Artificial Intelligence is becoming a convergence of neurobiology and computer science. It was once thought that intelligence can be created by statistical analysis of data, but there is now acceptance that awareness plays a large part in autonomy. However, an imitation of consciousness is actually the more difficult problem to solve.
Object recognition plays a fundamental role in creating awareness. Before you can tell a robot to open a door, it first must know what a door is. To compound the problem, not all doors are created equal and not all large rectangular objects are doors. Once the system is finally aware of the existence of a door, only then can the computational operations involved in opening the door commence.
If there were only a few objects in the world, this problem may be a little simpler. Obviously that's not the case and to classify multiple objects from multiple vantage points for image recognition software already in existence to build a database of is a much larger undertaking than even reading this sentence.
form & function can coexist
Down swoops the internet to save the day! Where some have compiled large collections of images and asked for the participation of others, only small groups of willing participants have been inclined to provide input. It is possible to accomplish classification of a large image data set by a practical self serving method. Bring forth the CAPTCHA; that loathed gatekeeper of obscured letters that strain even a human's capabilities of object recognition.
While Google has made a valiant effort to scan Shakespeare into the internet and save us all from the torment of spammers in the process, maybe it's time to move on. Instead, we can make something a bit more natural for humans and solve a problem pertaining to the next stage of computing.
Associating objects with language is a human ability that has been thousands of years in the making and is an excellent human verification test. Given that any single language is roughly common among all who speak it, people shown an image under the assumption that they speak that language should identify all objects similarly. Thus Minotaur shows an image containing a simple object and asks the user to freely identify the object. To determine correctness, the input is statistically compared to the input of others.
The CAPTCHA is a self serving data input that doesn't have to be overly complex or intrusive. However, if you're going to use one, make the results useful. Support the next generation of AI!