<?xml version="1.0" encoding="utf-8"?><!-- generator="wordpress/2.3.2" -->
<rss version="2.0"
	xmlns:content="http://purl.org/rss/1.0/modules/content/"
	xmlns:dc="http://purl.org/dc/elements/1.1/"
	>
<channel>
	<title>Comments on: Amazon Mechanical Turk</title>
	<link>http://www.chausse.org/2005/12/amazon-mechanical-turk/</link>
	<description>A blog about web technology and culture</description>
	<pubDate>Thu, 04 Dec 2008 18:41:12 +0000</pubDate>
	<generator>http://wordpress.org/?v=2.3.2</generator>
		<item>
		<title>By: Joel McCoy</title>
		<link>http://www.chausse.org/2005/12/amazon-mechanical-turk/#comment-4031</link>
		<dc:creator>Joel McCoy</dc:creator>
		<pubDate>Sat, 03 Dec 2005 19:37:34 +0000</pubDate>
		<guid>http://www.chausse.org/2005/12/amazon-mechanical-turk/#comment-4031</guid>
		<description>I had the same thought about "click-thru" services or AllAdvantage where the idea was to prompt an unnecessary behavior and then pay someone to do it, which clearly isn't sustainable. Micropayments for tasks which actually do need to be accomplished seems a far better thought out business model.

And I'm sure the potential for malicious HITs is substantial, but then I really do wonder what the HIT population will look like once Amazon stops giving us the Confirm Artists Names to tread water with. So far the non-Amazon HITs have been French/English translation, Podcast transcription, and Auto Part Description writing- I can't think of too many projects that have a substantial enough amount of work to be done that gracefully break into HITs as well as the examples we've seen so far.

For the CAPTCHAs (though I'm sure that was an example shot from the hip) I can't think of a way to guarantee any success with that method since most all of them have some timer, and it would be pretty difficult to get the image fed over to the Turk and garner enough responses to do majority rule acceptance before it would expire. Auto-accept is an option, but if someone's willing to help willfully circumvent CAPTCHAs for a penny, they're probably not above just submitting '69lol' over and over again.</description>
		<content:encoded><![CDATA[<p>I had the same thought about &#8220;click-thru&#8221; services or AllAdvantage where the idea was to prompt an unnecessary behavior and then pay someone to do it, which clearly isn&#8217;t sustainable. Micropayments for tasks which actually do need to be accomplished seems a far better thought out business model.</p>
<p>And I&#8217;m sure the potential for malicious HITs is substantial, but then I really do wonder what the HIT population will look like once Amazon stops giving us the Confirm Artists Names to tread water with. So far the non-Amazon HITs have been French/English translation, Podcast transcription, and Auto Part Description writing- I can&#8217;t think of too many projects that have a substantial enough amount of work to be done that gracefully break into HITs as well as the examples we&#8217;ve seen so far.</p>
<p>For the CAPTCHAs (though I&#8217;m sure that was an example shot from the hip) I can&#8217;t think of a way to guarantee any success with that method since most all of them have some timer, and it would be pretty difficult to get the image fed over to the Turk and garner enough responses to do majority rule acceptance before it would expire. Auto-accept is an option, but if someone&#8217;s willing to help willfully circumvent CAPTCHAs for a penny, they&#8217;re probably not above just submitting &#8216;69lol&#8217; over and over again.</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Jeff</title>
		<link>http://www.chausse.org/2005/12/amazon-mechanical-turk/#comment-4030</link>
		<dc:creator>Jeff</dc:creator>
		<pubDate>Sat, 03 Dec 2005 18:59:51 +0000</pubDate>
		<guid>http://www.chausse.org/2005/12/amazon-mechanical-turk/#comment-4030</guid>
		<description>Joel,

I didn't mean to imply that Turking was a waste of time.  In fact, I think it's great that someone finally found a legitimate "pay per click" business model that actually makes sense (remember all the "we pay you to click on ads!" sites back in the day?).  

What I'm wondering though is how long until Turking is used for evil? For circumventing CAPTCHAs, etc.  Clearly, it's inevitable..</description>
		<content:encoded><![CDATA[<p>Joel,</p>
<p>I didn&#8217;t mean to imply that Turking was a waste of time.  In fact, I think it&#8217;s great that someone finally found a legitimate &#8220;pay per click&#8221; business model that actually makes sense (remember all the &#8220;we pay you to click on ads!&#8221; sites back in the day?).  </p>
<p>What I&#8217;m wondering though is how long until Turking is used for evil? For circumventing CAPTCHAs, etc.  Clearly, it&#8217;s inevitable..</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Joel McCoy</title>
		<link>http://www.chausse.org/2005/12/amazon-mechanical-turk/#comment-4029</link>
		<dc:creator>Joel McCoy</dc:creator>
		<pubDate>Sat, 03 Dec 2005 18:45:08 +0000</pubDate>
		<guid>http://www.chausse.org/2005/12/amazon-mechanical-turk/#comment-4029</guid>
		<description>I think you're right on Turking being somewhere between aluminum cans and independant wealth, but it may be closer to the other side of the spectrum than you insinuated. Right now the system is at the lowest point (inre: profit-per-person) since its creation, this still means ~$8-13/hr for solid work on the Music HITs you mentioned.

This is through use of Greasemonkey/Opera user scripts for improved usability without wholly automating the process, which is of course against the TOS. As an example, here are my stats from 4 hours' work yesterday at the lowest yield I've gotten so far-
Submitted:2461&#124;Accepted:1991&#124;Rejected:467&#124;Pending:3&#124;Earned:$39.82

TMYK</description>
		<content:encoded><![CDATA[<p>I think you&#8217;re right on Turking being somewhere between aluminum cans and independant wealth, but it may be closer to the other side of the spectrum than you insinuated. Right now the system is at the lowest point (inre: profit-per-person) since its creation, this still means ~$8-13/hr for solid work on the Music HITs you mentioned.</p>
<p>This is through use of Greasemonkey/Opera user scripts for improved usability without wholly automating the process, which is of course against the TOS. As an example, here are my stats from 4 hours&#8217; work yesterday at the lowest yield I&#8217;ve gotten so far-<br />
Submitted:2461|Accepted:1991|Rejected:467|Pending:3|Earned:$39.82</p>
<p>TMYK</p>
]]></content:encoded>
	</item>
</channel>
</rss>
