The topic of artificial intelligence in 2017 has become one of the most attractive for discussion. There were so many commentators among the IT market participants, and the comments were so interesting and detailed, that in the final issue of CRN / RE for 2017, we were not able to discuss all of the issues proposed for discussion. Today we will talk about the pros and cons of AI solutions and the difficulties of its implementation.
What are the main advantages of solutions that are called today "artificial intelligence systems"?
Project Manager ST Smartmerch, System Technologies Group, Maxim Archipenkov I am sure that "pluses follow from expectations".
“Neural networks, unlike humans, do not have emotions and do not get tired,” says Arkhipenkov. - The human factor and all the errors and problems associated with the character of a person and his low working capacity are excluded - regarding the machine, of course. Neural networks do not have a performance threshold: if a person can check, for example, 100 parts for quality in a day, then the system will check them as many as the server capacities allow. The system is easier to scale: at the same plant, 100 people for quality control in one room are difficult to put.
Marketing Director CDNvideo Angelina Reshina also believes that the main advantages of AI systems "in the speed of data processing, the ability to train the system and save on human resources."
Cezurity CEO Alexey Chaley emphasizes that AI-based products are capable of performing tasks on a qualitatively different level: classifying images, translating text, classifying files, etc. ", Chaley notes.
“The main advantages of currently existing solutions are the ability to automate many areas of activity while minimizing human participation in this and expanding areas where it is possible to use software instead of human labor,” says the founder of the hosting company King Servers Vladimir Fomenko. - At the moment, AI is especially good at analyzing large amounts of data, where a person would take too much time, and conventional programs that do not use machine learning would not be able to achieve the necessary accuracy.”
I agree with colleagues and the director of the department of corporate information systems ALP Group Svetlana Gatsakova:“With the help of AI technologies, the speed and level of automation of processing large amounts of information is significantly increasing - while improving the quality and manufacturability. With the right attitude to new technologies, the completeness of data use increases, as well as the efficiency and quality of management decisions.
According to the CEO of Hawk House Integration Alexandra Ivleva,"AI technology is best suited for optimizing various kinds of mechanical activities, automating routine operations, and using it in hazardous industries." “Proper use of robotics on conveyor lines allows switching to a non-stop operation, optimizes enterprise costs, improves product quality, but requires a serious and lengthy commissioning stage,” says Ivlev. - Not many companies can afford to invest large amounts of money in such technologies, although in the future this will make it possible to significantly reduce the cost of production. The situation is similar with machine learning technologies: for each project, analyze a large sample of data, moreover, using individual algorithms, which requires time and resources. But after the introduction of automation, these operations will occur faster and cheaper than a person can do.”
“Let’s start with the fact that artificial intelligence systems are being developed to improve efficiency in the broadest sense of the word,” recalls the Director of Business Applications at CROC Maxim Andreev. - To implement new ideas and approaches, companies often need to take into account a huge number of factors that an ordinary person simply cannot keep in mind. One of the main advantages of artificial intelligence is the ability to take into account such a diverse number of factors in real time. In addition, unlike a human, an algorithm cannot get tired or change some information on purpose. That is, by introducing artificial intelligence, the company minimizes the possibility of errors caused by these factors. But there is a downside to this: a person can take into account additional details, while a poorly tuned algorithm will continue to work incorrectly. Another plus of artificial intelligence systems is replicability. Take as an example any business process in a company that takes an employee a year to train. Therefore, if we need 10 new employees, then we will spend 10 man-years training them. From the point of view of algorithms, everything is simpler and the cost of scaling the solution is much lower.”
Head of Development and Implementation of AV Solutions at Auvix Alexander Pivovarov believes that the most obvious and superficial pluses include increased efficiency, reduced routine operations and greater ease of use. “For example, if you take such a fairly simple thing as a system for booking and displaying the schedule of meeting rooms, then when you start to study it carefully, you see many opportunities to increase the efficiency of its use, reduce downtime, and so on using “smart algorithms,” emphasizes Pivovarov.
“The main task of digital transformation, one of the tools of which is AI, is to make processes run faster and more efficiently, companies spend less and earn more,” says ABBYY Russia CEO Dmitry Shushkin. - For example, one of our customers in the banking sector automated the processing of documents for opening an account for legal entities. The intelligent system itself types and recognizes documents, then extracts information from them and loads it into the required fields of the banking system. As a result, it takes less than 10 minutes to enter data from documents, 2.5 times faster than manually. The bank calculated that in 3 years it would save more than 270 million rubles on document processing.
According to Plantronics business development manager Alexey Bogachev,“One of the main advantages of AI systems is the ability to get some new materials that are simply not available to us. Since an ordinary person draws conclusions based only on his knowledge, but here we get a deeper analysis that can lead to completely unexpected conclusions. This way you can get a breakthrough in a certain area.”
“Man is accustomed to consider himself the crown of evolution, but we regularly face limitations,” says the CEO of FreshDoc.ru Document Constructor. Nikolai Patskov. - For example, hypersonic aircraft fly at a speed 10 times greater than the speed of sound, a human pilot is simply not able to control such a machine without the help of smart electronics. Human reaction and decision-making speed are not enough to work at such speeds. Artificial intelligence helps us to step over these limitations. AI allows people to react faster, protects against making mistakes, frees them from routine operations and decisions. Such systems can effectively replace a human expert in transportation, forecasting, trading on the stock exchange, consulting, and drafting documents. The use of "smart solutions" also affects the final cost of the product: after all, "robots" do not need to pay salaries, they do not get sick and do not go on vacation, and are not subject to a decrease in efficiency. We see huge potential in the development of intelligent solutions for a wide range of tasks. Participation in the development of this area may allow Russian IT entrepreneurs to turn the market and “ride” the information wave of human development.”
According to the director of business development and marketing of Konica Minolta Business Solutions Russia Zhamilya Kameneva everything, of course, depends on the class of solutions. But for the most part, they are aimed at optimizing and automating processes, saving resources - both tangible and intangible, working and personal time. “To put it simply, their task is to make our life easier,” Kameneva sums up.
“Firstly, such systems allow us to reveal what is hidden from the human mind,” says Navicon International Business Development Director Ilya Naroditsky. - No matter how good a person’s BI tools are, in some cases machine learning is indispensable: for example, if you need to process the statistics of operations on bank accounts of 1 million customers over 10 years. Already today, the machine search for hidden patterns that are not obvious to a person allows many companies to build a business strategy and create management decision support systems. Secondly, artificial intelligence technologies significantly increase the efficiency of all types of communications with consumers. Innovative technologies that can understand and analyze text and voice messages help reduce the processing time of incoming requests and respond to customer requests more quickly than before. Thirdly, such systems can relieve company employees from performing routine operations, which means freeing up their time for solving strategically important issues. The time spent on solving routine tasks could be used to solve creative problems.”
