# Can an AI be trained to generate the outline of a story?

I know that one of the recent fads right now is to train a neural network to generate screenplays and new episodes of e.g. the Friends or The Simpsons, and that's fine: it's interesting and might be the necessary first steps toward making programs that can actually generate sensible/understandable stories.

In this context, can neural networks be trained specifically to study the structures of stories, or screenplays, and perhaps generate plot points, or steps in the Hero's Journey, etc., effectively writing an outline for a story?

To me, this differs from the many myriad plot-point generators online, although I have to admit the similarities. I'm just curious if the tech or the implementation is even there yet and, if it is, how one might go about doing it.

As far as I am aware, this has not been done yet.

I see several problems with this. A neural network is basically a classifier, which matches an input to an output. Both input and output are usually numerical values, though they could be matched to concepts or words.

To train a NN you provide an appropriately encoded input, and the corresponding output. The NN learns the associations between the two, and can then classify unseen input accordingly. This has recently been used to transform images in a particular style etc.

What would the input and output be for generating screenplays? You could use previous scripts as inputs, but what would the output be? It could be narrative 'moves' of some sort, perhaps. So you could train an NN to recognise narrative elements from screenplays.

However, you are still not creating anything, but just recognising stuff. You would need some other input. I guess you could train an NN on "The Simpsons", get a narrative structure, and then present it with an Episode of "Friends" and see what happens. It won't be a new episode of a screenplay, though.

The other way round might work: you feed it narrative moves (a kind of story skeleton), and get a script out. But it would need a lot of (human) post-editing to be at all useful.

I think an NN is the wrong tool to use here. There has been work done with generating stories and screenplays, even way back in the early days of AI. But that was all based on symbolic AI, not on the kind of ML which seems to currently be en vogue. Have a look at James Ryan's website; he has recently written an overview over historic approaches to story (and screenplay) generation.

• "Hollywood" is notoriously risk-averse, and favors formula, so I suspect generating formulaic content won't be all that much of a challenge. I suspect GANs will be a piece of the puzzle. – DukeZhou Jul 2 at 21:03

Story generation is only possible if some preconditions are fulfilled. That means, it is not possible to train a neural network directly, that it will generate a plot. The easier step before is only to parse existing stories. For doing so, a semantic model is used for storing knowledge about a story. Such models are encoded with ontologies, linked data and in action languages like GOLOG.

On top of a semantic model a concrete story takes place. That means, in the ontology it is defined, that two persons are in the plot, and the concrete story fill the slot with names and attributes. What neural networks are able to do, is to parse these matchings. That means, example stories are mapped to example ontologies and the neural network can predict this decision by learning from example data.

A famous example for automatic story generation in a game is Facade It is not providing a neural network but a semantic model. A neural network can be trained on user interactions with Facade and is able to predict what the user and the plot will do next.

Economics Effecting Question Parsing

It is easy to accidentally misread the question as a practice question rather than a feasability question.

Is it possible for an AI to be trained on literary story/structure to generate them?

$$\Large\ne$$

Did someone train an AI system on literary story/structure to generate them?

Economics Effecting Question Phraseology

It is also easy to confuse wider AI research with the narrower field of machine learning simply because the later is the current focus of economic activity. The question used the term fad, but machine learning is probably going to sustain longer than technology fads.

Is it possible for an AI to be trained on literary story/structure to generate them?

$$\Large\ne$$

Is it possible for AI research to lead to automated literary story structure generation?

Socioeconomic Trends in Authorship Methods

Movie making, including screenwriting, is an art. We know that popular art emerges from new and unusual methods.

• Pollock threw paint from above onto a horizontal canvas.
• WaveNet is being trained to generated symphony music.
• Movie themes with stochastic structures and meanings develop cult following.

The development of sophisticated interrelationships of characters, their feelings, their transitions in belief, ontological questions of individual purpose and how it relates to a another person, a family, a nation, the world, or some principle riding above humanity is not a machine learning problem.

Behind the question asked here, a feasibility question, not an algorithm or convergence question, is the core AI challenge to nature.

Can a computer produce what a human mind can produce?

In thinking about this question, it's clear that the storyteller's training is not an operation that takes a tensor at its input and an expected tensor at its output. The current machine learning boom has not developed any system of intelligent agents that can generate what a literary expert would consider remarkably insightful story. That much is certain.

The trend in academic publications seems to be a strong reassertion of the claim of the MIT AI lab under Minsky, that any feasibility issue would give way to some new methodology or reformulation until all was proven feasible and all was realized in LISP (now in Python or Java wrapping C and controlling some hardware acceleration cluster). Whether this trend is more overoptimism, which we've seen before in AI, or just a matter of time, we will see.

We'll also see of plot-point generators replace screenwriters and eventually the entire studio system, including the generation of stars and parties they go to and the magazines that pseudo-mock their lives to generate star status can be simply simulated. It wouldn't be the first well established and lucrative field of work to be completely eliminated by technology advancements.

It also occurs to many that there may be blowback, either culturally like the return to buckets and monotone after the increase in popular musical sophistication in the 1970s or something more extreme like a mass emergence of Uni-bombers. We'll also have to wait and see about that too.

What seems certain is that research will continue to push the envelope and technology will continue to change even the world of literature and storytelling. New extensions of Alan Turing's Imitation Game will appear: "Can the subjects tell which movies have human screenwriting and which were artificially written?"

"Are those real human stars or are they generated stars playing those generated characters in those generated stories?"

• Much of this answer is sensible and deserves more attention. But what does "Economics Effecting Question ..." mean? Could you explain, maybe some different phrasing would help, since to me reading it it looks like either nonsense or a deliberately obscuring technical term which makes me feel like I don't understand the answer . . . or maybe just a highbrow joke that i don't get? – Neil Slater Nov 14 '18 at 12:10

2018 was remarkable in the creation of the first AI novel, by Ross Goodwin, called 1 the Road. All the raw material was generated by his program.

As far as I know, there isn't any system like you describe yet. However, there are some interesting approaches to narrative intelligence that can be found at the University of New Orleans Narrative Intelligence Lab site: https://nil.cs.uno.edu/

Hopefully those can be helpful in guiding a deep-learning approach to narrative generation problems.