These are relatively simple programs, which should be easy to stop. More concerning is the usage of computer programs that compose original material. Using learning techniques, researchers have written programs capable of creating original cs papers from a given corpus (ie SCIgen, and other techniques). How long until we see junk books created by these more intelligent (and much harder to detect) algorithms?
Can a robot write books? Yes, and they're absolutely terrible.
11–14 of 14 posts
Re: Can a robot write books? Yes, and they're absolutely terrible.
#12These are relatively simple programs, which should be easy to stop. More concerning is the usage of computer programs that compose original material. Using learning techniques, researchers have written programs capable of creating original cs papers from a given corpus (ie SCIgen, and other techniques). How long until we see junk books created by these more intelligent (and much harder to detect) algorithms?
You know, in this case 'harder to detect' means 'can write reasonably well'. What's the problem with getting reasonably well-written books that happen to have been written by a program? We already get reasonably well-written books that happen to have been written by ghost writers all the time, and nobody claims it's fraudulent that the person named on the cover didn't write the thing.
Re: Can a robot write books? Yes, and they're absolutely terrible.
#13These are relatively simple programs, which should be easy to stop. More concerning is the usage of computer programs that compose original material. Using learning techniques, researchers have written programs capable of creating original cs papers from a given corpus (ie SCIgen, and other techniques). How long until we see junk books created by these more intelligent (and much harder to detect) algorithms?
Re: Can a robot write books? Yes, and they're absolutely terrible.
#14I worked on a natural language processing project as my required directed study in college. We focused on language generation and took an extremely naïve approach to the problem since it turns out that literature on alternative types of language generation (as opposed to simple generation from n -grams) is not well-proliferated. Needless to say, it was very hard. Semantic representation by itself is very difficult to…
Not that your comment is wrong in any manner but I must admit that I cannot understand its relationship with the article.