Eli - I wrote a paper on these several months ago, and that was one of my main issues with them as well. They don't simply require that one be human in order to pass. They require further that one be an fully-able human (for want of a better term) with a certain degree of human intelligence. This will exclude a certain number of humans who should nonetheless have access to the resources or services that the CATPCHAs protect, and providers must consequently provide a workaround for these users.
3
John Saxton
12-10-2002
09:35 PM ET (US)
Serveral sites already employ this kind of test when you sign up for a new account. Spammers wanting to sign up for a bunch of accounts at Hotmail (or whatever) now just have their scripts prompt a human to answer the question, and the spammers just employ minimum wage workers to sit there doing "human required" tasks on behalf of the computer.
2
jerwin
12-10-2002
08:42 PM ET (US)
I tried (and failed) the pix test several times. The questions were phrased so as to elicit the plural, but often, only the singular was accepted-- i.e. "riddle" was accepted, "riddles" was not. Synonyms were rejected "infant" was wrong-- "baby" was correct.
The pictures were small, and sometimes it was difficult to discern the relevant imagery.
Perhaps a certain level of artificial intelligence is required to "grade" such tests, though I've seen more linguistic sophistication in TADS (Text Adventure Development System) programs.
And, of course, users of "lynx" (and associated screen reading software) are left out in the dark.
1
Eli the Bearded
12-10-2002
07:08 PM ET (US)
These all require sight.
The pix one in particular is bad because at the size the images are shown the objects might not be visible. One I got had five images that I could clearly see had water and one that looked like a buffet table. The answer was "boat" (or ship or several other similar words).