I saw a video recently that said that captchas learn from the user as a way to train AI, but how does it know that you selected the right option? I’m talking about the captchas that tell you to select all the fire hydrants or traffic lights
In: 4
Captchas will give you a set of answers to select where some of them it thinks it knows the answer already and some it is unsure about. If for example you can select 4 fire hydrants correctly then it will assume the 5th is correct as well. It then gives the same kind of test to many different people and if they all get the first 4 correct and pick the 5th as being a hydrant, it can confidently add that to the data set as being a hydrant.
Multiple people answer the same question. If 80% of the people say this red corner is a fire hydrant and 20% say it’s not, the AI is rather confident it is.
You remember that Millionaires quiz show? You could ask the audience which multiple choice answer they think was correct. If a large enough percentage votes for answer C, so should you. People who are right will often agree on a fact, while people who are wrong will give a number of different answers.
It doesn’t always use the selection you make to verify your humanness. It cat actually tell because a bot will either select completely random answers, not make any selections or make selections in such a quick, predictable patter that no human will act like that.
Plus, these usually prevent repeat access from robots so if 1 in 3 have known answers then most robots that can beat the rest of the captcha can only access 3 times before it is stopped.
Latest Answers