Lights, Camera, and AI
“Netflix has a global algorithm which is helpful in the recommendations to all the users and the company claims that the combined effect of personalization and recommendations is worth $1 Billion per year!”
Well, the above statistics are not much surprising as we know what potential the recommendation systems are having these days! More surprising are, the use-cases of Artificial Intelligence (AI)!
AI has paved its way into the film-making industry now!
Recently, a group of researchers presented their research in “The Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing” forming a deep learning model to analyze and predict the movie ratings based on the language data in the movie script!
This model would be helpful in some earlier stages of the creative production phase where it would help with making movie scripts! The model was trained using 980 movie scripts! This would help the movie-makers in identifying the potential impact of the movie on the audience before the production stage, ultimately saving time and money!
Before few months, a popular studio, Warner Bros. signed a deal with Cinelytic, that developed an AI-powered system that predicts the likelihood of a film’s success based on parameters such as actors, budget, and brand!
Well, the system cannot measure the artistic factors of any movie but this system can be used as a tool during the early production phases to formulate and classify ideas that are likely to get success or not!
Warner Bros isn’t the first one to open the gates for AI in this industry. In fact, 20th Century Fox has already been using a system known as “Merlin” for many years! Merlin makes use of Big Data and Machine Learning for matching movies to particular genres and audiences.
It uses computer vision systems for frame-by-frame analysis of movie trailers and suggests that “this trailer resembles movie A or B which were popular in C or D type of the audience”. This prediction would help to evaluate the movie trailer which gives an idea to makers about the promotion and advertising strategy decisions!
Let’s deep dive into the methodology of Merlin:
A) The first step is to scan the movie trailer (ere, Logan) and labels the objects that appear in the trailer such as “forest”, “light”, “beard”, etc, and counts their respective frequencies.
B) Now, the analysis of these objects is done. Say, How long was the “label” shown in the trailer or When the “label” was shown in the trailer, i.e. the temporal sequencing of the data is analyzed
C) Next, the comparison of the analysis done above is performed with other movie trailers to identify similar movies.
And, Merlin predicts what films might be of interest to those moviegoers who saw Logan. Below are the results:
After the movie’s release, the audience was asked about the previously seen by that audience. The below table shows the top 20 actual moviegoer audiences i.e. COMP-ACTUAL compared to the top 20 predicted audiences i.e COMP-PRED.
“We are headed toward a situation where AI is vastly smarter than humans. I think that time frame is less than five years from now. But that doesn’t mean that everything goes to hell in five years. It just means that things get unstable or weird,” — Elon Musk
