Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Pac-Man recreated with a GAN trained on 50k game episodes

blogs.nvidia.com

Post reply on HN
Print view View original

101–110 of 154 posts

Re: Pac-Man recreated with a GAN trained on 50k game episodes

#101

Post by rootusrootus »

SkyBelow wrote: post #89
lumost wrote: post #70

I've observed an increasing trend with modern AI research that A) It's no longer practical to incorporate into real products given the hardware costs for training and inference. In NLP, large transformer based models can easily occupy a 5 thousand dollar GPU at a paltry 50-60 words per second processing pace. B) The Research often demonstrates things that existing non-ML toolchains were already reasonable at doing wi…

It reminds me of older electric cars. They worked. They could get you around. But they did so at higher price, slower speeds, or other factors that a consumer wasn't willing to pay for. Thus they were regulated to novelty items that some work campuses would have and concept cars that would be shown from time to time. But eventually we reached the point where they are not only competitive, they may be positioned to fu…

It seems like there is not a lot of novel technology in an EV; what made them viable was steady improvements in battery technology so that the intersection between range, price, and weight is in a place that is financially feasible. Combined with continued improvements in battery, the already huge gain in fuel efficiency (for a typical car at least) will be what causes them to overtake ICEVs.

Re: Pac-Man recreated with a GAN trained on 50k game episodes

#102

Post by YeGoblynQueenne »

SkyBelow wrote: post #89
lumost wrote: post #70

I've observed an increasing trend with modern AI research that A) It's no longer practical to incorporate into real products given the hardware costs for training and inference. In NLP, large transformer based models can easily occupy a 5 thousand dollar GPU at a paltry 50-60 words per second processing pace. B) The Research often demonstrates things that existing non-ML toolchains were already reasonable at doing wi…

It reminds me of older electric cars. They worked. They could get you around. But they did so at higher price, slower speeds, or other factors that a consumer wasn't willing to pay for. Thus they were regulated to novelty items that some work campuses would have and concept cars that would be shown from time to time. But eventually we reached the point where they are not only competitive, they may be positioned to fu…

Electric cars of the era you describe could at least go. The approach described in this paper doesn't "go". Its goal is to "learn a simulator [to train a robotic agent] simply by watching an agent interacting with the environment".

The learning part is done by training on the output of an existing simulator. Obviously, if you already have the output of a simulator, you don't need to learn a simulator so the only real use of this approach is to train a simulator by watching a human agent interacting with the environment. So the idea is to eventually apply this thing to training robots to interact in the real world by watching how humans do it (this is not clearly stated in the paper; it's my charitable interpretation of the very poor motivation of learning a simulator from the output of an existing simulator).

The only problem is that creating an accurate simulation of a human's interaction with the real world (let alone an accurate simulation of the real world) is prohibitively expensive. Thus, for the time being, this kind of idea works only in very, very simple environments like PacMan.

So we don't yet have an expensive prototype electic car that nevertheless can do everything an er other kind of car can do. We have... not sure what. The description of an idea with an example of how such a thing could be achieved in the far future, when we can train robots to act like humans by watching us.

Re: Pac-Man recreated with a GAN trained on 50k game episodes

#103

Post by DeonPenny »

lumost wrote: post #70

I've observed an increasing trend with modern AI research that A) It's no longer practical to incorporate into real products given the hardware costs for training and inference. In NLP, large transformer based models can easily occupy a 5 thousand dollar GPU at a paltry 50-60 words per second processing pace. B) The Research often demonstrates things that existing non-ML toolchains were already reasonable at doing wi…

Researchers are not entrepreneurs. It's for the entrepreneurs and business to find a way to make this random game into something profitable.

Re: Pac-Man recreated with a GAN trained on 50k game episodes

#104

Post by Terretta »

tlarkworthy wrote: post #56
Terretta wrote: post #53

Earlier quoted context omitted.

Aside, but, I'm said this is an acceptable use of the term AI, to mean "big statistics".

I find it amusing you seem sad because one hyped term was used, only to go on to prefix the word statistics with "big". Statistics was comfortable with large quantities since inception.

I agree with you, ‘big data’ is silly too.

Re: Pac-Man recreated with a GAN trained on 50k game episodes

#105

Post by esoterica »

YeGoblynQueenne wrote: post #93
esoterica wrote: post #90

Earlier quoted context omitted.