“Such systems make it possible to make decisions for a person in those areas where it is permissible,” says Atak Killer CEO. Rustem Khairetdinov. “While previously automated systems made decisions only within the framework of well-defined “if-then” scenarios, then today’s and tomorrow’s systems will be able to make decisions under vaguely defined conditions and with insufficient information, which previously could only be done by a person.”
Acronis Development Director Sergei Ulasen also notes: artificial intelligence systems solve many tasks that previously required the involvement of a person. At the same time, they often function faster and have a predictable result and quality of work.
“AI technologies really work and help improve business processes, at least partially freeing the human intellect from the routine for creativity and creating something new,” emphasizes the CEO of Preferentum (IT Group) Dmitry Romanov.- It is easy for them to assess the economic effect. For a large class of systems using machine learning methods, their ability to become “smarter” as they work is a definite plus.
According to the marketing director of Vocord company Sergei Shcherbina, the main advantages are that based on "chaotic" facts, poorly structured, unclassified or incomplete information, AI makes accurate predictions. “Relying on them, we get a fundamentally new level of accuracy and speed of decision-making where simple, linear rules do not work,” Shcherbina continues. - Huge arrays of data are constantly replenished, but by themselves they cannot solve problems, AI is just what is needed to analyze them. We already know many examples of the successful application of AI in medicine, in the analysis of global and local economic and social processes, in solving engineering and technical problems, in making investment decisions, and in security systems. Innovations in the field of AI will make it possible to automate a fundamentally wider range of business processes. Thus, in the field of video surveillance and security, for the first time it will be possible to guarantee, without the participation of an operator, to detect potentially dangerous incidents 24/7, to identify wanted persons. There are already many examples of successful applications of AI.”
The main plus, according to the co-founder of the shikari.do service Vadim Shemarova, is that AI systems are trainable. “For example, if we want the system to be able to distinguish people's messages where they want to buy something from messages where they want to sell something, or determine the subject of messages, we do not need to compile a detailed list of words and phrases that express intentions. , mood, theme, etc. We select a lot of example texts on the topics we need, “train” the system on these examples, and then it itself begins to understand the essence of unfamiliar texts,” Shemarov says.
Head of the Research Center for Robotics and AI Regulation, Senior Associate at Dentons Andrey Neznamov also believes that the possibility of learning (supervised learning or self-improvement) can be called the main plus of technologies that are commonly called "AI".
What are the difficulties in implementing these systems?
Briefly summarized, the main advantages of AI technologies, according to IT market experts, are reaching new levels of productivity, automation, efficiency, analysis, learning, decision making, predictability, and learning. However, since this is a new direction, experts see even more difficulties than advantages. Suffice it to say that almost every speaker named his difficulty.
“This is a completely new area. Each task that is being solved now is RnD in its purest form: you need to define, systematize, come up with a solution, implement this solution and implement it, - emphasizes Maxim Arkhipenkov. “This is a creative process that requires a high level of science and expertise both directly in the field of application of this solution - whether it be FMCG, space, medicine, or in the field of implementation of neural network systems.”
According to Alexander Pivovarov, the difficulty is "in finding a balance between hype and real utility, the difficulty of making these technologies invisible to the consumer and the absence of errors in their work."
Dmitry Karbasov believes that "the key difficulty of these projects is related to the unpredictability of the result." “Let's say, when buying a CRM system, the customer clearly understands the functionality that the system offers him, and how he will use this functionality,” says Karbasov. - These are processes, data entry forms, reporting, etc. When implementing an AI system, it is very difficult to predict the result without implementing the project, the disclosure of technologies and algorithms will tell practically nothing to a person without a mathematical education and practical experience, and among customers there are only a few top managers with such a background . The implementation of pilot projects helps, the methodology of which we have debugged and which we use in 99% of projects.”
“There are certainly a lot of difficulties,” Maxim Andreev reflects. - The main one, perhaps, is the lack of large enough data sets for training artificial intelligence. This requires historical data. Let me explain what I mean: for one large company, we made a sales forecast for transportation services - we predicted the weight of cargo and the direction of transportation. We could not achieve good forecast accuracy in any way, we began to figure out what was the matter and found out that in the historical data that was stored in the company, somewhere the weight was taken into account with packaging, and somewhere without. At the same time, there is simply no sign by which this factor could be tracked. That is, once in the past this information did not play a role, but now everything has changed. That is why it is so important to collect all the data that can be collected “on demand”. Technologies for collecting and processing data are constantly evolving, and companies can already implement Data Lake technologies, which are becoming an excellent platform for training artificial intelligence. Another difficulty is that the algorithms themselves are still quite small. Therefore, before the introduction of the company, it is necessary to conduct research. This allows us to find out whether, under specific conditions, on specific data and for specific business processes, it will be possible to build AI, the costs of which would not exceed the value that it gives to the company.”
Anna Plemyashova believes that the main problem is the complete absence or insufficiency of data to build accurate models. “For industrial enterprises, where such solutions require significant investments in infrastructure, this is a delayed economic effect: you must first start collecting and accumulating data, and then you can move on to solutions using intelligent systems. Transitional BI solutions and real-time data visualization allow bringing economic benefits closer, says Plemyashova. - Another difficulty is the need to restructure the business process when introducing intelligent systems. That is, it is not enough to buy such a solution and put it like a flower in a vase or an application on a computer. It is necessary to make this decision friendly to the business process: create, reconfigure or even cancel some operations, retrain people, optimize staff.”
“These systems are based on data and big data,” reminds Sergey Ulasen. - Significant computing resources are needed to train models, and an appropriate infrastructure is required to store big data. Therefore, the implementation of AI systems requires a significant investment in hardware.
In turn, the collection and preparation of data require great organizational efforts, and often the development of new software that helps in data analysis.”
Svetlana Gatsakova believes that the difficulties are primarily "in insufficient attention to the limits of applicability of each specific AI technology, to pitfalls." And also "in the weak interpretability of the results (after all, for example, a neural network does not explain its conclusions), in the difficulties of forming homogeneous data sets for training and testing models." Another difficulty is "blind faith in data and a lack of attention to manager's intuition and those factors that are difficult to measure and integrate into DDM* processes." On this, according to Gatsakova, "complexities specific to Russian organizations" are superimposed. “This is the low availability of reliable data about the external world of the organization and the resulting risk of becoming isolated on internal information, that is, turning into a kind of autistic organization. In addition, this is a small (compared to leading Western companies) penetration of the DDM culture, limited mainly to graduates of Western business schools.
