Artificial intelligence “conquers” the IT world : Corporations buy up competitors developing neural networks
Google has bought Belarusian company AIMATTER who developed the neural networks for smartphones. Last year the Belarusians launched a self-learning app that can change the color of hair in photos. This was enough to attract the attention of the American corporation.
Such transactions occur more often because the rapid development of computer technology — especially in terms of computing power — expands our capabilities to an unprecedented level.
The pace of application of cognitive services in a variety of areas is doubled annually. But the question arises — do we all understand what the definition “artificial intelligence” means? Soon it will be literally everywhere thanks to the introduction of “smart things” and “smart services” but media and cinema are modelling the public opinion of a very contradictory image rather dangerous than useful technology. So is AI an enemy or a friend?
On this topic there are incessant discussions. All news feeds are full of reports about new achievements of neural networks, machine learning algorithms etc. Based on the information background of recent months it was in the neural network segment a new breakthrough occurred and without surprising — Unsupervised machine learning systems are able to work without a person or even instead of it.
For artificial neural networks to “live their own lives” they must learn to feel, abstractly think based on independent sensory observations of the environment and to have intuition and imagination.
With all these tasks AI can not cope yet and is unlikely to be able in the foreseeable future therefore it only imitates the intellect — it is incapable of realizing itself. So it is absurd the world would be ruled by such an “intellect”.
But it could be enough experienced dealing with a huge array of varied applied problems. Especially if it is connected with complex calculations not adapted for human brain. Designer can change the color of hair on the photo in such a qualitative and rapid manner so that the skin can not be also tinted.
Neural networks are the result of unsupervised learning: the computer receives a task, calculates a function and tries to find out the solutions through analysing examples. First ше colored the entire head, then large fragments, then it grew smaller and now AI learned to color separate hairs. According to the developers at the same time their neural network is “not smarter than an ant”.
To teach AI to recognize objects or hair the same a sufficient quantity of marked images are loaded into the system and on the basis of analysis the neural network finds common lineament, regularities and other features of these images.
There will be errors but AI will learn and eventually can associate the pictures with specific characteristics. This method is similar to the game form of teaching little children when they are explained distinguishes between cats and dogs.
The difference is that an artificial neural network learns from a million examples at once tunes its invisible neural connections until it reaches capabilities of a state where the problem will be solved with the utmost precision: 99% of 100%. Why not 100%? Because the neural network operates with probabilities but not with absolute values. For this “1” is “99 probabilities out of 100, that is 1” and that is the result of training.
Over time neural networks will become more perfect. The goal of scientific researches is to teach AI to fully simulate the work of the human brain copying all the behavioral factors of the biological analogue in synthesis with inaccessible people.
If AI is combined with the blockchain technology in decentralized system format it will be possible to create a huge supercomputer in the form of the one global neural network comprising local neural networks — in the same smartphones.
The possibilities of such AI + blockchAIn look almost limitless. It’s no wonder that transnational corporations are literally hunted behind AI developers — if such a decentralized project is implemented then their monopoly will be dealt a powerful blow.
Prepared by team NeuroSeed