Unwitting Distributed Genetic Programming via AJAX | Hacker News Reader

Unwitting Distributed Genetic Programming via AJAX

hampshire.edu

22 pointspsawaya12 comments

Mininova tried this [1] and got something like ~1000 MIPS out of it. That was using Firefox 1.5 though, running their IPS counter on Chromium dev on a modern Core2 duo gives me 71428571 IPS, more than a 30x improvement over the per-user numbers they cite.

Back of the envelope math: ceteris paribus that would put a 3-million pageviews/day somewhere along the lines of a dual-Itanium box today. Facebook pushes ~500 million pageviews so has access to ~300 CPUs worth of computing power in the form of its visitors' javascript engines (~5,600,000 MIPS).

[1] http://blog.mininova.org/articles/2005/11/17/mininova-the-ja...

I detest anyone who would try to sap my spare computing resources without first informing me. With consent and full disclosure, however, I would likely participate in such a system.

Would you put a bit of javascript on your site that ran GP on your visitor's computers? With an option to disable it, of course.

I would consider an opt-in system. For example, if I were paid $x for every 1000 iterations run, I could see myself offering users a choice between running an iteration of GP on every page load or viewing one more ad of equivalent value.

Of course, I'm sure there are many others who would see nothing wrong with using a few milliseconds of the unsuspecting masses' time. I'm just not one of them.

What if the use of your computing resources were limited to only one mutation/tests? I.e., something that would slow your page load down a few milliseconds, then no more?

If you can spare one millisecond, then surely you could also spare two...

http://plato.stanford.edu/entries/sorites-paradox/

I detest anyone who would try to sap my spare computing resources without first informing me.

So adopt the maxim: People who steal my computing resources aren't entitled to correct results. :)

My thoughts exactly :). I treat marketing surveys and loyalty card application forms similarly, unless I really really like the company.

Why? Your reaction seems sort of knee-jerk to me.

Most people are perfectly ok with advertisements on web pages. I would guess that it wouldn't sap your resources more than downloading an advertisement or two.

In fact, for the average user, it might be less noticeable, given that you can load the page first and start loading the code/data for genetic programming afterwards.

Hey, if you don't have NoScript and FlashBlock turned on, you're already explicitly giving any and every Web site permission to do whatever they want with your spare (and not-so-spare) computing resources.

I've always thought this sort of thing would make for an interesting way for publishers of websites to monetize their traffic. Instead of showing ads, a publisher could make it so each page view meant a small unit of work was completed.

A company could do business like an ad network, but instead of matching advertisers with publishers and serving ads, it would match groups needing computing resources with publishers and handle everything involved with serving out units of work and collecting the results.

I like the idea, although it limits the kind of "fitness evaluations" that one can carry out.

For instance, you aren't going to be compiling and running c-programs client side... http://news.ycombinator.com/item?id=1084015

Reply on news.ycombinator.com