“AI helps to automate many processes and replace low-skilled employees, but at the same time it requires control from developers, whose work costs, of course, are higher,” says Angelina Reshina. “The system’s learning needs to be controlled so that it does not go beyond the acceptable limits.”
According to Sergey Shcherbina, the difficulties lie in outdated equipment and weak infrastructure, legacy hardware and software platforms, which in difficult economic times and with limited budgets, few people dare to change. “The human factor also has an impact,” Shcherbina emphasizes. - Here there is a shortage of qualified personnel, and an insufficient level of competence, or the conservatism of leaders. Moreover, not everyone understands why this is necessary and why spend money on modernization, when “the old fashioned way” everything seems to work anyway.”
“Among the difficulties in building AI systems, first of all, it is necessary to note the shortage of personnel,” notes Andrey Sykulev. - There are very few specialists, because the requirements here are extremely high: in addition to programming skills, one must master a rather complex mathematical apparatus and have knowledge and experience in subject areas. Quite often, the “showstopper” is the poor quality of data and the lack of infrastructure for their integration. Another important issue is data security, because data consolidated for AI operation can become a target for attack or be used, to put it mildly, for other purposes.”
Alexei Bogachev also believes that one of the main difficulties is personnel. “As with everything new, the question is how to work with it. Since the applied application of any technology requires qualified specialists, and this is a very young direction, it is therefore quite difficult to find people who would understand this.”
The staffing problem has another side. “The main difficulty is that not many top managers of enterprises understand what artificial intelligence is and what its practical application is,” recalls Dmitry Karbasov. - Yes, almost all of them have heard about AI, everyone knows that AI helps to optimize business processes, reduce costs, make certain functions more efficient (logistics, analysis of consumer behavior, forecasting production load and sales volumes, etc.). But few of the customers understand that in order for AI to work as it should, it is necessary to formulate a business task and criteria for its success in business terms. In other words, the customer must understand which of the parameters should be instructed to analyze the AI system and how to deal with the received data from the point of view of making managerial decisions.”
“Two factors can be singled out as the main difficulty in implementing such solutions: human and technological,” says Nikolai Patskov. - The first is the problem of a small number of experts able to interact with artificial intelligence systems. This problem is gradually being solved, the market realizes the value of such specialists and more and more employees master the skills necessary for the developing market. The technological factor can be attributed to the lack of computing power: now we are again developing ideas that we will be able to implement only with the advent of more powerful machines. But given the projected growth in productivity (an increase of 1000 times in the next 10 years), we believe that the evolutionary development of technologies will at least not slow down.
According to Aleksey Chaley, there are three main difficulties: “The first is people . There are very few people in the world who are able to work in frontier areas, who at the same time understand the subject area (in our case, virus analysis), are well versed in mathematics, statistics and machine learning, and also know how to program at least a little. The second is data for training . This data must be taken somewhere, and then marked up. Data is very difficult to obtain. Because of this, by the way, the progress of AI development is hindered, since researchers do not have the opportunity to experiment with models. It is not enough to just be a talented analyst and programmer - without data, it is impossible to create anything in the field of AI. And the third is the cost of infrastructure. The initial investment in infrastructure can be quite significant.”
“In order for artificial intelligence to solve business problems well, the technology must be “customized”, Dmitry Shushkin believes. - Any machine, like a person, needs to be trained on actual data in order to make accurate decisions. To teach such a system, one first needs to collect or synthesize a large amount of well-labeled data - for example, information about finance, production, customer service, and so on. In a large business, it is easier to prepare and collect such data, since many companies already use streaming data entry systems from various types of documentation, this corporate information is streamlined and structured. The creation of such arrays in medium and small businesses is still less accessible.”
Zhamilya Kameneva calls the high cost of such solutions, the length of projects and the long return on investment (2-5 years minimum) one of the main difficulties. “Secondly, like any new tool, long and painstaking work is needed to create a market for consumers of these technologies,” continues Kameneva. “In addition, I would like to note the lack of highly qualified personnel on the market - the vast majority of foreign vendors and only a few scientific institutions are engaged in artificial intelligence systems in our country.”
According to Dmitry Romanov, the main difficulty, surprisingly, is psychological: “People are used to expecting absolute accuracy from a computer. AI systems have a probabilistic output. They can make mistakes, give wrong answers, and in this they are like a person. Users sometimes tend to overestimate the power of smart technology.”
Vladimir Fomenko is sure that in a few years, as soon as this technology ceases to be new and becomes more understandable, there will no longer be much difficulty in its implementation. “There will be systems or programs that can create AI systems or programs.”
But Rustem Khairetdinov believes that there is no difficulty in implementation - “both the mathematical apparatus, and the algorithms implemented in software, and computing power are now available almost “out of the box” or “from the cloud”. “The difficulty is rather in the formulation of the problem, the construction of a model for analysis. Soon we will face the fact that pure mathematicians, as they are now called data scientists, will be less in demand than specialists in other fields (doctors, technologists, security specialists, linguists, etc.) with knowledge of the principles of machine and “deep” learning” ” , - emphasizes Khairetdinov.
* DDM (English Digital Diagnostics Monitoring) - a function of digital control of the performance parameters of the SFP transceiver (as well as SFP + and XFP). Allows you to monitor in real time such parameters as: voltage, module temperature, bias current and laser power (TX), received signal level (RX).
Project Manager ST Smartmerch, System Technologies Group, Maxim Archipenkov I am sure that "pluses follow from expectations".
“Neural networks, unlike humans, do not have emotions and do not get tired,” says Arkhipenkov. - The human factor and all the errors and problems associated with the character of a person and his low working capacity are excluded - regarding the machine, of course. Neural networks do not have a performance threshold: if a person can check, for example, 100 parts for quality in a day, then the system will check them as many as the server capacities allow. The system is easier to scale: at the same plant, 100 people for quality control in one room are difficult to put.
Marketing Director CDNvideo Angelina Reshina also believes that the main advantages of AI systems "in the speed of data processing, the ability to train the system and save on human resources."
Cezurity CEO Alexey Chaley emphasizes that AI-based products are capable of performing tasks on a qualitatively different level: classifying images, translating text, classifying files, etc. ", Chaley notes.
“The main advantages of currently existing solutions are the ability to automate many areas of activity while minimizing human participation in this and expanding areas where it is possible to use software instead of human labor,” says the founder of the hosting company King Servers Vladimir Fomenko. “At the moment, AI is especially good at analyzing large amounts of data, where a human would take too much time, and conventional programs that do not use machine learning would not be able to achieve the necessary accuracy.”
I agree with colleagues and the director of the department of corporate information systems ALP Group Svetlana Gatsakova: “With the help of AI technologies, the speed and level of automation of processing large amounts of information is significantly increasing - while improving the quality and manufacturability. With the right attitude to new technologies, the completeness of data use increases, as well as the efficiency and quality of management decisions.
