

听录译者

收集自网络

翻译译者

[B]hugue

审核员

审核团O

视频来源

https://www.youtube.com/watch?v=2VyhmbEjs9A

相关推荐