The A.I. in Evolution is learning by recognising patterns in the words the author uses, and sometimes even strings of words. It then takes words or strings and cross-references them with all of Helden’s works to find the words or combinations that show up the most, and deems them as being ideal because of their frequency of use by the original author. This method doesn’t work however because there are not specific strings of words that should always be used. In the early generations of Evolution, the “poetry” is about what one would expect. There are a bunch of random words spewn across the screen that don’t make any sense. As it grows, the A.I. filters more words out of the ideal list and practically starts from scratch. This can often be seen around when the A.I. has reached its twenty thousandth generation. In every generation, the A.I. will write a word then delete it. It generally grows out of this phase within a hundred thousand generations, and then starts attempting sentences again. This bit of time when the computer seems lost is presumably when it has run through all of Helden’s works once. This has given it a sample of all his work, but it hasn’t been able to cross-reference it yet. After this, there is some slight improvement, but still nothing that can be called poetry. The A.I. will start writing clauses and fractions of sentences, but never anything cohesive. From looking at how it writes, one can tell that it will work through word by word when writing. This means that if two “ideal word strings” have a word in common, the A.I. will often mesh the two together to create something that makes even less sense than the original bunch of words. This is in the attempt to make its own unique sentences, because the A.I. knows that copying and pasting the phrases wouldn’t be original; and while these new sentences truly are original, it’s because they make zero sense.