According to the CEO of Hawk House Integration Alexandra Ivleva, "AI technology is best suited for optimizing various kinds of mechanical activities, automating routine operations, and using it in hazardous industries." “Proper use of robotics on conveyor lines allows switching to a non-stop operation, optimizes enterprise costs, improves product quality, but requires a serious and lengthy commissioning stage,” says Ivlev. - Not many companies can afford to invest large amounts of money in such technologies, although in the future this will make it possible to significantly reduce the cost of production. The situation is similar with machine learning technologies: for each project, analyze a large sample of data, moreover, using individual algorithms, which requires time and resources. But after the introduction of automation, these operations will occur faster and cheaper than a person can do.”
“Let’s start with the fact that artificial intelligence systems are being developed to improve efficiency in the broadest sense of the word,” recalls the Director of Business Applications at CROC Maxim Andreev. - To implement new ideas, approaches, companies often need to take into account a huge number of factors that an ordinary person simply cannot keep in mind. One of the main advantages of artificial intelligence is the ability to take into account such a diverse number of factors in real time. In addition, unlike a human, an algorithm cannot get tired or change some information on purpose. That is, by introducing artificial intelligence, the company minimizes the possibility of errors caused by these factors. But there is a downside to this: a person can take into account additional details, while a poorly tuned algorithm will continue to work incorrectly. Another plus of artificial intelligence systems is replicability. Take as an example any business process in a company that takes an employee a year to train. Therefore, if we need 10 new employees, then we will spend 10 man-years training them. From the point of view of algorithms, everything is simpler and the cost of scaling the solution is much lower.”
Head of Development and Implementation of AV Solutions at Auvix Alexander Pivovarov believes that the most obvious and superficial pluses include increased efficiency, reduced routine operations and greater ease of use. “For example, if you take such a fairly simple thing as a system for booking and displaying the schedule of meeting rooms, then when you start to study it carefully, you see many opportunities to increase the efficiency of its use, reduce downtime, and so on using “smart algorithms,” emphasizes Pivovarov.
“The main task of digital transformation, one of the tools of which is AI, is to make processes run faster and more efficiently, companies spend less and earn more,” says ABBYY Russia CEO Dmitry Shushkin. - For example, one of our customers in the banking sector automated the processing of documents for opening an account for legal entities. The intelligent system itself types and recognizes documents, then extracts information from them and loads it into the required fields of the banking system. As a result, it takes less than 10 minutes to enter data from documents, 2.5 times faster than manually. The bank calculated that in 3 years it would save more than 270 million rubles on document processing.
According to Plantronics business development manager Alexey Bogachev, “one of the main advantages of AI systems is the ability to get some new materials that are simply not available to us. Since an ordinary person draws conclusions based only on his knowledge, but here we get a deeper analysis that can lead to completely unexpected conclusions. This way you can get a breakthrough in a certain area.”
“Man is accustomed to consider himself the crown of evolution, but we regularly face limitations,” says the CEO of FreshDoc.ru Document Constructor. Nikolai Patskov. - For example, hypersonic aircraft fly at a speed 10 times greater than the speed of sound, a human pilot is simply not able to control such a machine without the help of smart electronics. Human reaction and decision-making speed are not enough to work at such speeds. Artificial intelligence helps us to step over these limitations. AI allows people to react faster, protects against making mistakes, frees them from routine operations and decisions. Such systems can effectively replace a human expert in transportation, forecasting, trading on the stock exchange, consulting, and drafting documents. The use of "smart solutions" also affects the final cost of the product: after all, "robots" do not need to pay salaries, they do not get sick and do not go on vacation, and are not subject to a decrease in efficiency. We see huge potential in the development of intelligent solutions for a wide range of tasks. Participation in the development of this area may allow Russian IT entrepreneurs to turn the market and “ride” the information wave of human development.”
According to the director of business development and marketing of Konica Minolta Business Solutions Russia Zhamilya Kameneva everything, of course, depends on the class of solutions. But for the most part, they are aimed at optimizing and automating processes, saving resources - both tangible and intangible, working and personal time. “To put it simply, their task is to make our life easier,” Kameneva sums up.
“Firstly, such systems allow us to reveal what is hidden from the human mind,” says Navicon International Business Development Director Ilya Naroditsky. - No matter how good a person’s BI tools are, in some cases machine learning is indispensable: for example, if you need to process the statistics of operations on bank accounts of 1 million customers over 10 years. Already today, the machine search for hidden patterns that are not obvious to a person allows many companies to build a business strategy and create management decision support systems. Secondly, artificial intelligence technologies significantly increase the efficiency of all types of communications with consumers. Innovative technologies that can understand and analyze text and voice messages help reduce the processing time of incoming requests and respond to customer requests more quickly than before. Thirdly, such systems can relieve company employees from performing routine operations, which means freeing up their time for solving strategically important issues. The time spent on solving routine tasks could be used to solve creative problems.”
“Such systems make it possible to make decisions for a person in those areas where it is permissible,” says Atak Killer CEO. Rustem Khairetdinov. “While previously automated systems made decisions only within the framework of well-defined “if-then” scenarios, then today’s and tomorrow’s systems will be able to make decisions under vaguely defined conditions and with insufficient information, which previously could only be done by a person.”
Acronis Development Director Sergei Ulasen also notes: artificial intelligence systems solve many tasks that previously required the involvement of a person. At the same time, they often function faster and have a predictable result and quality of work.
“AI technologies really work and help improve business processes, at least partially freeing the human intellect from the routine for creativity and creating something new,” emphasizes the CEO of Preferentum (IT Group) Dmitry Romanov. - It is easy for them to assess the economic effect. For a large class of systems using machine learning methods, their ability to become “smarter” as they work is a definite plus.
According to the marketing director of Vocord company Sergei Shcherbina, the main advantages are that based on "chaotic" facts, poorly structured, unclassified or incomplete information, AI makes accurate predictions. “Relying on them, we get a fundamentally new level of accuracy and speed of decision-making where simple, linear rules do not work,” Shcherbina continues. - Huge arrays of data are constantly replenished, but by themselves they cannot solve problems, AI is just what is needed to analyze them. We already know many examples of the successful application of AI in medicine, in the analysis of global and local economic and social processes, in solving engineering and technical problems, in making investment decisions, and in security systems. Innovations in the field of AI will make it possible to automate a fundamentally wider range of business processes. Thus, in the field of video surveillance and security, for the first time it will be possible to guarantee, without the participation of an operator, to detect potentially dangerous incidents 24/7, to identify wanted persons. There are already many examples of successful applications of AI.”