I don't know where you got the idea that the purpose of an academic paper is to produce the schematics for a working commercial product. The fact that it's novel and interesting is sufficient grounds for producing a paper about it. Why do you think projects like these are solely the domain of blog posts?

That is not true for academic papers in computer science and particularly in machine learning were contributions are expected to have some sort of practical application that is at least possible to foresee. Indeed, a common criticism against papers in machine learning is "what is a real-world application of this approach?". If you can't answer that convincingly then it's unlikely your paper gets published, particular…

> That is not true for academic papers in computer science and particularly in machine learning were contributions are expected to have some sort of practical application that is at least possible to foresee.

This is 100% not true for computer science and I don't know why you think it's true. There's are entire branches of computer science (e.g. complexity theory) that are highly theoretical and are certainly not geared towards immediate practical application.

Even in more applied branches of computer science like ML there are many papers that are geared more towards the fundamental research side of things.

> saying that it aims to "learn a simulator [for robotic agents] by simply watching an agent interact with an environment".

Saying that a paper could lead to a useful application down the line is not the same that saying that the paper is intended to describe the schematics for a specific useful application.

Re: Pac-Man recreated with a GAN trained on 50k game episodes

#106

Post by eeZah7Ux »

lumost wrote: post #70

I've observed an increasing trend with modern AI research that A) It's no longer practical to incorporate into real products given the hardware costs for training and inference. In NLP, large transformer based models can easily occupy a 5 thousand dollar GPU at a paltry 50-60 words per second processing pace. B) The Research often demonstrates things that existing non-ML toolchains were already reasonable at doing wi…

> 3-4 orders of magnitude higher resource consumption

Just 10000x? More like 6 orders of magnitude

Re: Pac-Man recreated with a GAN trained on 50k game episodes

#107

Post by ars »

Bjartr wrote: post #63
simias wrote: post #51

Earlier quoted context omitted.

The other answers to your comment are great but I also want to point out that you can feel your brain doing that if you play some videogames for a long time. You can feel that you develop an intuition for the way the game will behave even if the in-game physics are very different from our real word, showing the plasticity of our brain. If we had a fully hardcoded physics simulator in our head we couldn't "think with…

This reminds me of a story when VR was just becoming big. Someone had been playing the game "Job Simulator" in the restaurant level where you play as a cook in the kitchen of a restaurant. The player was constantly looking in the fridge for things and realized they didn't have to open the door since the game didn't stop their head from clipping through. So they would just stick their head through the door when they n…

If I play a racing game for a while, I find it pretty difficult to drive safely right afterward.

I have not actually had any incidents, but I have to consciously not crash into things, or drive wildly.

Re: Pac-Man recreated with a GAN trained on 50k game episodes

#108

Post by QuixoticQuibit »

madenine wrote: post #31
cooljacob204 wrote: post #20

Earlier quoted context omitted.

Servers are already running AI to try and counter cheating in quite a few games.

2018 video on Valve's use of deep learning for anti-cheat in CSGO https://www.youtube.com/watch?v=ObhK8lUfIlc

Neat, I wasn’t aware of this application of DL. Very cool!

Re: Pac-Man recreated with a GAN trained on 50k game episodes

#109

Post by danans »

mettamage wrote: post #75
danans wrote: post #73

So AI can learn to play the games, but can it learn to create them? That is, can it create compelling cultural novelty? The most successful examples I've heard are computer generated classical music, but generating abstract art like classical music is simpler than generating art that relies on coherent language and narrative, like opera, or the blues, or a good novel.

There has been research on it. I forgot the source and can’t find it easily, but if you email Julian Togelius, he might know. He doesn’t focus on game AI research and is the only name I can remember at the moment. Edit: oh, Togelius also seemed to have researched it. https://www.google.com/amp/s/www.technologyreview.com/2017/1...

Cool. Thanks for that link!

Re: Pac-Man recreated with a GAN trained on 50k game episodes

#110

Post by longtom »

itchyjunk wrote: post #8

Thought experiment: Is my brain also generating reality without a physics engine and is filling in details that it thinks should be there? If true, what are the advantages/disadvantages of it?

This is a compression artifact. You cannot process everything that is going on, so your brain fills in the missing or uncertain parts with a similar (abstract) computation that it already has experienced in a similar manner. It is abstract in the sense that it only replicates some aspects of the real thing. Perceiving such a compression/abstraction is better for increasing reproductive success than receiving just noise when data is missing or has low resolution.
Post reply on HN