Tuesday, 3 July 2007

Human Computation Projects

This is interesting stuff. From Wired. So I wrote some thoughts on it.

First we started linking all those idle processors around the world to do distributed network calculation projects, mostly analysing data, like SETI or decoding complex molecules. The PS3 is doing a similar thing with the Cell chip, why only have PCs do it when there are so many consoles around the world? [And at the same time buy a Blu Ray player instead of a standalone. Since so may consoles have DVD drives I'm surprised standalone DVD players still sell.]

Now we are starting to use all those idle Human moments to do something equally important. It's all about efficiency which I like. There are some tasks which only Humans can do and computers can't and it's about harvesting that ability in a useful manor. The challenge is doing it in a way that's fun, because otherwise people won't pick it up and "do it", and the way to get people to do it is to make it a game and "play it".

Which is what this guy Luis von Ahn has and is trying to do. He invented Captchas which are everywhere now, the pictures with scrabbled letters you type to verify you are a Human and not a bot on website. * The problem is motivating people to do something, the traditional way is paying money, or this way of making it fun, I found it interesting that the people who are combating Captchas are hiring 3rd world people to crack them or paying with porn. Perhaps this is all proof that Human nature is selfish.

[*Frankly I'm not sure why anyone would need to be at Uni to come up with these ideas, but it must be useful having the backing of a PHD advisor Professor. If anything I could say it was his programming skills which enabled this. In the electronic age, programmers are the new world builders, the new gods. There must be thousands of people coming up with bright ideas outside of Uni, I could do this, and if you want to talk about efficiency you should talk about harvesting these people, forming groups of individuals who specialise in certain skills, and then putting these groups together to make it happen.]

So out of all of these, the way of making something fun would seem to be the best option to me. Bring some fun into the world. Really it's addictive and repetitive behaviour, which the user (interesting phrase- who's using who here?) or player finds, well.. addictive. I find it interesting that he lied for his first game, the image matching game Matchin, he called it The ESP game. The game itself might be fun and addictive but it proves that perception is equally important for providing that hook, if he hadn't called it that but "Image matching game" I imagine a lot less people would have been interested and maybe it wouldn't have taken off. That's marketing for you. I could put this on an equal par with psychology experiments or studies where the experimenters lie too, but I'm not sure how ethical that really is anyway. It's interesting because I wonder how many other business got started by a lie somewhere along the line.

"Well, if it isn't like that, it won't be fun, and no one will play it," he replies.

This is the problem faced by all of von Ahn's human computation projects. People will contribute their brainpower, but only if they're given an enjoyable, time-killing experience in exchange. Play is the unexpected glue that lashes human brains together into a global overmind. So to build a good human-computation project, you can't merely be a scientist; you also need to be a videogame designer.

This, as it turns out, is a significant hurdle, because few academics are trained in game design. Von Ahn tries to find students who have done it themselves: Law worked for Ubisoft, and others have programmed their own indie Flash games. "Game design is a funny thing," von Ahn admits. "There are people out there who are really good at it, but it's not clear that they can teach it. It's a very intuitive process. It's an art."

This is why I'm interested in this subject even more. God, imagine how much we could have got done it we had harvested Ever"crack" players killing rats for fun. These standalone games necessitate being very simple, draw them in on a simple task which keeps them hooked for hours, but imagine if we could meld it with complex games? I'll come to that in a bit. For now the opportunities come in crobaring in useful human computational tasks to daily activities. Stuff we are going to do anyway. reCaptcha uses a word that when recognised goes to correcting a database entry. I think all this efficiency and using our wasteful cycles is good but there is room for abuse, things should be kept fun, but some horrific psychosocial experiments could be conducted on the race like this and I wouldn't like to see this used for anything sinister. We already have a nation watching Big Brother, even if Human behaviour is one of the most interesting things to watch, it kinda proves they'll sit a round and watch anything.

//Here's one the Sci-fi-ists might not have considered, if we prove that computers can't reach real AI without a human component, and we grow or plug in a human brain component and connected it to machines - matter of time really - once that is done perhaps that computer will force it on another, the computers will be using us not as batteries but as their "learning" component. Scary thoughts.//

You see I grew up with Douglas Adams books, and while this did kinda make me see the world in an absurdest way it already made me think along the lines of people as computer components/program in a giant computer. Life, the Universe and Everything, anyone? So my thinking already is well suited to this. You never know it may actually be true...seems like it more and more everyday.

You look at the Human race and their behaviour, their nature, group behaviour and you really can see them as some type of program. However much you want to infer into that, that part is true. You just have to step back a bit and look at the bigger picture.
"Basically, I want to make all of humanity more efficient by exploiting the human cycles that get wasted," he tells me over lunch at a diner near his lab in Pittsburgh. "As humanity goes online, it's becoming an extremely advanced, large-scale processing unit."
So what types of things could Humans teach that computers can't do? Path finding, Perception, teaching computers to Lie, Humour?

The first way would be analysing human choices in game rather than actual real world computational problems, but then what is learned could be used for real world applications.

Some of the simpler games could be puzzle solving by humans. If you think of that Draw to Play game, it's kinda unique, would a computer be able to work that out? You use the mouse to draw your own platforms while controlling a little man around some obstacles. If you run out of "ink" to draw platforms you have to restart. It's quite a challenge. What if it could analyse the patterns Humans had drawn and then work out how to complete a freshly drawn puzzle challenge, that kind of thinking could then be applied to something else.

FPS- What if the in-game player choices of millions and millions of games being played of Counter Strike, or Doom, or Quake or UT were analysed to make a better military robot? It's combat in a 3d environment by millions of people around the world constantly.

RTS - same thing.

RPGs? Path finding and player choices about what is valuable in their inventory.

Or the second way, rather than analysing player choices in game, these actual computational applications could be put in-game. They talk about advertising in-game now, an idea I hate, but if as part of the game play they could cro bar in a challenge from the real world, it would get solved as part of the in-game puzzle. Interesting thoughts to ponder.

Massively Multiplayer Online Games offer the most diverse player choices imaginable and would also be able to put in actual computational challenges. I'm not sure how players would feel being treated like Guinea pigs, as long as it didn't interfer with the flow and fun, but you could always offer them level up experience or gold etc. The idea of gaming is that the game should remain focused on gaming - ALWAYS.

-hmm interesting thought - how about making a MMOG that' completely honest about being for Human computational projects and ram it full of them, then the players drawn to that would play mini games in the Massive online world solving real world computation problems while "leveling up". Patent pending that thanks.

Unfortunately the only applications that spring to mind for the first method in-game analysis of choices, seem to be militaristic, game of def con anyone? There just aren't enough games out there about feeding the worlds population. And Humans seem to be inheritably destructive in their play...because that is fun.

The thought occurs - that that is because it's fun to survive. Which is a game the Human race has been playing for a very long time. We play as a means to test our abilities in a safe environment, and learn - which is the essence of play, for the time when we are in a dangerous environment. We could teach computers how to survive.