Comment Re:Captcha 2.0? (Score 1) 267
I think the problem there would be that the text of the question itself would become the point of attack for the automated solver.
But I like the 3-D idea. My thought was that answering the question correctly doesn't require a "perfect" answer, but rather something that is score as being "good enough".
Differences in language or terminology would be included in the answer, so essentially each image in the test library would have a series of words and scores for those words.
Like any approach there will certainly be new-found complexities, but overall I just think that making the test based on imagery essentially makes for a far superior test than one based on letters. The evolution of the text-based tests just arrives at our current situation where obfuscation of the letters makes it hard for people too.
In the case of images, training a computer to "understand" an image would, as far as I can see, be vastly more complicated than training it to see 26 letters and 10 numbers under some messy lines and distortions.
But I like the 3-D idea. My thought was that answering the question correctly doesn't require a "perfect" answer, but rather something that is score as being "good enough".
Differences in language or terminology would be included in the answer, so essentially each image in the test library would have a series of words and scores for those words.
Like any approach there will certainly be new-found complexities, but overall I just think that making the test based on imagery essentially makes for a far superior test than one based on letters. The evolution of the text-based tests just arrives at our current situation where obfuscation of the letters makes it hard for people too.
In the case of images, training a computer to "understand" an image would, as far as I can see, be vastly more complicated than training it to see 26 letters and 10 numbers under some messy lines and distortions.
