Artificial Intelligence DQN and WCS.

25.01.18 в 09:54 other 2611

In the recent past by the standards of the development of information technologies, in 2015 Google created artificial intelligence based on neural networks, which was able to analyze the condition around itself and draw conclusions about its further education. The name of the new offspring from Google was given in abbreviated from the term "deep Q-network" - DQN.

The DQN started training in common arcade games (Pakman, Tennis, Space Invaiders, Boksing and other classics). The results were amazing: DQN surpassed in 22 of 49 games the successes of the best players in the world!

"He plays very carefully, as if a human behaved after a number of defeats, but a person is not able to react so quickly in a calm condition" - !!!!!

The program was taught in practice. Rather, she did it herself, because she was not given any rules. Trial and error method. And it's not surprising that she was able to get comfortable - the processing speed of graphic information is 2 million pixels, which is thousands of times faster than a human could do.

"This cool program learned about the trick with punching a hole in the ceiling of the blocks and throwing a ball there, but this is a profi Breakout trick!" - !!!

Unfortunately, the history of DQN development is over, and the specific reasons for such a rapid cessation are not called anywhere. But the development of neural networks continued.

Let's look in 2017.

One of the offices of IBM has created "Watson for Cyber Security" - a management system of cognitive information security centers. She was able to learn natural languages and analyze cyber threats. That is, security analysts provide reports to the program in the language of ordinary people - the program reads them and conducts an "investigation", distinguishing false threats from real ones. And this decision was very profitable for IBM, because at the time of 2017, IBM calculated how many man-hours a year it took to process the false threats - the result was about 20 thousand hours. Very impressive, isn't it?

By the way, the program was related to cognitive defense systems, it meant ... a possible potential for overlap with a human. If you know about the famous cyberpunk games, then you understand what it's all about. The system of protection, communicating in the human language, exploring cyber threats (hacking, tracking, interception of control, etc.) is good. But who said that she can not work "in the opposite direction" - to conduct cyber attacks? It is enough just to create interfaces for the interaction of a human neural network and a processor with this kind of program. For example, as in "Syndicate", "Deus Ex" and other scientific "fantasies" - any cyberattack is the cost of one idea.

However, the reality is not so complicated and scary – artificial intelligence is still far from perfect. At least, there are still a couple of years prior to full development of artificial "clever" if you believe Gartner. And judging by the degree of utilization of neural networks (according to data from Spiceworks, 30% of companies employing more than 1,000 employees, are using AI in their IT activities), forecast, Gutner can come true. And most importantly – the using of such programs is very justified! They are like nootropics to your budget, because ahead of the workers. People need to pay salaries, to equip the premises according to the working conditions for full-time work to hire 2-3 people, and most importantly – the risk of error in the case of large scale processed digital information. For example – artificial intelligence system in the company Daqri is watching more than 1 thousand devices internal environment. It is able to track almost any breach of security or locked for suspicious activity(shipment of large amounts of data, a port scan on the subject of "loopholes", etc.) that nearly eliminates the possibility of corporate espionage.

But all this is only the beginning, the beginnings of the development of artificial intelligence without limits. Now we have artificial intelligence technology for advanced quick antivirus-firewall, which does not need man-made database, and then not always, and people still needed to ensure cyber security, the full application of neural networks and solving problems of artificial intelligence.




伴随着会联网的广泛使用,随之带来一系列的危险。首先涉及到匿名和安全性。如果信息不设防范手段,您的数据很可能被入侵 — 这里举个例子,关于在公用网络中的私人的信件。

最初的互联网预想让空间没有国界化,无论在在那里都可以绝对匿名地获取任何信息。现在很多国家都出现了禁止访问某些资源的情况 。

12月 2017


In the recent past by the standards of the development...

Recently, VKontakte, the most popular social network in Russia, has...


​The most common methods for organizing network anonymity are the...

You probably know the video-tuning service specializing mainly on...

Thousands of cybercrimes are committed daily in the world, large...

Good afternoon! Now we are talking about such an important...

There are many types of advertising promotion of their services....