The main plus, according to the co-founder of the shikari.do service Vadim Shemarova, is that AI systems are trainable. “For example, if we want the system to be able to distinguish people's messages where they want to buy something from messages where they want to sell something, or determine the subject of messages, we do not need to compile a detailed list of words and phrases that express intentions. , mood, theme, etc. We select a lot of example texts on the topics we need, “train” the system on these examples, and then it itself begins to understand the essence of unfamiliar texts,” Shemarov says.
Head of the Research Center for Robotics and AI Regulation, Senior Associate at Dentons Andrey Neznamov also believes that the possibility of learning (supervised learning or self-improvement) can be called the main plus of technologies that are commonly called "AI".
What are the difficulties in implementing these systems?
Briefly summarized, the main advantages of AI technologies, according to IT market experts, are reaching new levels of productivity, automation, efficiency, analysis, learning, decision making, predictability, and learning. However, since this is a new direction, experts see even more difficulties than advantages. Suffice it to say that almost every speaker named his difficulty.
“This is a completely new area. Each task that is being solved now is RnD in its purest form: you need to define, systematize, come up with a solution, implement this solution and implement it, - emphasizes Maxim Arkhipenkov. “This is a creative process that requires a high level of science and expertise both directly in the field of application of this solution - whether it be FMCG, space, medicine, or in the field of implementation of neural network systems.”
According to Alexander Pivovarov, the difficulty is "in finding a balance between hype and real utility, the difficulty of making these technologies invisible to the consumer and the absence of errors in their work."
Dmitry Karbasov believes that "the key difficulty of these projects is related to the unpredictability of the result." “Let's say, when buying a CRM system, the customer clearly understands the functionality that the system offers him, and how he will use this functionality,” says Karbasov. - These are processes, data entry forms, reporting, etc. When implementing an AI system, it is very difficult to predict the result without implementing the project, the disclosure of technologies and algorithms will tell practically nothing to a person without a mathematical education and practical experience, and among customers there are only a few top managers with such a background . The implementation of pilot projects helps, the methodology of which we have debugged and which we use in 99% of projects.”
“There are certainly a lot of difficulties,” Maxim Andreev reflects. - The main one, perhaps, is the lack of large enough data sets for training artificial intelligence. This requires historical data. Let me explain what I mean: for one large company, we made a sales forecast for transportation services - we predicted the weight of cargo and the direction of transportation. We could not achieve good forecast accuracy in any way, we began to figure out what was the matter and found out that in the historical data that was stored in the company, somewhere the weight was taken into account with packaging, and somewhere without. At the same time, there is simply no sign by which this factor could be tracked. That is, once in the past this information did not play a role, but now everything has changed. That is why it is so important to collect all the data that can be collected “on demand”. Technologies for collecting and processing data are constantly evolving, and companies can already implement Data Lake technologies, which are becoming an excellent platform for training artificial intelligence. Another difficulty is that the algorithms themselves are still quite small. Therefore, before the introduction of the company, it is necessary to conduct research. This allows us to find out whether, under specific conditions, on specific data and for specific business processes, it will be possible to build AI, the costs of which would not exceed the value that it gives to the company.”
Anna Plemyashova believes that the main problem is the complete absence or insufficiency of data to build accurate models. “For industrial enterprises, where such solutions require significant investments in infrastructure, this is a delayed economic effect: you must first start collecting and accumulating data, and then you can move on to solutions using intelligent systems. Transitional BI solutions and real-time data visualization allow bringing economic benefits closer, says Plemyashova. - Another difficulty is the need to restructure the business process when introducing intelligent systems. That is, it is not enough to buy such a solution and put it like a flower in a vase or an application on a computer. It is necessary to make this decision friendly to the business process: create, reconfigure or even cancel some operations, retrain people, optimize staff.”
“These systems are based on data and big data,” reminds Sergey Ulasen. - Significant computing resources are needed to train models, and an appropriate infrastructure is required to store big data. Therefore, the implementation of AI systems requires a significant investment in hardware.
In turn, the collection and preparation of data require great organizational efforts, and often the development of new software that helps in data analysis.”
Svetlana Gatsakova believes that the difficulties are primarily "in insufficient attention to the limits of applicability of each specific AI technology, to pitfalls." And also "in the weak interpretability of the results (after all, for example, a neural network does not explain its conclusions), in the difficulties of forming homogeneous data sets for training and testing models." Another difficulty is "blind faith in data and a lack of focus on manager intuition and those factors that are difficult to measure and integrate into DDM processes." On this, according to Gatsakova, "complexities specific to Russian organizations" are superimposed. “This is the low availability of reliable data about the external world of the organization and the resulting risk of becoming isolated on internal information, that is, turning into a kind of autistic organization. In addition, this is a small (compared to leading Western companies) penetration of the DDM culture, limited mainly to graduates of Western business schools.
“AI helps to automate many processes and replace low-skilled employees, but at the same time it requires control from developers, whose work costs, of course, are higher,” says Angelina Reshina. “The system’s learning needs to be controlled so that it does not go beyond the acceptable limits.”
According to Sergey Shcherbina, the difficulties lie in outdated equipment and weak infrastructure, legacy hardware and software platforms, which in difficult economic times and with limited budgets, few people dare to change. “The human factor also has an impact,” Shcherbina emphasizes. - Here there is a shortage of qualified personnel, and an insufficient level of competence, or the conservatism of leaders. Moreover, not everyone understands why this is necessary and why spend money on modernization, when “the old fashioned way” everything seems to work anyway.”
“Among the difficulties in building AI systems, first of all, it is necessary to note the shortage of personnel,” notes Andrey Sykulev. - There are very few specialists, because the requirements here are extremely high: in addition to programming skills, one must master a rather complex mathematical apparatus and have knowledge and experience in subject areas. Quite often, the “showstopper” is the poor quality of data and the lack of infrastructure for their integration. Another important issue is data security, because data consolidated for AI operation can become a target for attack or be used, to put it mildly, for other purposes.”
Alexei Bogachev also believes that one of the main difficulties is personnel. “As with everything new, the question is how to work with it. Since the applied application of any technology requires qualified specialists, and this is a very young direction, it is therefore quite difficult to find people who would understand this.”
The staffing problem has another side. “The main difficulty is that not many top managers of enterprises understand what artificial intelligence is and what its practical application is,” recalls Dmitry Karbasov. - Yes, almost all of them have heard about AI, everyone knows that AI helps to optimize business processes, reduce costs, make certain functions more efficient (logistics, analysis of consumer behavior, forecasting production load and sales volumes, etc.). But few of the customers understand that in order for AI to work as it should, it is necessary to formulate a business task and criteria for its success in business terms. In other words, the customer must understand which of the parameters should be instructed to analyze the AI system and how to deal with the received data from the point of view of making managerial decisions.”
