TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Why I’m Remaking OpenAI Universe

224 点作者 evc123将近 8 年前

14 条评论

Houshalter将近 8 年前
Why not use game emulators? With popular NES emulators you can advance the game frame by frame. You can read the raw memory addresses that correspond to the score. You can dump the memory at any time and reload the game to a specific game state. You can even manipulate the games in many fun ways by messing around with the game memory. Or give an AI algorithm access to memory addresses as additional information, instead of relying on pure machine vision, if you want to do that..<p>Here&#x27;s an example of a guy who made a general game playing algorithm that brute forces it&#x27;s way through any NES game: <a href="https:&#x2F;&#x2F;www.youtube.com&#x2F;watch?v=xOCurBYI_gY" rel="nofollow">https:&#x2F;&#x2F;www.youtube.com&#x2F;watch?v=xOCurBYI_gY</a> This isn&#x27;t necessarily interesting from an AI perspective - the playing algorithm is just brute force. But it shows what can be done with the platform, easily reloading to previous states and exploring counterfactual futures (which is exactly the sort of thing RL algorithms do.) He also has a cool algorithm for finding the objective function of an arbitrary game, by watching a human play, and seeing what memory addresses increment. Which is a lot more easy to use than writing OCR code to read the score and game over states from the screen.
gdb将近 8 年前
(I work at OpenAI.)<p>Great project. We&#x27;ve found that the VNC Universe environments are hard for today&#x27;s RL algorithms primarily due to the their async nature. We&#x27;re currently working on a new set of Universe environments without VNC; I&#x27;m very happy to see others inspired by the core ideas of Universe as well.
评论 #14638580 未加载
评论 #14639003 未加载
评论 #14638139 未加载
评论 #14638227 未加载
评论 #14638211 未加载
hackpert将近 8 年前
This is great. Using HTML5 games in a headless browser makes a lot of sense because the need for VNC is circumvented. However, I think that while OpenAI&#x27;s implementation is certainly not the best, having access just the information on the screen is not a bad idea in itself as a (maybe optional) constraint. With access to the game&#x27;s internal state we don&#x27;t even need RL for solving a large number of games - algorithms like NEAT are sufficient.
评论 #14638996 未加载
daveguy将近 8 年前
According to the author, &quot;Universe never really took off in the AI world.&quot;<p>That&#x27;s a bit premature for a project that was just released less than 7 months ago, isn&#x27;t it?<p><a href="https:&#x2F;&#x2F;blog.openai.com&#x2F;universe&#x2F;" rel="nofollow">https:&#x2F;&#x2F;blog.openai.com&#x2F;universe&#x2F;</a><p>Edit: that said the project seems to have some interesting and needed improvements (esp time adjustment). Glad to see dialog between muniverse and openai here.
evc123将近 8 年前
<a href="https:&#x2F;&#x2F;github.com&#x2F;unixpickle&#x2F;muniverse" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;unixpickle&#x2F;muniverse</a><p><a href="https:&#x2F;&#x2F;github.com&#x2F;unixpickle&#x2F;demoverse" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;unixpickle&#x2F;demoverse</a>
strin将近 8 年前
Awesome project.<p>Despite the flaws, the nice thing with VNC is its universality to support any apps on a computer. Using HTML5 in a browser limits the scope of things we could encapsulate as environments, and makes it less &quot;universe&quot;.<p>However, there is a difference between the universality of the tech stack and the exposed interface. In my opinion, the future universe would be rich clusters of RL environments with unified API, each of which implemented using different underlying technology to meet the desired synchronicity and frame performance.<p>HTML5 could deliver one of such clusters.
评论 #14638437 未加载
dswalter将近 8 年前
I&#x27;m a little surprised, but this seems like a good idea. HTML5 certainly has a brighter present and future than flash, and skipping the OCR stem should save quite a few CPU cycles.
zzh8829将近 8 年前
I am also working on related project. Flash and HTML5 games in chrome are great but they are very far away from the initially promised full blown GTA5, Starcraft and other complex envs. I am in process of remaking the Universe framework for host machine, since running those computation intensive games at reasonable frame is nearly impossible inside docker or virtual machines.
misiti3780将近 8 年前
Did openAI really unofficially abandoned universe ?
评论 #14636887 未加载
评论 #14636744 未加载
namuol将近 8 年前
Funny, I have an old (unfinished) HTML5 space-exploration game by the same name:<p><a href="https:&#x2F;&#x2F;github.com&#x2F;namuol&#x2F;muniverse" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;namuol&#x2F;muniverse</a><p>If I had more time I&#x27;d submit a PR to integrate it...
make3将近 8 年前
I wonder what&#x27;s happening with OpenAI. Most big names are leaving.
评论 #14640775 未加载
zach417将近 8 年前
I echo all of your issues with running Universe. I have a decrepit Macbook, and it was actually not possible for me to use it at all.
评论 #14638710 未加载
tomjacobs将近 8 年前
Missed opportunity for a Rick and Morty Microverse reference here as the name
Cellestro将近 8 年前
Congratulations on the initiative, it looks very cool! Indeed, we found that running asynchronous environments, while possible, proved to be too cumbersome for research. We&#x27;re now working on a synchronous set of environments for universe that are easier to use.