Artificial Intelligence progress in next level.
Creative functions have always been considered the prerogative of man. Since the advent of the first computers, a person has been thinking – is it possible to make him do not purely computational work within the rigid framework of programs, but use it for creative work, such as writing texts, playing chess, editing images and pictures.
We moved from the idea to the implementation, although the features of AI still make themselves felt. Some things, such as playing chess, where the computer is able to calculate millions of possible combinations every move and choose the best one, the artificial intelligence does very well.
On May 11, 1997, the Deep Blue supercomputer defeated world champion Garry Kasparov in a series of 6 matches, and convincingly proved to everyone that he has no equal in the game of chess. Since then, people have been competing with each other, not even trying to beat the supercomputer, more than that over the past years their performance and calculation speed has increased tenfold.
How is creativity
But with purely creative work, such as writing texts, the success of artificial intelligence is not so impressive. So in April 2019, a new author Beta.Writer appeared.
The fact is that there is a scientific content platform SpringerLink, which combines a huge number of various scientific articles written by students and researchers. Reading them all, sorting out and highlighting the most valuable thoughts is a huge job, a large staff is required. Here, scientists from the University of Frankfurt wrote a special self-learning AI algorithm, the same Beta.Writer for collecting and summarizing all articles on one topic into a single book that is convenient for human reading and perception.
The first book dedicated to lithium-ion batteries, which was written by this very Beta.Writer, has also appeared, it’s just boring to read it, there is no single coherence and consistency of the narrative, but in scientific literature, where there is no need to take care of especially original expressions and the creative component of AI, it can applied quite successfully.
However, artificial intelligence is also used in conventional automatic translators, such as Promt, or the built-in browser – Yandex and Google. You can understand what is written on the page with such a translation, but you will not be able to enjoy reading high-quality and non-trivially written text.
The problem is that when writing texts, AI can only rely on plots, forms and styles that are in its database, that is, once created by people. This seriously limits the creative potential of AI and it cannot create anything fundamentally new at the present stage, so writers and a large army of copywriters can sleep peacefully – AI will not take away their bread in the near future, but it is not known what may happen next, the development of AI is underway leaps and bounds, so it is possible someday world literature will be enriched by a masterpiece written by AI.
What’s the deal with text generation?
At the time of writing, the most advanced development is the RuGPT-3 Model from Sberbank. The ruGPT-3 model was trained on 600 gigabytes of Russian texts and deployed using the Christofari supercomputer.
Models are able to continue the program code, texts in Russian and partially in English. To do this, the user needs to formulate a “seed” – a phrase that the model will add.
Painting and Artificial Intelligence (AI)
In 2018, artificial intelligence painted a painting and it was sold for 28 million rubles. At the Christie’s auction, an unusual lot went under the hammer – a picture drawn with the participation of artificial intelligence. Initially, a computer work of art was estimated at only 7-10 thousand dollars.
Conclusion
AI is good at generating abstract objects, but today it is not very good at making meaningful texts.
It is widely used in generating images based on previously obtained examples. There is considerable progress in this matter, but there is still no end to the work, although it is quite possible that in the near future AI will be able to fully generate meaningful texts.
According to experts, it will soon be very difficult to distinguish between the work of AI and humans.