“Two factors can be singled out as the main difficulty in implementing such solutions: human and technological,” says Nikolai Patskov. - The first is the problem of a small number of experts able to interact with artificial intelligence systems. This problem is gradually being solved, the market realizes the value of such specialists and more and more employees master the skills necessary for the developing market. The technological factor can be attributed to the lack of computing power: now we are again developing ideas that we will be able to implement only with the advent of more powerful machines. But given the projected growth in productivity (an increase of 1000 times in the next 10 years), we believe that the evolutionary development of technologies will at least not slow down.
According to Aleksey Chaley, there are three main difficulties: “The first is people. There are very few people in the world who are able to work in frontier areas, who at the same time understand the subject area (in our case, virus analysis), are well versed in mathematics, statistics and machine learning, and also know how to program at least a little. The second is training data. This data must be taken somewhere, and then marked up. Data is very difficult to obtain. Because of this, by the way, the progress of AI development is hindered, since researchers do not have the opportunity to experiment with models. It is not enough to just be a talented analyst and programmer - without data, it is impossible to create anything in the field of AI. And the third is the cost of infrastructure. The initial investment in infrastructure can be quite significant.”
“In order for artificial intelligence to solve business problems well, the technology must be “customized”, Dmitry Shushkin believes. - Any machine, like a person, needs to be trained on actual data in order to make accurate decisions. To teach such a system, one first needs to collect or synthesize a large amount of well-labeled data - for example, information about finance, production, customer service, and so on. In a large business, it is easier to prepare and collect such data, since many companies already use streaming data entry systems from various types of documentation, this corporate information is streamlined and structured. The creation of such arrays in medium and small businesses is still less accessible.”
Zhamilya Kameneva calls the high cost of such solutions, the length of projects and the long return on investment (2-5 years minimum) one of the main difficulties. “Secondly, like any new tool, long and painstaking work is needed to create a market for consumers of these technologies,” continues Kameneva. “In addition, I would like to note the lack of highly qualified personnel on the market - the vast majority of foreign vendors and only a few scientific institutions are engaged in artificial intelligence systems in our country.”
According to Dmitry Romanov, the main difficulty, surprisingly, is psychological: “People are used to expecting absolute accuracy from a computer. AI systems have a probabilistic output. They can make mistakes, give wrong answers, and in this they are like a person. Users sometimes tend to overestimate the power of smart technology.”
Vladimir Fomenko is sure that in a few years, as soon as this technology ceases to be new and becomes more understandable, there will no longer be much difficulty in its implementation. “There will be systems or programs that can create AI systems or programs.”
But Rustem Khairetdinov believes that there is no difficulty in implementation - “both the mathematical apparatus, and the algorithms implemented in software, and computing power are now available almost “out of the box” or “from the cloud”. “The difficulty is rather in the formulation of the problem, the construction of a model for analysis. Soon we will face the fact that pure mathematicians, as they are now called data scientists, will be less in demand than specialists in other fields (doctors, technologists, security specialists, linguists, etc.) with knowledge of the principles of machine and “deep” learning” ” , - emphasizes Khairetdinov.
It all started with the scientific and technological revolution, which served as a powerful impetus for the development of technology. It was then that the transition from an industrial society to a post-industrial one took place. Of course, Nikola Tesla with his alternating current, Alexander Popov with the invention of radio, Alexander Bell - thanks to him, humanity got acquainted with the telephone are considered geniuses who turned the usual picture of the world upside down.
It is worth mentioning the people who until recently created or continue to work in this fertile field. Bill Gates, Mark Zuckerberg, Elon Musk are outstanding minds who have made and continue to make a significant contribution to the development of society today. They move forward our new, high-tech world. And very soon a new miracle will appear before the eyes of people. The indefatigable Elon Musk said that in ten years it will be possible to write messages using the "power of thought." Relatively recently, he would have been called crazy or an eccentric, but in the good old days they could have been counted! But in the twenty-first century, the world has become more tolerant and inquisitive. However, it is difficult to surprise humanity, fed up with a large number of new products, informs.
So what can interest our generation and take technology to the next level? The answer is artificial intelligence and nanotechnology. The creation of artificial intelligence will lead to the emergence of new directions, as well as the expansion of the functions of existing ones, such as speech recognition and synthesis, prediction, cluster analysis, and many others. Developments have been going on for a long time, but to create a fully functional model, a new technical solution, known as a "quantum supercomputer", whose computing power can provide full functionality, will be required.
The implementation of these ideas has its global pros and cons:
The first plus is the production factor. Today, the presence of a person is necessary, he evaluates the quality of the work performed and eliminates technical malfunctions.
In the future, artificial intelligence will manage the entire chain on its own. It is assumed that he will do it an order of magnitude better than his creator.
The second is research. The study of space, the depths of the ocean or the earth's core will become safer, provide more opportunities. Where a man cannot pass, a machine will cope.
The third is medicine. Diagnostics, surgery, rehabilitation, implants.
Of the minuses can be identified:
The main thing is the replacement of man by machines, which will lead to mass unemployment of the population. What to do with millions, billions of unemployed people? The question is still open.
The second is failures in the operation of global information and production networks. This can create global problems.
In 2003, there were disruptions in the power supply system in Canada as well. New York, Detroit, Toronto, Montreal, Ottawa were left without electricity. Traffic jams, hundreds of thousands of people locked in the subway, facts of looting, loss of life, fluctuations in world oil markets.
This is such an annoying call. The reasons were various. Lightning strike, failures at nuclear power plants, but the fact remains. Fifty million people were left without power for several hours, and this led to panic and confusion. Some behaved like lost children, others were worse than animals.
The world is very fragile, and the coating of human civilization is very thin.
The third is the seizure of power on the planet by the AIs, the enslavement or complete destruction of people. Today, such a scenario is considered only in science fiction films and books. But it is not the first time for mankind to make a fairy tale come true. And not necessarily a fairy tale with a happy ending.
But let's be optimistic. We believe in human genius and new names in the world of high technologies and humane ideas. Civilization has stood on the brink more than once, but every time people appear with advanced, non-standard thoughts that do not allow them to fall into the abyss.
Error in the text? Select it with your mouse! And press: Ctrl + Enter
Popular
The leader of the political force "Civil Position" Anatoly Gritsenko, in terms of the level of sympathy from the electorate, significantly outstripped the co-chairman of the "Opposition Platform - For Life" faction Yuriy Boyko.
The scandalous prima ballerina of Russian show business, dancer, and more recently, singer Anastasia Volochkova, once again shocked the audience with her antics, the site reports.
The answer to the question why socialists all over the world promise to make someone dramatically happy, to shower money that they will take away from someone, is very simple. The thing is that this is the only economic
According to the website, it is no longer a secret to anyone that one of the most influential Ukrainian oligarchs, Igor Kolomoisky, has a great influence on some politicians in the upcoming presidential elections.
