AI Goes to The Movies

Elias Kruger

What if you could predict the next blockbuster? I am sure directors, producers, and studio executives would want to hire you. Hence it is not surprising the emergence of machine learning startups claiming to predict box office success. Consider Israeli startup Vault that distilled 30 years of movie and box office data into 400,000 features to estimate a movie opening weekend. The company collects these features from upcoming movie productions, feeds into a neural network that outputs a probability of success for the movie. Scriptbook follows a similar path but focuses on screenplay only. Pilot uses a recommendation system similar to the Netflix algorithm to predict both first-weekend gross as total domestic ticket sales. These are just a few of the companies entering the business of movie success prediction.

Yet, a full AI takeover of Hollywood is far from likely. While machine learning algorithms are becoming a common staple in the decision process, most studios still rely on traditional projections based on marketing and past successes. As in any established industry, old habits are hard to die out. With that said, we should see algorithms slowly providing insight and support not just for box office estimates to establishing niche audiences. Just like in other industries, propensity models can be a great tool to help market special-interest movies to narrower audiences. In an age of customization where streaming options further multiply options available to movie watchers, Hollywood studios better get ahead of the curve or else see their major projects disappoint.

There are, of course, limitations in using machine learning for any of these applications. A big one is the algorithm’s reliance on historical data. A true breakout movie would likely not be picked up by machine learning that relies on past success. Even with improved tools and greater insight, making movies will continue to involve risk and uncertainty.

Reality Changing Observations:

Q1. Will AI be able to create movies in the future? If so, would you watch it?

Q2. Is creativity inherently human or can it be emulated by machines?

Q3. Is the age of the blockbuster over or will Hollywood continue to thrive in an age of instant streaming and niche markets?