Ultimate goal of the development of the AI concept, while an unattainable ideal limited by the current technological level of civilization.
What is artificial intelligence ( AI)? The truth is that the definition itself, as well as examples of AI, depending on the era, were described in different ways.
For example, in the 60s of the XX century, a conventional calculator seemed to be a real embodiment of artificial intelligence. We have no doubts about the progressiveness of the technology at that stage, but today we understand that the calculator and artificial intelligence are not at all the same thing, although its invention was a breakthrough that automates and accelerates the calculation process many times.
Although the 90s were remembered for the relatively widespread use of computers - a new milestone in progress, in 1997 a historic game of chess took place between the famous grandmaster Garry Kasparov and the IBM Deep Blue computer, which won the total games.
Then this victory was cited as an example of the victory of artificial intelligence over man, but today we understand that it was an improved, but still known, principle of searching through a chess tree - sorting through a huge number of combinations for dozens of moves forward and in a split second.
Today, most consumers are daily faced with technologies that use elements of artificial intelligence, for example, voice assistants in smartphones and individual devices, such as Siri, Alexa, Alice.
How AI works
AI in a broad sense is a complex of technologies that allows a machine to imitate the cognitive and creative functions of a person, otherwise - make decisions in a manner similar to a person. There are different views on the classification of AI.
For example, they distinguish “ narrow” AI, which automates decision-making in a specific area, “wide” - working in more diverse fields and various tasks without major changes in methodology, as well as “ general”, which is closest to the capabilities of the human brain.
"General" AI is rather the ultimate goal of the development of the concept, while an unattainable ideal limited by the current technological level of civilization.
The main difference between the modern AI concept and other examples of automation and optimization of decision-making processes ( “scripts” and simple programs) lies in the learning ability and, more importantly, the self-learning system. AI in its modern interpretation is able to understand, classify knowledge, reason and draw conclusions, interact with the user.
In analyzing the concept of AI, one cannot fail to mention such a concept as neural networks. It was the creation of artificial neural networks ( ANNs) that made it possible to bring machine learning to a new level and create a new kind of it, the so-called " deep learning".
The principle of deep learning is based on a simplified simulation of the activity of the human brain. Education through the INS called deep, because the network learns to recognize the information in layers, and the more layers of a network has, the " deeper" is the training and the more accurate the result to which we aspire.
It is the implementation of the principles of artificial neural network in machine learning that has become one of the major breakthroughs in the application of artificial intelligence today. Obviously, an artificial neural network was created on the basis of human knowledge about biological processes in the brain.
By the way, the opportunity to create an artificial neural network mankind owes to the Spanish scientist and physician Santiago Ramon i Cahal, Nobel laureate, who at the turn of the XIX-XX centuries, worked on the study of the human nervous system and became the founder of modern neurobiology.
Of course, an artificial network is not as perfect as a biological one, but it uses the same principles, and its further development, by and large, is only a matter of time.
Learning artificial neural networks requires serious computing power, and until the early 2000s it was insufficient. Today, thanks to the development of computer technologies, in particular graphics accelerators, the processing speeds of large amounts of information have increased many times, but they are still far from the capabilities of the human brain.
This is the reason for the current state of AI: “narrow” is already fully working, and “wide” is just beginning to develop, but is still far from effective use.
However, even in the “ narrow” AI domain, there are many interesting use cases. In medicine, it helps in diagnosing diseases, for example, analyzes the likelihood of developing skin cancer by the image of birthmarks.
Recently, a similar task was tested on the analysis of the patient's retina in order to determine the risk of developing diabetes. It should be noted that AI in this case acts as an assistant for the doctor, but does not replace it. The system cannot and should not make the diagnosis itself, but it can help doctors do it faster based on more information.
It is worth noting that the recognition and analysis of pictures and videos today is one of the most successful types of application of AI. Employees of the IBM laboratory and the Massachusetts Institute of Technology are actively working on training ANNs using large amounts of information, helping artificial intelligence not only recognize images, but also talk about them.
Another scenario for applying applied AI is unmanned vehicles. Such cars using cameras and various sensors in real time recognize the situation on the road or objects: signs, stripes, a person, a dog, an intersection , etc. This is an applied task, which is also quite successfully performed by a neural network.
Application of AI in business
IBM's Applied Artificial Intelligence technology is known under the Watson brand in honor of IBM founder and first president Thomas John Watson. In 2019, according to an IDC report , IBM leads the global AI systems market, with a third consecutive year.
IBM has an analytical division of the Institute for Business Value, which is engaged in research for business, communicates with the heads of client companies from different business fields and from these surveys forms an expert opinion.
According to a recent study by IBM Institute for Business Value, 94% of companies believe that AI is a competitive advantage of the business. At the same time, only 5% of companies today use artificial intelligence. Why is that?
The work indicates that 80% of the data is either not available, or not reliable and not amenable to analysis, and 65% of employees usually do not trust analytics in their company. At the same time, 81% of companies do not understand what data is needed for artificial intelligence.
Some still do not trust artificial intelligence, but changing this attitude is quite simple: to trust technology, you need to understand how it works.
As a rule, today companies cannot analyze what data they need to analyze in order to improve business processes, and also why this or that information can be valuable for AI. After all, the data themselves do not mean much. We need them when we understand what we can get from them, and AI is the best assistant in this sense.
AI today can help businesses make predictions and analytics, automate and optimize different processes, and as a result, make faster quality decisions and be much more effective in the market. Indeed, today one of the most important qualities of any brand is flexibility, the ability to adapt to the needs of customers and respond to external changes.
For example, using AI, a network video service gives personalized recommendations on what to see for yourself based on your views and ratings. Many people like personalized content, and they continue their subscription.
AI also saves and optimizes costs. Take, for example, the issuance of loans at a bank. If these are loans for $ 100 thousand, you can afford an employee in the state to issue such loans. And if we are talking about microloans, it will be very expensive. The AI system can analyze the data that the client enters, and in seconds decide whether to give him credit. This is a good example of an algorithm based on input data analysis, as well as a quick and cheap solution.
How to start implementing AI in business
IBM calls its approach to the introduction of AI in business a “ ladder to artificial intelligence”. Four successive stages of AI implementation can be distinguished:
Make data accessible and understandable.
Any company usually works with several different sources and types of data needed for daily operations. Therefore, first you need to structure your approach to access and data management.
Organize and prepare data.
In the data arrays that the organization collects, some elements may be missing; the data may be incomplete. You cannot use such data as the basis for machine learning. The machine will make mistakes in the analysis of information.
Analyze this data, build a predictive or optimization model, test and adjust it.
Now that the data is of a clear nature and the necessary “ purity”, the results can be trusted.
Consistently introduce AI elements.
Obtained on the basis of correct and “clean” data and tested on the basis of quality models, they will ensure transparency and confidence in the results.
Thus, AI will help a person make informed decisions on the basis of all the information, and in other cases allow the machine to make lightning decisions based on operational analysis.
