I wonder if one could construct a metric for how 'interesting' a drawing is- then throw it at a genetic algorithm? I'm guessing most likely the system is inherently unstable, so two 'interesting' parents don't necessarily create 'interesting' offspring

You don't have to use crossover, you can just make small mutations. That's much more likely to result in interesting offspring.

Would be great to see some articles on their technical approach, I assume they have used techniques to reach those scales based upon the fact the average pair of players are too far away from each other to be visible

They trust client data. It makes scaling a lot easier (and cheating/griefing).

i second that, immediately went to check if the source code is available, but sadly nothing.

It seems hard to general to get your hands on any modding code. Apparently the modding culture doesn't encourage opensource as much as other hacking circles.

Modding seems very 'credit-heavy' and it can be difficult to get credit for your work when a bigger, better team can simply fork your code and call it their own. Then again I don't have any experience in this field so I could be completely off-target.

Maybe it's my experience of living in central London but I can certainly think of many 'demographics' that would be interested in soundproof bedrooms on the street...

Nice to see PCA in an HN article, it's a very powerful tool.

For those struggling to get the example in this article, I find PCA easier to understand given visual examples, and in less dimensions (try http://en.wikipedia.org/wiki/File:GaussianScatterPCA.png)

Note how this dataset is two dimensional in nature, and PCA yields two vectors. The first gives the direction of the greatest variation, and the next gives the variation orthogonally to the first.

An awesome use of PCA is for facial detection, a method called 'Eigenfaces' http://en.wikipedia.org/wiki/Eigenface

I wrote a blog post with more detail, and lots of intuitive examples. see http://jeremykun.wordpress.com/2011/07/27/eigenfaces/

FYI Eigenfaces was a ground-breaking theory when introduced...almost 25 years ago. It's no longer used in any serious way for practical face recognition applications.

eigenfaces are very cool. i did a bunch of work with them a few years ago.

I think it depends on the genre at hand. Massive AAA singleplayer titles that tend towards being more ‘interactive cinematic experiences’ would suffer from this certainly, but I’m sure (as other have posted) more free-form games such as Minecraft it can only be a boon for. With the former, a play-through potentially makes your own experience feel a bit stale, whereas with creative games it more likely than not just stimulates your own imagination about how the possibilities of the sandbox the game gives.

Surely censoring the functionality is completely impossible, due to the dynamic nature of JS?

alert("blah")

can just as easily be made to be

window["alert"]("blah"), where the string 'alert' could be made by a very convoluted manner. Solving this would be akin to solving the halting problem, no?

Static typing is nice, but does this support type inference?

Some keyboard shortcuts would be fantastic (Especially delete!)

I would think the average person isn't lucky enough have an inventory of multiple thousand dollars of high-end computing equipment. This kind of system would be perfect in developing countries, or as the project suggests, for education.

Many people in developing countries have mobile phones, and smartphones are rapidly becoming the computer of choice for them.

Also, remember that today's high-end is tomorrow's low-end. Why develop stuff for the past? Don't think of the developing world as a dumping ground for our worn-out ideas and wares, because that's not what it is.

At first I was wondering about the benefits of using such a system vs. something like deploying the raspberry pi en masse (which is a full system for under $50), but after some thought it comes down to the fact that the average utilization of any computer most of the time is very low, but we really notice the CPU power when we do something intensive (although not very often). Each user on this system could potentially feel the same power as a full desktop, as the probability of multiple users requiring intensive computation at the same time is relatively low. Kind of like how web providers quote 10MB per customer, but their infrastructure only has to cater for a tiny fraction of the potential full utilization.

The hardware in question very likely has computing power (and associated cost) not that far off from that of the Raspberry Pi.

You also have to factor in the cost of the machine used to drive these USB display drivers -- with 4 users on a $500 machine, you're still talking about $125/user.

I'm still wondering why we couldn't just make thin clients using raspberry pi.