Cognitive computing - work faster than thought

Cognitive computing is a trend of the last few years. These are technologies that are being developed by many specialists at a very fast pace and help a person cope with a huge flow of information. Moreover, this flow is very deep and wide, figuratively speaking, this is the entire flow of information generated by humanity. The human brain is the most powerful system capable of analyzing unstructured data arrays, processing them and “laying them out on the shelves.” But even this tool does not cope with the information flow of modernity, therefore, a person has put computers, both ordinary personal and super-productive systems, at his service. But there was a problem of a different nature, namely, the need to structure the data that is being processed. Every day, humanity generates about 2.5 quintillion bytes of data, and 80% of them are unstructured. And this means that these 80% are invisible to modern computer systems created using conventional technology.

Come to the aid of cognitive computing, technologies that partially repeat the features of the human brain and are able to work many times more efficiently than their organic forerunner. It is worth mentioning here that we are talking only about a small part of the brain functions responsible for processing and analyzing information coming from outside. You can also talk about self-learning system, with certain assumptions. But, nevertheless, cognitive technologies are capable of much, simplifying life for both a single person and business structures.


')
Such systems can be used in a huge variety of different areas and areas, including banks, materials science, business optimization, urban infrastructure management, environmental assessment, research in various fields of science and medicine. The main task of cognitive technologies is to enable a person to work with unstructured data in a convenient way.
At the same time, systems of a new type are gradually being created, which not only follow a given algorithm, but are able to take into account many third-party factors during operation, self-study, use the results of past calculations and third-party resources (Internet, as an example). The architecture of the new systems will differ from the von Neumann architecture.


John von neumann

As is known, the principles of von Neumann read as follows:

The principle of uniformity of memory . Commands and data are stored in the same memory and are apparently indistinguishable in memory. Recognize them only by use; that is, the same value in a memory cell can be used both as data, and as a command, and as an address, depending only on how it is accessed. This allows the teams to perform the same operations as on the numbers, and, accordingly, opens up a number of possibilities. So, cyclically changing the address part of the command, you can provide access to consecutive elements of the data array. This technique is called the modification of commands and from the standpoint of modern programming is not welcome. More useful is another consequence of the principle of homogeneity, where commands from one program can be obtained as the result of executing another program. This possibility underlies the translation - translation of the program text from a high-level language to the language of a particular computer.

Targeting principle . Structurally, the main memory consists of numbered cells, and any cell is available at any time to the processor. Binary codes of commands and data are divided into units of information, called words, and are stored in memory cells, and the numbers of the corresponding cells are used to access them — addresses.

The principle of program management . All calculations provided by the algorithm for solving the problem should be presented in the form of a program consisting of a sequence of control words - commands. Each of which prescribes a certain operation from a set of operations implemented by a computer. Program commands are stored in successive memory cells of the computer and are executed in a natural sequence, that is, in order of their position in the program. If necessary, with the help of special commands, this sequence can be changed. The decision to change the order of execution of program commands is made either on the basis of an analysis of the results of previous calculations, or unconditionally.

The principle of binary coding . According to this principle, all information, both data and commands, is encoded with binary digits 0 and 1. Each type of information is represented by a binary sequence and has its own format. A sequence of bits in a format that has a specific meaning is called a field. In numerical information, the field of a sign and the field of significant digits are usually distinguished. In the command format, you can select two fields: the operation code field and the address field.

The following are the main elements of a cognitive computing system (Redbook IBM - Rob Hay - "The Epoch of Cognitive Systems"):



In order to comply with its purpose (simplification of a person’s work with his information environment), cognitive systems must be:

• Adaptive . It is necessary to study changes in the information environment, including changing goals and objectives. Also, when analyzing information, it is necessary to consider unpredictable factors. Cognitive systems must be able to process dynamic data and provide the result in real time or close to it.
• Interactive . They must interact with the user so that he, the user, feels comfortable getting the desired result. Also, such systems should be able to work with other systems, devices, cloud services and people.
• Self-taught . The work of cognitive systems should be based not only on new data, but also on the results of their work in the past. They should “remember” previous iterations and access this information if necessary.
• Contextual . They must understand, identify, and highlight contextual elements, such as value, time, location, user profile, goal, process, and task. Be able to access multiple sources of information, including structured and unstructured data, as well as input devices.

Also, cognitive systems can be integrated or use existing information systems (including systems with von Neumann architecture), be able to work with various interfaces and tools.

The scope of cognitive systems is very extensive:

In business, cognitive systems can detect problem areas in infrastructure, the daily routine of an enterprise, and other elements. As a result of the elimination of bottlenecks, the productivity of workers and the efficiency of the work of entire departments increase. Significant funds and time of employees, as well as machine time are saved.

Cognitive technologies can be useful in business in many cases, including the following:

• Involvement: understanding the aspects of each individual, the cognitive system can carry out individual interaction with users, customers.
• Expertise: a cognitive system can be used to audit an enterprise, including a financial issue. As a result, the company provides an individual approach to the client.
• Products and services: the same systems help to constantly improve the services of some companies, showing weak points that could be improved and improved.
• Discoveries: yes, in scientific research conducted by corporations or individual scientists, cognitive technologies can bring many benefits, since the most implicit data can be taken to the surface, studied and recorded.
• Decision making: identifying an increasing number of dependencies in any information, working with this data, the system can improve decision making in a single company or its division.

The applied application of the capabilities of the IBM Watson cognitive system - the study of trends

In health care, cognitive systems help to gradually move towards such a goal as an individual approach to treating a patient. This is especially true in difficult cases, for example, with cancer. Analyzing human DNA and comparing additional factors (place of residence, stress, etc.) helps to treat much more effectively than before. By studying the genotype and characteristics of the body of a particular person, doctors can prescribe the most effective in this or that case medicines and procedures.

In cooking, cognitive systems may offer something completely unexpected, open a new sphere, add new combinations of products. Already, some cognitive systems (for example, Watson) can make recipes for various dishes based on a predetermined list of products.

In sports, cognitive computing helps in real-time to assess the training of various athletes, as well as to recruit teams of players with the characteristics necessary for the coach.

And this is all - only a small fraction of what cognitive systems are capable of. In more detail on this subject we will talk in one of the following articles.

Many organizations and governments of different countries are working on the creation of cognitive systems. But at the moment the most perfect and complete cognitive system, including a huge number of subsystems and elements, is IBM Watson. The IBM blog has already written about this system more than once, and we plan to continue talking about it, because every day the system is being developed and improved. This is an attempt to operate with the categories of the future, not just thinking, but thinking ahead of the curve.

Source: IBM Materials.