Such is our faith in technology that we tend to overlook the ‘how’ in favor of focusing solely on the result. In the case of artificial intelligence technologies, provided that the results are accurate enough, companies, institutions, and individuals are likely to trust them without caring how they were reached.

Without data, there would be no artificial intelligence. With only a small amount of data, artificial intelligence would be, well, not so intelligent. The truth is that AI systems thrive on data, and the more of it we give them, the better they are able to formulate an accurate understanding of what they are being asked to quantify.

Glenn Gruber, a digital strategist with Anexinet recently stated in an interview that “the more data we put through the machine learning models, the better they get. It’s a virtuous cycle.”

This rule of thumb is no different when it comes to AI-assisted moviemaking systems. The complexity of the tasks that they are being asked to aid filmmakers with requires enormous amounts of data in order to give these systems the perspective they need to provide accurate insights.

In this article, we are going to examine the different types of data sources that allow AI-assisted moviemaking to be a powerful tool in the filmmaker’s arsenal by helping them to understand their audience better.

What Kind of Data Does AI Use?

The term ‘Big data analytics’ first appeared in the mid-2000s and refers to the use of large data pools in predictive analytics, user behavior analytics, and other forms of advanced data analytics methods, rather than for what one might expect, i.e. referring to a specific quantity of data.

Indeed, the quality of the data is more important than sheer quantity. However, as indicated in the introduction, the more relevant data that these systems are able to analyze the better. Generally, speaking, there are two types of data sources – structured and unstructured.

Structured data refers to a type of data that is easy for a system to understand as it is presented in a coherent format that makes it very easy to index and interpret. Until the advent of advanced analytics systems, structured data was the only real data that could be analyzed by computers.

Unstructured data is data that is complex to understand as it involves varying degrees of abstract interpretation or processing to understand. As such, it requires far more sophisticated systems that have much greater computation power.

The difference between the two can easily be illustrated by the simple example of trying to identify the age of any given person. For structured data, a date of birth is sufficient for the system to accomplish this while for unstructured data such as a picture of a face, the system would require much more computation to accurately predict that person’s age.

We will start with examples of structured data in AI-assisted moviemaking systems.

Read Also

Past Movie Data

Data on past movies is the most important structured data set AI-assisted filmmaking systems use. Examples include everything from scripts, information on which actors played which parts and in which genres, actors’ physical information such as height, country of origin, etc., and how much films grossed by region or country.

While there is always the tendency to assume data beyond a certain age might be outdated and therefore irrelevant, this is not the case when it comes to AI systems.

Certainly, during initial training, AI systems will make assumptions that relate to past audience demographics and are therefore not totally accurate. However, with guidance from the data scientists that train them, they will quickly understand why, for example, Charlie Chaplin’s Modern Times (1936) would not be as successful were it to be launched today, but also why it’s performance in 1936 is a useful guide to contemporary releases.

This example highlights the importance of AI systems having access to a wide variety of data sets. Past movie data represents vital more concrete information that the AI system can use to help gain a wider perspective on what movies are about and how audiences responded to them. Given that it is possible to identify the reasons behind past events, AI systems can use these insights as a guide to understand what is happening in the modern era.

Theatre and Live Streaming Audience Data

Both historic, as well as the most up to date viewing habits data, is vital if AI filmmaking systems are to achieve a high level of accuracy regarding gross earnings or audience numbers for any given film.

Indeed, the importance of this data type is becoming increasingly obvious as audience viewing habits shift away from movie theatres to online video on demand streaming sites.

Since this shift effectively ends a business model that stretches back to the very beginning of film, namely, the main release being into cinemas, movie production companies the world over have found themselves plunged into the dark when it comes to trying to understand the financial returns that they can expect to get from their films from online sales.

Thanks to streaming platforms keeping precise records of their users’ viewing habits, we are now able to access accurate data that includes everything from genre tastes, viewing times, whether and when audiences turn off a film before it is finished, etc.

The insights that this data allows AI systems is far beyond what can be collected by movie theatre attendance data and therefore helps AI to dramatically step up the accuracy of its predictions.

Read Also

Unstructured Movie Data

The unstructured data in AI-assisted moviemaking refers to the films themselves. Creating systems that can understand and accurately make predictions relating to what is happening onscreen is the most complex part of movie AI development.

In essence, the unstructured data is any part of the mise-en-scène or staging of the action. This includes visual components such as lighting, actor’s performances, setting, etc., as well as soundtrack elements such as dialogue and music. Other aspects include how the scene is edited, i.e. cutting frequency, as well as the type of shot (closeup, longshot), etc.

Today’s systems are learning how these elements combine to create reactions in the viewer. At the current time, AI-assisted moviemaking systems are able to break scenes down into their genre ‘ingredients’ in a manner that allow filmmakers to get a clear picture of how audiences will see their movies.

As these systems develop, they will get better and better at identifying even the most subtle elements in scenes such as symbolism and subtle emotions as well as understanding how audiences will react to them.

So, for example, not only will AI systems be able to understand why the crucifix scene in William Friedkin’s The Exorcist (1973) is so shocking given the religions cultural symbolism of the cross, but they will be able to also understand how contemporary audiences will react differently to those in 1973, etc.

Such developments are many years off, but only require AI systems to have access to enough data to make them possible. Thanks to our online habits, we are providing oceans of accurate data that will allow AI-assisted moviemaking to take big leaps in a relatively short space of time, leaps that will help filmmakers both understand and connect with their audiences in a way never seen before.