Neural storytelling: how AI is attempting content creation

Deep learning and artificial neural networks now power some of the most advanced predictive and prescriptive big data analytics tools, as well as a new generation of intelligent gizmos such as self-driving cars or more familiar apps like Google Translate, which massively improved the quality of translations after rolling out a new neural net-driven version of the product.

Gradually, machines are becoming increasingly better at understanding word context and different elements in stories. So it’s logical to wonder if AI can become an augmented breed of a storytelling animal?

Three successful experiments in neural storytelling

The early attempts indicate that indeed AI can be taught creativity. Three years ago Ryan Kiros from the University of Toronto published an open-source project on GitHub called a neural storyteller. The neural network was trained on a bunch of romance novels to deliver somewhat tongue-in-cheek descriptions for images:

The algorithm’s artistic abilities may seem modest but it’s important to remember how complex human speech can be for machines. If you have ever conversed with a chatbot, you know that they aren’t particularly great with lengthy queries or memorising (not even generating) a long string of unstructured information (a.k.a. a story) without any direct commands.

Modelling several paragraphs of coherent text with a high-level structure remains an open problem as most natural language processing algorithms can only accurately generate word-by-word summaries. They cannot think ahead and map out a good story plot.

So far, the scientific focus has mainly been on teaching machines how to write a good sentence – a seemingly simple task that not every human can accomplish well enough every time. Researchers from Facebook have recently raised the bar further and decided to experiment with hierarchical storytelling.

The team sourced over 300,000 human written stories from Reddit and fed that data as a ‘summer reading’ to a neural network. The algorithm, after learning what and how others wrote, was tasked with creating a multi-stage story that would be relevant to a particular writing prompt. After applying several different approaches to teaching the network the basics of writing and helping it optimise the output, the team received over one hundred short stories drafted by AI:

You can easily tell that Stephen King didn’t compose that alien story, but would you guess that it was generated by an AI – given nothing more than a short prompt and a bunch of writing about different topics? Most human reviewers did not.

What are the perks of teaching machines to understand unbounded context and come up with quick stories? For starters, knowing how to compose unique responses and even quick storylines translates to better micro and macro sales conversations in the future. AI-powered chat and customer support bots would become capable of holding more effective discussions and intelligently responding to customer queries, no matter how complex they are. Additionally, new algorithms could help marketers deal with such tasks as creating product image captions and product descriptions for images; or better – producing descriptive video content for the visually impaired in a matter of clicks. Creative machines can also help marketers achieve a new level of personalisation, especially for conversational UIs.

Finally, algorithms may not yet excel in writing Booker-worthy narratives, but they can assist us with creating more emotionally-loaded stories, based on data. A recent joint project by McKinsey and MIT Media labexplored how AI can be deployed to identify the key emotional arcs in video stories and prompt the creators when to dial up on those emotions to gain a stronger response from the viewers.

Every successful written or video story incorporates several major emotional arcs – those crucial moments when we cry together with the main hero or feel dominance after watching a James Bond film. Those generated emotions drive further action on our part. With the help of AI, McKinsey further managed to estimate how certain storylines in videos tend to generate more engagement online:

While the developed algorithm cannot create stories of its own, it can zero in on how creators can ramp up their stories, amend dialogues, dial up on plot twists or just add a dramatic tune at a crucial moment to reinforce the emotion on display. Such AI advisory can steer us towards telling better stories that prove certain actions, rather than fly under the viewer’s radar.

AI holds great potential for the content creation industry. By no means should it be viewed as a threat. It's an immense opportunity to bring in more real-life, personalised experiences to digital interactions.

George Karapalidis is head of data science at Vertical Leap.