This article is from the public number: KnowingAI known Chi (ID: Knowing_AI) , author: Yu meow, title Image from: Oriental IC

The cover article will not be written?

The shit can’t pass the article generator to write 10,000 words in minutes, to ensure originality, never re-sample!

Is this generator artificial intelligence?

Although it was written in Python, unfortunately it is simply a looping random content selection script.

The article will start with a celebrity quote, pick up a sentence with no practical meaning, then a sentence, and then a transition sentence that is also meaningless. Then pick up famous celebrities, take sentences, and order sentences… and start again and again.

Set up the statement library, randomly select statements from them in order, and then combine them. The principle is not difficult. What is rare is that the author tries to figure out the ideas and ideas.

If artificial intelligence is introduced, will the effect be better?

It’s hard to say. At present, automatic text generation, that is, the machine writing is usually based on statistical methods. First train the neural network with a large amount of corpus, and then let the machine continuously predict the next words according to the keywords, and finally form the sentences and paragraphs.

The text generated in this way is inevitably subject to logic errors, out of nothing, and the effect is not even as good as the rigid template text generation method.

Can’t you solve it?

Maybe you can ask the authors of the Stars New Literature Award, the author of “The Day of Robot Writing Novels.”

The practice of them is to stipulate the structure of the novel, set the import, progress, end and other large modules, and then disassemble these modules into more detailed beginnings, environment descriptions, etc., step by step and then generate according to specific requirements. Statement, then combine the statements.

The result of this is that the generated novel content looks very reasonable, but it is hard to say that this is the creation of the machine, after all, its framework is entirely from humans.

Another artifact GPT-2 can produce English text of fairly good quality, because the dataset – its training corpus is a high quality response from 8 million Reddit users. In the face of Reddit’s discussion of many topics, it gives pretty good output. When discussing more technical and uncommon content, it will instantly become an artificial mental retardation.

Let the machine write an article, and the road is long.

This article is from the public number:KnowingAI Zhizhi (ID : Knowing_AI) , author: Yu meow