Some plants for the apartment not only perfectly decorate the home, but also purify the air, and also serve as a prevention of many diseases. As the site informs, experts told which
The candidate for the presidency of Ukraine, the leader of the "95 quarter", actor Volodymyr Zelensky shared a new photo with his fans, where he showed how his day off goes, replyua.net reports.
NUTRITION BEFORE WORKOUT
Pre-workout nutrition is based on the consumption of alternative energy substrates (mainly carbohydrates) in order to keep the body's energy reserves intact for as long as possible. Proper pre-workout nutrition is a great way to replenish your energy levels and plays an important role in increasing the effectiveness of your workout. Need to consume food 60-90 minutes (depending on metabolism and food volume) before training. Food should contain grams: from 25-40 protein, 70-90 carbohydrates and no more than 15 fat.
Carbohydrates
Glycogen stores are in huge demand during intense strength training. Glycogen is a sugar that is stored in the liver and muscles. Since anaerobic exercise does not involve the saturation of the blood with a large amount of oxygen, the body is not able to break down fats and use them as a main source of fuel. Instead, the body must use both sugar stores, the one stored in the muscles and the one supplied by the liver to the blood.
Most of your pre-workout meals should be complex carbohydrates. Complex carbohydrates have a low glycemic index (GI). GI is a measure of the immediate effect of carbohydrates on blood glucose levels. Simple carbohydrates are easier to digest and thus have an immediate effect on blood glucose levels, which means they have a high GI. Conversely, more complex carbohydrates take longer to digest and therefore have less of an impact on glucose levels and have a lower GI.
But why is all this important and what is the point of their consumption? Low GI carbohydrates (complex) are broken down over a long period of time, and breakdown products (simple carbohydrates that are formed from digested complex ones) are steadily released into the blood for a long time. This avoids the ups and downs of energy and performance and helps maintain an anabolic state in the later stages of a workout.
As a general rule, pre-workout meals should consist of grains − oatmeal, brown rice, whole grain bread, sweet potatoes, durum pasta, legumes, nuts.
Squirrels
Proteins are known as the building blocks of muscles. They are made up of smaller units - 9 amino acids that cannot be produced in the body and must come from food or supplements (Essential Amino Acids). Proteins that contain all the essential amino acids are called complete proteins. All animal products (meat, eggs, dairy products) are complete proteins and must be added to the diet before and after training.
Protein Sources:
Meat (beef, turkey, chicken)
Fish (salmon, tuna)
Eggs
Dairy
nuts
Another pre-workout strategy is to take advantage of the increased blood flow to the muscles being worked, as this is when the muscles are most sensitive to nutrients.
The lack of amino acids has always been a limiting factor for protein synthesis, so by including protein in your pre-workout diet, you will contribute to the accelerated delivery of amino acids to muscle tissues.
Try to avoid the presence of fats in the diet before training. Fats greatly slow down the digestion process. Since the human body increases blood flow to those organs that need it, being in a state of heavy digestion, a loaded stomach takes precedence over muscles, which is not good. Therefore, those grams of fat that you get along with your carbohydrate and protein sources will be quite enough.
An example of a pre-workout meal
Chicken breast - 200 gr. (45 gr. b.)
Brown rice - 300 gr. finished product (65 gr. ug.)
Whole grain bread - a piece of 50 gr. (20 gr. ug. + 7 gr. b.)
Juice - 300-500 ml
Oatmeal - 300 gr. (60 gr. ug. + 10 gr. b.)
Fat-free cottage cheese - 200 gr. (44 gr. b.)
Green banana - 1 piece (30 gr. ug.)
Water - 300-500 ml
PRE-WORKOUT SUPPLEMENTS
So, you have had a good meal with a wholesome meal, added carbohydrates to the body to replenish glycogen stores and provided some complete protein. Now you need to immediately provide the body with additional nutrients in the form of supplements to increase the effectiveness of the workout. Sports nutrition is quickly absorbed, therefore, it should be take 15-30 minutes before workout. The following is a list of some popular pre-workout supplements:
Whey Protein- perhaps the most important supplement both before and after training. Provides you with protein and branched chain amino acids, which will be delivered to muscle cells as quickly as possible during training.
Creatine - increases muscle volume and energy, and also retains water in the muscles, which contributes to good hydration. It is a safe supplement.
BCAAs are undoubtedly essential amino acids in any bodybuilding diet. They promote muscle growth and recovery. However, the need for their use may be questionable. After all, protein powders (especially whey protein concentrate, not isolate) already have an excellent set of amino acids. Therefore, there will simply be no point in using BCAA, and it is worthwhile to better see the label on your whey protein in advance.
NO2 - nitric oxide, dilates the blood vessels so that more blood can be delivered to the muscles. This means that more nutrients can be delivered to the muscles.
Caffeine is an excellent stimulant that provides the body with energy and helps to concentrate. Caffeine works in the opposite direction of creatine (the former acts as a diuretic, the latter stores fluid), so you should choose one.
Leukic Hardcore - a complex of nutrients that maintains an optimal level of insulin in the blood and creates favorable conditions for maximum growth of muscle tissue.
Nano Vapor - a complex of special biologically active compounds, stimulates the anabolism of muscle cells and prevents the catabolic effect.
An example of a pre-workout shake
Whey protein - 2 scoops (about 40-50 gr. b.)
Creatine - 5 gr.
BCAA - 5-10 gr. (depending on the composition of the protein, take only BCAA or only protein)
NO2 - 2 capsules
Water - 500 ml
nano vapor - 2 scoops (50 gr.)
Leukic Hardcore - 1 Serving (6 Capsules)
Water - 300 ml
Send your good work in the knowledge base is simple. Use the form below
Students, graduate students, young scientists who use the knowledge base in their studies and work will be very grateful to you.
Hosted at http://www.allbest.ru/
Introduction
Fundamentals of artificial intelligence
Machine and Human Capabilities: Examples, Practice and Analysis
Pros and cons of using artificial intelligence in management
Conclusion
List of used literature
Introduction
Our world may be nearing disaster. It is not difficult to come to such a conclusion with any sober analysis of the state of our planet. Economic stagnation, poverty, rampant inflation, massive unemployment, overpopulation, political strife, terrorism, wars and the danger of their outbreak, as well as the threat of a doomsday, have not bypassed a single corner of the globe. Of course, mankind has always faced many problems, but today's problems, of course, seem to be more significant than those of the past. Now we seem to have indeed reached the point where something very essential must be given up. Unfortunately, it is customary to lay the main blame for this on the development of technology, i.e. just what humanity throughout its history has tried to find a solution to many problems.
