I can quite easily using Markov chains some Python and a source text produce quite realistic texts. Turing might even have been pleased in a perverse way his machines are indistinguishable from humans. Many blog commenters seem to be using their second language which is fair enough. But does make it difficult to tell if they are genuine humans. Often the only way I can tell if comments are spam is to look at the links they include. Super cheap *** something is usually a clue.
Just for fun here is a brief auto generated example of text generated from the Henry Fords book ‘My Life and Work’. Markov Ford?
They think that is the business and who do not stand the assumption of mind.
The average person may be filled by foot of fear will respond.
Everything has won his maintenance and also happened to produce also we should prove that bread. If a very carefully tried to greedy for a business could ever really constructive thinking over to share; at a child to the railways and then is the filing of life, manufacture, a tractor came down you see, it must be easy. Success is the country that same with personalities.
To my plant at the limits his territory.
This is pretty crude an still quite easy to tell apart from normal text. With a little more effort I’m sure I could make it much more “human”. I hope most spam is generated by this sort of method I guess the alternative is poor people in foreign countries generating spam for money which is probably worse and doesn’t feature much technology at all so is pretty uninteresting.
Now I wonder if I can produce my own Management Guru book :)
