Operating systems and hardware technologies seem to have come to some degree now. We are also experiencing easier-to-use operating systems at the same time as the speed of our computers increases. However, the fact that faster and more "easy" use is now a criterion, does not mean that it will be the same for the next decades. This is only possible with new artificial intelligence exercises if an operating system that we have in our mind in the traditional sense and the hardware technology that develops in this way will be able to proceed after that. Artificial intelligence studies seem to be the most important research field and driving force of our day and near future. It is inevitable that there will be a transition from the operating system to the intelligent operating system if there is a switch from the phone to the smartphone. In other words, at the end of the next decade, it is highly probable that the concept we refer to as operating systems will turn into a "living" artificial intelligence.
Big companies like Google and Facebook have the most expertise in artificial intelligence in the world. As such, artificial intelligence becomes more involved in the life of the end user every day. For example, Facebook is organized with an artificial intelligence software that keeps users up-to-date with updates on their news sources and constantly learns about personalized ad submissions. "Deep learning", a method inspired by the way neural networks work in artificial intelligence, is one of the biggest breakthroughs of recent years. Today, Google faces the challenge of being the most advanced company in this technology. Although there are relatively few experts in deeper learning, most of them are working under Google. Especially since bought DeepMind, a London-based company, it seems that all the trumps have gone to Google. In this regard, Google has started to share its work with the world, partly by doing it right. In addition to training for the training of new specialists, he has been using open source as part of the TensorFlow artificial intelligence software library he uses. Facebook has also shared some of the code for artificial intelligence studies (DeepMask, SharpMask, MultiPathNet) in the same way.
Such "generous" behaviors of these big companies can be explained as follows: Firstly, the world of information and even the makers of it can now see that everyone is profitable if their knowledge is shared. We know that it is not easy for commercial companies to come to this confusion. The next generation of IT specialists can be a source of change, with open source, free software and systems based on unrelated labor partnerships such as Linux. It is also a fact that keeping the threads in one hand is not going to accelerate in order to be able to make progress in the tricky situations like artificial intelligence. Big firms want to focus their computing world on the same goal at the global level by pointing to the direction they should go because they will be the ones who will benefit most from it.
As soon as DeepMind worked on the company's deep learning method, surprising results emerged. The beginning of the story took place in simple arcade games, where artificial intelligence achieved little success and succeeded in games. Similar to the way babies learn by playing, the software has developed its own experiences and learning style. Unlike the "Deep Blue" software developed by the IBM company that defeated former world chess champion Garry Kasparov, DeepMind is not configured for a single game. He has the power to operate his deep learning ability in many different areas. It is no coincidence that DeepMind's founder, Demis Hassabis, also showed a superior development in childhood chess. At the age of 13, he was the second highest ranked chess player in the world under the age of 14 (Judit Polgar, the youngest of the legendary Polgar family) was the first.
After the chess, there was only one castle that man could sink, GO game (also known as Baduk). Until today, the most advanced GO game software had difficulty in defeating even a man of average skill. Since the Go game had complex possibilities at the time of its inability to calculate, it remained a game in which software that used old methods such as Deep Blue could not provide superiority. Being able to make instinctive decisions based on experience is one of the key elements in the Go game. With the software AlphaGo developed by the Google DeepMind team, the final fort was conquered in 2016. AlphaGo defeated world champion Lee Sedol 4-1 and gave the first signal of what the deep learning method could lead to in the future.
This new kind of artificial intelligence has even begun to function quickly in different areas. It will surely provide many benefits in medicine. There are studies in the UK about early detection of eye disorders by artificial intelligence. It is not hard to predict the contribution of artificial intelligence to the 5000-minute television archive on lip reading, which would allow the reader to read the lip as well as a professional person in this regard, which would bring security and "surveillance" systems. Another important development is the language luck.
In the near future, the deep learning method has been incorporated into the Google Translation system and in a short period of time, very striking results have been achieved. Especially for languages other than English, there has not been a long time worthwhile journey, but we can say that the new kind of artificial intelligence has stepped up a lot in this regard. In the deep learning method, software is not only able to produce results from other people's experiences but also from others. Knowing this opportunity, Google has added features that will enable users to contribute to the development of artificial intelligence on the Google Translation page. After you have spoken the languages you know, you can interpret the translations of the software as true / false between those languages and make it easy to understand the nature of the language of the software by translating the examples given to you. If you want the artificial intelligence to make better translations into English, you have the ball. In addition, the text that Google Translation has obtained from the internet for English language analysis can also be a fun way to spend time. Seeing quoted text and phrases, the system can not stop thinking about what services Google is using.
If the Google translation system makes a jump, it is still possible to find some errors. At first glance, especially when trying to translate long sentences, there is a difficulty in finding out where the cues differ from each other in meaning. It is possible for mistakes to arise while translating such cues as it can have different meanings depending on the structure of the cümlenen in which it is contained and its context. On the other side, it will be surprising if the artificial intelligence in the very near future witnesses itself to such problems. Artificial intelligence also makes a big difference in terms of translating native languages into spoken languages other than spelling and text reading systems (see: WaveNet).
Even the experts who created it are surprised by the way Google's translation works. Artificial intelligence succeeded in making healthy translations even between the two languages for which the translation method was not defined. For example, since there is a translation method and knowledge between English-Korean and English-Japanese, it is natural that the software translates between these languages. Even though there is no such translation method between Korean and Japanese, when the software requested such a translation (although it was prevented from being used as a bridge in English), the experts entered a small circle of scarcely encountered results. It was suggested that the software could do such a translation (without using English), but that it could be possible by translating the common features found in the languages the software has so far studied into its own language. This comment is short; It implies that the software is a kind of abstraction activity on its own. If this assessment is true, it means that one of the most complex capabilities of the human mind is artificially imitated for the first time. It may be a matter of time when a machine capable of doing this could cross the human mind in almost every field.
The next step in an artificial intelligence that dominates natural languages will be computer languages or other formal languages. The point reached by artificial intelligence about formal languages has not gone beyond the level of "reading"; We have not yet created a "writable" artificial intelligence. We can, however, hope that a natural intelligence-dominated artificial intelligence will develop a faster insight and skill in relatively less abstract, but more regular computer languages or symbolic logic systems, if we go from the definition of 'formal language is a subset of natural language.' This means in a sense that; Artificial intelligence in the near future will not be able to read the language in which it was created but also write it with it. It is not hard to imagine that this ability will give him infinite possibilities in the direction of his own decisions, or in accordance with the task he is able to carry out, from scratch, the necessary coding. It is also possible to develop a unique software language that we do not know.
For the first time in history, humanity is experiencing a phase in which it has to realize how it works. Experts are not fully able to explain how artificial intelligence's algorithm and hardware structure, despite human manipulation, has achieved the results of software because of the nature of the deep learning method. In that face, all this noise is stunning, but at the same time you can hear footsteps of a scary future.Everything we write right now can be a reference to our "talented Mr. Machine" that will emerge in the near future, or evidence against us. Then we can speak to him already:
Wellcome Samaritan.. or the Machine?
Comments
Post a Comment