Can AI Make a Hit Netflix Movie?
The futurist book Life 3.0 by AI researcher Max Tegmark opens up with a story about a benevolent AI taking over the world by making blockbuster movie hits. First, the AI makes thousands of auto-generated movies at low cost, then becomes very profitable, developing a following. Finally, the AI expands into news media, influencing elections, but wielding it’s power in the pursuit of democracy, universal social services, and efficient markets.
While the overall moral of the story is about the power of AI and the impact it could have on society, I found it interesting that the baseline belief required was that an AI could make a great movie. Despite what you believe about technology and power, I’d like to explore if that core belief is true.
As computers and AI get smarter, the remaining frontiers of human capability are creativity, imagination, and soul. Ideally, a great movie should have those human dominated characteristics to be successful, right?
Here I want to explore the concept. Can an AI make a good Netflix movie?
AI in Filmmaking today
In the industry today, media companies are embracing more and more technology to produce, distribute, and enhance their creative pursuits. Good examples include:
- Spotify uses AI to recommend weekly tracks and serve up ideal next songs to you, increasing your time on the app.
- Adobe suite products take the busy work out of a lot of the creative process. Their AI features have simplified and enhanced film making, editing, animation, and design.
- Netflix uses machine learning to pick the images it show you of recommended movies.
- Film companies optimize box office revenue and propose script/character changes based AI models which review the scrip and predict the opening weekend sales with accuracy.
Yes, the tools that humans use to express their creativity in film have digitized, but that does not imply AI creativity. Just because this article was written on a computer does not make it any more robotic, and in fact may have enhanced creative freedom. I was not distracted in the making of this article by needing to refill my quill ink.
So AI is a useful tool for creatives, but can it be creative?
Some developing technologies in AI certainly show progress toward generating some amount of simulated creativity. AI models have been emerging with improved ability to generate characters, scripts, scenery, and pictures. These “generative” technologies, one good example being Generative Adversarial networks (GANS), can be trained on huge amounts of data and produce human-like outputs. For example, the GPT-3 model has trained on writing from around the internet, and when given a writing prompt, can develop some interesting short stories.
Is this technology similar to when you mashed up Wikipedia, Lexus nexus, and encyclopedia Britannica articles in middle school to write your research paper on manatees [Plagiarism / mashup / open sourcing / wisdom of the crowd], or is this no different than standard human creativity which pulls from prior examples? Avatar reads a lot like Dances with Wolves and Disney’s Pocahontas. The Hobbit’s hero journey has the same bones as a lot of mythology.
Aside from my opinion on the derivative nature of human creativity, I do think these technologies show a lot of progress toward generating reasonable characters, scripts, and edits for Hollywood movies.
Filmmaking technologies of the future
Here are some examples of technologies today that could be used to generate the next great Netflix movie.
- AI as an Auto-Muse — As a start, AI can now convert your spoken words into pictures. If you dream of fantastical characters and would like to track them as ideas for your next film, the Dall — e model from Open AI would help. Further, generating worlds and scenery will be possible with the technology behind Microsoft Flight Simulator. In the same way that CGI opened up a lot of possibilities for movie-maker creativity, text to image models will be able to act as a live drawing muse for thoughts of creatives. “Computer, begin generate images of a radish walking a dog!”
- Neural Voice Actor — Given a few dozen recorded tones and words, a neural network can create a pretty convincing facsimile of a famous person’s voice. Using this technology, my Google Home device speaks exclusively like John Legend. Conversely, the neural voices can turn text into believably human speech, which may replace voice actors in the future.
- GPT-3 generated scripts — The large transformer model from Open AI has demonstrated an incredible ability for “creative” writing. Based on a massive, internet-wide training data set including human dialog, chat, other scripts, GPT-3 has the potential to generate interesting and convincing written dialog. For example this screenplay was written by the GPT-3 model. OpenAI’s GPT-3 Now Writing Screenplay For A Short Film With A Plot Twist (analyticsindiamag.com)
[warning the video gets a bit violet, but in defense of GPT-3 it was trained on examples from humans on the internet/reddit/twitter]
The next great Netflix movie
So can a computer generate an award winning flick? Probably not any time soon, but the tools and models are in place for AI to augment the creative process more and more.
Fully computer generated films will likely start as short animated films similar to the Pixar shorts powered by GPT scripts, generated dialog, custom neural network voices, and robust animated worlds. The first iterations of these will be clunky like the GPT-3 script above, but one of the great talents of digital technology is that it can improve iteratively based on consumer feedback. In fact, AI generated technology could even develop into custom films tailored to the user preferences and prior history of likes. After learning the characters you enjoy in Great British Bake off, the model could generate a whole animated season with those characters and improve episodes 2–10 based on your approval of episode 1.
So, while far away from AI creating world changing film, the foundational technology is being built today. AI in filmmaking will certainly be used to enhance human creativity and productivity. In it’s best state, this is the achievable vision of AI as a partner/companion which enhances humanity and brings our human creativity to life. On top of providing access to more people around the world who may not have expensive equipment or training, AI will enable people who may not be employed as creatives to bring their visions to the world. Ideally, the introduction of AI will only increase the number of people making a living off of their creativity.
I would expect that in advance of artificial general intelligence (AGI), there will be AI-based movies. While I don’t expect them soon, the first round will certainly be better than that new Nick Cage Jiu Jitsu movie.
Originally published at https://www.linkedin.com.