Domestic users, who are the majority of ordinary people, poorly perceive the idea of what neural networks are and how they work. Meanwhile, it is no longer a technological exotic.
AI picture and image recognition is already widely used in healthcare, security, aviation, robotics, quality control, and more.
Neural networks are highly trainable. Now they are successfully managing not only visuals, but also copyright. Ultimately, this can lead to a real revolution in marketing, journalism, cinematography …
WHAT CAN NEURAL NETWORKS DO?
It is quite simple to explain to a layman what a neural network is. It is a mathematical model that mimics the work of nerve cells networks in living organisms. Just as in living nature, in artificial nets the main element is neurons, which form various connections and layers of different number, depending on the complexity of the neural network and the tasks it solves.
Huxleў has already talked about a unique experiment on “resurrection” of Taras Schevchenko’s voice using a neural network. This technology is widely used wherever visual recognition is required: various symbols, signatures on documents, faces, bank cards … But the neural network is actively learning human thinking and imagination, in particular, to create stories for people and instead of people.
The authorship of AI is no longer only the stories and scripts, but also the soundtrack for films and even the selection of actors. It is quite possible that in the near future it will not be difficult for a neural network to make a film or series on a turn-key basis — alarming news for directors, scriptwriters, composers and actors. The biggest story is the participation of AI in writing the script for the cult Black Mirror series. Well, a full-fledged literary text still requires human participation, but how long will this situation last? With the creation of advertisement of Lexus, for example, a neural network did quite smart.
FANTASY IS NOT ENDLESS
How is it even possible to recreate the creations of the free and unpredictable human imagination with the help of computer algorithms? The fact is that the “average” amount of what a person can imagine, of course. For the first time, even before the invention of computers, William Wallace Cook tried to algorithmize inspiration in 1928. Cook himself was a popular tabloid writer and eventually summarized his experiences in Plotto: The Master Book of All Plots. It contained about 2000 plot twists, from which, like a Lego set, any story could be assembled.
They say that even Alfred Hitchcock himself kept Cook’s book on his desktop and periodically peeked at it, creating scripts for his film masterpieces. The world-famous work of Vladimir Propp Morphology of the Folktale or The Thirty-Six Dramatic Situations book by the French literary critic Georges Polti can be regarded as such generators of pre-computer plots. Its meaning is clear already from the name itself: no matter how hard you try, showing miracles of ingenuity, everything you have composed will in one way or another go back to 36 basic plots.
THE FUTURE OF THE FILM INDUSTRY
Artificial intelligence today is quite capable of composing descriptions on its own, without human help, according to predefined templates and algorithms. Roughly speaking, you take 36 plots, plus 2000 plot twists and go! Neural networks choose the right words themselves and take previous events into the account. Moreover, it is no longer a person who evaluates the creative abilities of AI, but on the contrary, AI evaluates the quality of human’s work. The ScriptBook company, based on a neural network, has created a program that predicts the success of films at the box office using the information loaded into it about 6500 scenarios.
So, having analyzed the scenarios of films produced by Sony Pictures for 2015-2017, it was able to predict the failure of 22 of them within two years. The ScriptBook program will give you a complete picture of how well your characters are elaborated, how this or that audience will react to them for only $ 5,000. It seems that such diagnostics of each scenario that attracted the attention of the film studio costs decent money, but considering that two dozen failed pictures will cost much more, cooperation with a neural network seems to be very profitable.
NEUROMARKETING WINS BUT DOESN’T CONVINCE
The neural network is gradually turning into a much better specialist in high-quality copyright than its creators. At the same time, it is free from subjective assessments, such as “it’s my vision”. An experiment with the content of a telegram channel, in which marketing specialists participated, showed that in 50% of cases, the copyright of a neural network cannot be distinguished from a text written by a person even by professionals. Moreover, the audience voted for AI reflections, and not digests and complitations, which on a rational level resemble more structure of thinking of machines.
The experiment was carried out using GPT-3, the third generation of the most advanced language neural network invented by Elon Musk and now owned by Microsoft. The neural network is trained on the basis of the content of Internet pages and books with a volume of text data of more than 40 GB. There are projects in which AI independently creates horoscopes based on the texts of Sorokin and Pelevin.
GPT-3-based tools can do more than social media posts writing. They are able to compose quality advertisement and marketing materials for any purpose, product and audience. Often they act as a source of inspiration for the author, prompting him for a non-standard solution, which is then creatively refined. A neural network, in comparison to a living person, in a real communication situation has only one drawback — it is not capable yet of giving quantitative arguments in a dispute, defending its position. But, apparently, this disadvantage is not going to last for a long time…