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

131–140 of 154 posts

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

#131

Post by angel_j »

anigbrowl wrote: post #127
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…

Your comment reminds me of snide dismissals of personal computing in the 1980s - looks worse than [existing thing], very expensive, requires an army of technicians, probably explodes unpredictably, overall just a terrible investment.

Followed by decades long monopoly over expensive new technology, likely crushing waves of innovation and usage, in exchange for slow, buggy, not-even-automated spreadsheets; plus the hemming of governments and organizations into endless software rental contracts and platforms, that maybe don't add value as fast as Moore's Law.

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

#132

Post by lumost »

angel_j wrote: post #131
anigbrowl wrote: post #127

Earlier quoted context omitted.

Your comment reminds me of snide dismissals of personal computing in the 1980s - looks worse than [existing thing], very expensive, requires an army of technicians, probably explodes unpredictably, overall just a terrible investment.

Followed by decades long monopoly over expensive new technology, likely crushing waves of innovation and usage, in exchange for slow, buggy, not-even-automated spreadsheets; plus the hemming of governments and organizations into endless software rental contracts and platforms, that maybe don't add value as fast as Moore's Law.

that's a fair criticism, however it's not a given that any product category will end up with the same growth curve of PCs, the web, cloud, and mobile.

thinking back on the PC, Web, cloud, and mobile growth periods, each came with a transformative change in the capabilities, cost, and availability of computing power. We're only seeing two of these aspects in AI today - with the cost side heading in the wrong direction.

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

#133

Post by pmiller2 »

YeGoblynQueenne wrote: post #77
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…

Or, to paraphrase: great paper, no application.

A lot of the ML techniques we're putting to use today came out of papers from the 80s. These techniques weren't practical to do much beyond toy examples, but they were proven to work. Now, we're able to use those very techniques to teach a machine to identify cat videos with 75% accuracy using unsupervised learning! [0]

---

[0]: https://www.wired.com/2012/06/google-x-neural-network/

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

#134

Post by quickthrower2 »

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?

It means the original programmers of PAC-MAN didn't "understand" the game either :-). There was never an "engine"

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

#136

Post by eru »

pbhjpbhj wrote: post #19

What're the total number of possible states in Pacman?

It's not so much the total number, but the dimensionality.

Eg if you have a blob that can move in x and y coordinates, it doesn't get more complicated just because you move from 32-bit floating point numbers to 64-bit floating point numbers. Even though the number of states squares.

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

#137

Post by eru »

justicezyx wrote: post #67
dontreact wrote: post #66

I predict methods like this will be used to create a simulation environment that approximates the real world. Reinforcement learning folks are running out of game engines win inside of. I believe the best way to transcend this coming limitation will be to take a learning based approach similar to this.

I mean, it's pretty clear that we are already living in such simulation, right?

What evidence do you have? How would the world look different, if we weren't in a simulation?

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

#138

Post by voldacar »

Terretta wrote: post #53
cooljacob204 wrote: post #20

Earlier quoted context omitted.

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

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

AI these days just means principal component analysis

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

#139

Post by m463 »

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…

Lots of new stuff sucks.

The english longbow was vastly superior to the first guns that came out, which were slow to load, unreliable to shoot, didn't work in the rain, and were extremely inaccurate.

3d printed stuff sucks. It's not very strong, it's not very pretty and takes a long time to create.

Meanwhile, take a look at this ML photo processing technique:

https://www.youtube.com/watch?v=bcZFQ3f26pA

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

#140

Post by 101404 »

bobly_today wrote: post #128
101404 wrote: post #126

Earlier quoted context omitted.

Because it's a pile of if-else statements shuffled until it was able to perform a single specialized thing. Any housefly would by a thousand times as "conscious".

What makes you so certain a house fly doesn’t have a form of rudimentary consciousness? It might be a spectrum.

How about a computer? It has if else statements too.
Post reply on HN