Cats have known what a computer is for a long time. Why else would they run across their owners’ keyboards? But, thank self-learning software for film and web video: Computers have only known what a cat is for a few days!
Google no longer automatically identifies human faces in photos on the Internet, but also cats. What is revolutionary about this trend is not that Facebook and Co. will be able to tag cats on the internet in the future. It’s the artificial intelligence behind it. Google’s YI HALO 360-degree camera, for example, already uses the so-called Jump algorithm to independently assemble 360 movies from 17 (!) individual cameras into a 3D VR experience.
Independent computer learning is on the rise. It will also have serious consequences for image films and product videos:
Evolution to revolution
What is revolutionary about being able to recognize cats is the fact that computers have taught themselves this ability on their own! While there is still debate about what A.I. (artificial intelligence) can and may be for humanity, there is almost without exception agreement that self-learning software for film will provide innovation on a previously unimaginable scale:
The computer independentlycreates a photo-realistic entertainment experience. This matches the user’s individual digital footprint and thus perfectly reflects their preferences.
Many single images
What works for a single frame is also valid for 25 frames per second if there is enough capacity. It’s only a matter of time before computers will soon be automatically analyzing, cataloging, and automatically editing film and video sequences with AI without human intervention.
It’s old cheese. And this has significantly fewer holes than the dark cave-aged Swiss Emmental: Surveillance software for security purposes can already recognise faces independently. In real time. This is self-learning software today. The step until the self-learning software for film comes on the market is foreseeable.
The most successful machine learning method is called Deep Learning. Deep Learning means that the learning process of the computer takes place in a multilayer neural network (or just: multilayer network).
This brings computer learning closer to the function of the human brain. However, to stay with the example of cats, the sight of a few cats is enough for a toddler to know what a cat is and to recognize a cat. Self-learning software for film is different.
The computer has to look at millions of cat pictures… – but thanks to the internet this is possible without any problems and thanks to the power of modern chips also in reasonable time.
Each time the computer looks at a cat image, the computer calibrates the weighting of each neuron in the layers. The number of layers thus decreases with each recalibration, until in the end only two layers are necessary: Cat yes. Cat no.
The advantage of this approach is that the computer uses the neural process to find out independently and without human intervention how to recognise cats in the best possible way. The computer acts intelligently (that’s why we talk about artificial intelligence). The disadvantage: The way, which led to the recognition of the cats, is no longer recognizable and comprehensible for the experts, who built the computer, as well as for the operator(s) of the computer.
Counting backwards from 5 and rotating once around its own axis is enough… – and our image editing software will tell us what we see! Self-learning software for film is helpful for archiving. And nice for all the movie databases. But not only.
Automated image processing
Apart from experimental works and music videos, the editing of a video follows not only the story to be told, but above all conventions and rules.
You don’t have to be a quantum computer to find the number of factors that have to be taken into account in image processing (or that should be deliberately disregarded) ludicrously small in mathematical terms today. This simplifies the process for the self-learning software for film.
Any search algorithm on the Internet is more complex and requires more brainpower and memory than mapping the rules and conventions of image editing in computer software! For the interpolation of image transitions in montage or for immersive 360 videos, computers already provide automated processes in image processing.
With the fact that software will learn on its own in the future and thus develop itself further, things will only get really fun in post-production. A huge part of human filmmaking is already digitized and available online. In every genre and in every quality. From the cat video on YouTube to the fat Hollywood ham with Tom Cruise.
Machine learning is still avant-garde. Other than Amazon and Google, hardly anyone is doing this at a high level.
Adobe’s Premiere will still ask for the desired genre and a default length at the beginning. A few updates later, the software will suggest what can be created from the footage immediately after the shot material has been imported. The computer will therefore not stop learning.
One breath after that, the editing program will also analyze the learnings from the digital distribution and take them into account in the digital production of the film. After all, the computer knows who clicks away when and where in the film. Self-learning software for film becomes a cinema mogul.
For a small fee, I as a user will get an individual film cut tailored to my digital footprint on the Internet. And which takes into account my own personal, private likes and dislikes.
Paradoxically, every blockbuster will then be different in terms of technical editing, and for this very reason will have the same psychological effect on every consumer.
Virtual acquisition of content
Buster Keaton and Charles Spencer Chaplin already knew: What is not convincing in the script does not become better just because it is recorded on celluloid. This rule of thumb will lose its absolute validity for filmmakers in the future thanks to artificial intelligence.
Today, feature films are already animated in the computer without a single real day of shooting. The editing computer of the future will put 1 and 1 together and generate the perfect material for the perfect film itself. And constantly further optimize. Until the perfect movie.
The future of the film experience increasingly does without actors in the flesh and brings virtual heroes.
NZZaS, 15 May 2016
Disney says that in 5 years, digitally animated images of people will be indistinguishable from real people. I assume that representatives of a US company listed on the stock exchange are rather cautious with such statements. In other words, it will take 3 years instead of 5 years to get there. Self-learning software for film is not a question, it’s just a matter of time.
Self-learning software for film and video
Computers have been able to capture texts since time immemorial. In the brave new world, therefore, instead of raw material, the script is loaded into the computer at the front. In the back, the finished film is delivered to the user via the Internet in the desired resolution. The only question is whether any human being is still writing the script and whether any human being is still watching the film?
The cats namely, they run to there already long ago no more over the keyboard. The pussycats of the future are connected to the Internet thanks to wireless. They look like cats, behave like cats and are as much cats as the seals that are already used in Japanese old people’s homes to sweeten the old age of the inmates. On the iris will conjure up the videos of cats, guessing, self-learning software for film.
Related literature on the subject of self-learning software for film
Superintelligence – Scenarios of a Coming Revolution, by Nick Bostrom, Suhrkamp Verlag, ISBN: 978-3-518-58612-9
This article was automatically translated into English using AI. If you would like to help us improve the quality, we would be happy to hear from you.