[Eagle Creek]

Originally Posted by antialiasis

Because CAPTCHA images are computer-generated images containing letters and/or numbers with varying amount of distortion. Computerized pattern recognition programs can be programmed to read the distorted letters in CAPTCHA images without much trouble, but identifying an object in an image is a dramatically more difficult task requiring far more elaborate, advanced AI.

Say you took your avatar and asked in NoSpam! what that image shows. The average human should without too much trouble be able to answer "eagle". However, a computer doesn't have years of experience of looking at pictures of animals and eagles with a mind-numbingly elaborate neural network that specializes in recognizing objects such as animals in visual data, and thus it will not have any idea what the heck your avatar is. Letters are simple patterns that a bot can be programmed to recognize; recognizing 3-D objects in 2-D images from any angle is something that we humans have been specializing ourselves in for millions of years of evolutionary history and a varying number of years of personal experience. Not to mention that the bot would have to keep a database of rules to recognize an absolutely astronomical number of objects in order to be able to take on an arbitrary image with no clue whatsoever as to what it contains. Artificial intelligence just can't do it, at least not yet.

HI,

I understand the system of images being read etc.. (very nice explanaiton btw!). But what I don't understand is this. You say:

Or you can go with inserting a simple image with HTML and ask what is shown on the image.

Then you still have exactly the same, dont you? You present a simple image like .JPG and then it can be read? Or am I wrong understanding what you are writing .
I will definetly going to use this hack .