Human Computation...The Web as training grounds for AI

Step through the looking glass to see where human computation could lead us...

The whole idea behind human computation is intriguing.  The premise is that our technology is only as smart as we make it, and I quite agree.  It is time to step away from using the web as simply a time-sucker from our daily lives.  We need to stop filling it with the same garbage and refuse that fills our streets and the millions of acres of landfills around the country...and the world.  Can we make the web better?  Can we make it more efficient?  Are you motivated yet?  Probably not, knowing most busy-bodies in the world today.  Well, the good news is that you may be doing your part already, totally unaware of your active participation in this process known as human computation. Designers have produced games that are geared towards developing computer algorithms which would be used, in turn, to develop better programs, search engine results, etc... Although the game designers do not try to trick you into playing, many of you may not have known (if you're like me and you just dive in without reading) that the games you have played are producing such data. 

Many of you may have seen Google's new Image Labeler or perhaps you have heard of the ESP game.  These two particular games use human input to label images on the web. There are other games, as well, that go a step further and label particular objects within images.  Such data, given by humans, can then be used to train computers to recognize certain objects within any given image for which it has data.  Other projects include the input of "common-sense" knowledge, which could help computers, for instance, filter data in order to return more applicable results in search engines.  But does it stop there? Of course not! 

The future applications of such methods are obvious.  If we can develop programming languages that center around the human intellect, then we can pass human intelligence onto machines, thus creating AI (artificial intelligence).  Images are most important, at first.  If we can develop digital libraries that can teach computers how to "see" on the web, then the ultimate step would be to create a robot in the physical world that can see (of course this would take a continual evolution of these human computation methods and the refinement and application of the data it produces to computers, as well as the development of other algorithms needed for other robot processes).  After all, images are simply snapshots of our world.  Combine these snapshots, along with some "common-sense" knowledge and movable parts and a robot can cross the street without getting run over by a car.

The implications of human computation are truly fascinating because it opens the realm of the web.  The web can be seen  as a a singular, living organism, throwing out data left and right.  Programs developed with human computation in mind, simply harness this data and organize it into usable training data for new programs.  The continual refinement of data and programming will undoubtedly foster the first truly intelligent AI.  We already have programs that refine themselves...programs that teach themselves is one way of putting it.  Of course, the best thing about robots is that we can program them to be the way we want them to be.  Or, can we...?

Since I am an educator, let's talk education. The immediate implications of human computation research for educators could result in higher efficiency of student research on the web, interactive learning programs that respond to a student's previous answers in a very specific way, thus enabling the program to teach the student based on cognitive infrastructures already in place within the student's brain, maps that can talk about history, plant life, animal life, etc... that are common to the area specified on the map (including map search functions; search for "monkey" and you can be shown every place in the world where monkeys live and all about them), and calculators that can verbally explain computations and help students with common errors in processing (more complex than a help menu like the one in "Windows").  The applications are endless because human computation opens the window to redefining the web to a living, breathing, human computer that can eventually program itself, meaning humans can develop programs so computers can eventually map enough data that comes from humans to program an infinite number of human processes in binary form using algorithms (or maybe some new type of processing yet to come).

Whatever comes of human computation in the future, one thing is for ceratin.  It's going to be interesting. Kudos to Dr. Luis von Ahn and his team!

First update to this post:
I found this video online which actually shows the development of a tool that will lead to cyborgs.