Programming the Turk?

Wikipedia says "The Amazon Mechanical Turk (MTurk) is one of the suite of Amazon Web Services, a crowdsourcing marketplace that enables computer programs to co-ordinate the use of human intelligence to perform tasks which computers are unable to do".

OK.. and Amazon's page at https://www.mturk.com/mturk/welcome says that you can "Make Money by working on HITS (Human Intelligence Tasks)". You "work from home", "Choose your own work hours", and "Get paid for doing good work". In case you still didn't understand, a graphic shows you how it works:

I'm surprised they didn't have exclamation marks after "Earn money".

I think if you are living in desperate poverty in a third world country, the compensation offered for these tasks might be reasonable. These tasks might take 15 seconds or so to complete and pay pennies. Pennies as in one or two pennies. For example, a vitamin company wants to see if a picture of its products appears on certain websites. They have 1600 or so websites to look at and are paying 2 cents per site. At that pay scale, you'd need to find the image link in 12 seconds to earn just $5.00 an hour - and of course the most you can earn here is 1600 x .02 or $32.00. This isn't going to pay your rent.

However, let's go back to the original premise: this is "to co-ordinate the use of human intelligence to perform tasks which computers are unable to do". But who decided that a computer was unable to do this task? Some human, of course, and frankly most humans (outside of the readers of this site, of course) have barely a clue what their computers really can do. I'd bet that the HIT for the image lookup might have been written by someone in advertising who thinks that the task requires a human because there's no such function in Microsoft Excel. But in the real world of computers, it is of course possible to compare images.

Nor are these people even looking for tasks that difficult. One I looked at yesterday was paying a similar amount for finding out if Wikipedia had a reference to a company and if the company itself had a web page. Again, Microsoft Excel fails miserably at this but I bet most of my readers could hack out code for this in their sleep. So..

So you'd think that coders would jump on this. I don't think they have, and it;'s probably because of the scope of the HITS. If there are only 1600 websites to process at 2 cents each, is it worth my time to write specialized code that will at best net me $32.00? Probably not, though many of these requests are very similar so after hacking out a few you might gain some efficiency - if you could get the code ready to run in 5 minutes or less you'd be making decent money, right?

Well, yeah, but you've still got to report the results back. No problem, you say, writing a little CGI POST isn't exactly rocket science. Right you are, but Wikipedia has a warning:

Programmers have developed various browser extensions and scripts designed to simplify the process of completing HITs. According to the Amazon Web Services Blog, however, Amazon appears to disapprove of the ones that automate the process 100% and take out the human element. Accounts using so-called automated bots have been banned.

I find that a little disturbing, don't you? Why on earth don't they set up a site that HELPS automated solutions? The requesters obviously just want results and are simply too naive to realize what can and cannot be done by computer. Why put artificial barriers in the way of those who do know?

Oh well - it is interesting and might be worth peeking at. Somebody just might post a task with enough available work to make it worthwhile even with these impediments.