Technology has been accompanying man for thousands of years and is nothing more than the combined result of mankind's aspirations for a better life. Now, however, there are people who believe that the development of technology, on the contrary, worsens rather than improves life. The challenges people face today vary in scope, from social upheaval driven by technological change, unemployment, pollution and the threat of nuclear annihilation, to alienation and dissatisfaction with work and its concrete outcomes. To this we can add the following. It is possible that the complexity engendered by technological progress is responsible for the untreatable ills of the economy, and that technical systems that become so complex that people will soon lack the knowledge and understanding to manage them are beginning to pose a significant danger.
Naturally, the question arises, how to solve these problems? Can the inanimate creations of technology find answers to the questions it itself has posed, and to the myriad others that drive the human race to despair? Are machines themselves capable of coming up with solutions that elude the human mind? In this term paper, I would like to prove that in principle this is possible and, moreover, in the future it must happen.
Such a statement is not just a dream. It is based on the discoveries that are being made day after day in various laboratories of the world, the most fruitfully working in the field of computer technology. For a long time, it was erroneously believed that the output of a computer could only be what was put into it at the input. This notion has certainly been confirmed over the past 30 years in most of the work related to data processing. Now, however, it is irrefutably proven that something completely new can be obtained from computers, namely knowledge. This knowledge, in turn, can take the form of original ideas, strategies, and solutions to real problems.
Until now, the knowledge created by the machine is not of great practical importance; it is not capable of curing those deep ailments with which our world is sick. This was to be expected: after all, a biologist, having begun the synthesis of living matter, at best expects to get only a virus, and not an adult horse. But, no doubt, in time it will be possible to direct computers to search not for solutions in chess or some other game, but for more pressing problems facing society. And most likely he will find them.
Of course, this will take a lot of time, but if a person sets himself such a goal, he will achieve it sooner or later. I would like to believe that the day will come when poverty, hunger, disease and political strife will be tamed, and new knowledge will play a role in this, received by computers, acting as our assistants, not slaves. In addition, the mental and artistic potential of a person will receive completely different development opportunities, which are difficult to even imagine today, and the gates of human imagination will open wide as never before.
And we must not miss our chance, although it may not be easy. We will have to completely abandon the traditional technical approach with its main goal - to maximize the economic effect of the use of machines and move to a strategy aimed at making the processes occurring in systems quite understandable to people. To do this, computers must learn to think like people, i.e. if the computing systems of the next decade do not fall into the "human framework", they will become so complex and incomprehensible that a person simply cannot manage them. The inability to cope with such complex systems at first will simply lead to breakdowns (in terms of the many applications of these systems that are available today); if we are talking about military warning systems, control systems for nuclear power plants or global communications systems, then getting them out of our control can lead to catastrophes on a worldwide scale.
Fundamentals of artificial intelligence
Since the beginning of the 80s, a new stage has begun in the work on artificial intelligence - the creation of industrial and commercial samples of intelligent systems. The industry producing such systems began to develop, which means that potential consumers of its products appeared. What distinguishes intelligent systems from other creations of the human mind? What can we expect from their appearance in the near future?
The key term of artificial intelligence is the term "knowledge". With a certain degree of approximation, one could say that intelligent systems are systems that use knowledge. This is what they differ from other artificial systems (including software systems that were implemented on computers in the era preceding intellectual systems), based, as a rule, on the same computers.
Remaining at a metaphorical level, we can say that before computers "understood" how to execute the program entered into them, but "did not understand" what they were doing, and with the advent of intelligent systems, computers learned to "understand" how to build the necessary for problem solving program and what this program does. Let's explain this important idea. With the traditional method of solving a problem on a computer, the essence of the problem itself, its meaningful interpretation, was known to the programmer who prepared the program for the computer. These could be various programs: for playing "backgammon" or "go", calculating the trajectory of the spacecraft, or calculating payroll. When these programs were introduced into the computer, the content side of the tasks disappeared - the computer, by virtue of its design, carried out the commands of any of the programs in a qualitatively identical way. In the “computer-programmer” pair, only the first one knew what the computer was doing, and the computer, like a powerful adding machine, simply performed the necessary transformations and calculations.
There was danger in this traditional scheme. It consisted in the indissolubility of the pair "programmer - computer" in solving problems. The programmer, like a "galley slave", had to interact with the machine, "indifferently" grinding any information entered into it.
The emergence of intelligent systems testified to the breakdown of this paradigm. If knowledge is introduced into the computer memory about how programs are built from the conditions of the problem and what this or that task means in a given problem area (i.e., how the goal of the task is interpreted and what are the possible connections between the initial situation and the goal), then the functions of the programmer will be perform by the computer itself. She will automatically, on the basis of the knowledge she has in her memory about the problem area, about the tasks that may arise here, and about the ways to solve them, will be able to independently draw up the necessary program and execute it.
This moment is fundamental. The knowledge entered into the computer now allows her to "understand" what she must do when the need arises to solve a problem. By the way, when exactly this need arises, the computer also "knows" itself (although the requirement to solve the problem may come from outside - from the user of the system).
This is how the main tasks that face that branch of artificial intelligence, which today is increasingly called knowledge engineering, are formed. What are these tasks? First of all, it is the task of collecting the knowledge that is needed by the computer. This task is far from being as simple as it might seem at first glance. After all, in addition to the knowledge that is embodied in various texts, professionals have a lot of knowledge that cannot be found either in manuals, or in instructions, or in monographs. This is the knowledge that is usually called experience, skill, professionalism. Often, an experienced specialist does not even suspect that he has vast knowledge. It seems to him that he “just works and that’s it”, and a colleague who has not yet gained experience looks at him with envy, not understanding why everything goes wrong for him. To be able to obtain this knowledge from an expert expert, to be able to present it in a form suitable for entering into computer memory, is the first and very non-trivial task of a knowledge engineer. But this is not enough. Accumulating knowledge obtained from various sources, one must constantly take care that they do not form a contradictory system: any new knowledge must be linked to previously existing ones. The emergence of new knowledge may require some kind of restructuring of the previously created knowledge base. This requires special management procedures. Developing and manipulating such procedures is the second task of the knowledge engineer.
Receiving information from the surrounding world, analyzing emerging situations, a person constantly refers to the information stored in his memory. Attracting what is already known to understand the new, a person, using his knowledge, sort of completes the input descriptions, replenishes them. In any conversation between two people, understanding the remarks is possible only because a lot of additional information about the subject of the conversation is stored in the memory of the interlocutors. And computers for replenishment of knowledge should have a set of similar procedures. For this purpose, the so-called pseudophysical logics are used: temporal, spatial, causal and others. With their help, input descriptions are replenished, which ensures their understanding. In addition to replenishing descriptions in knowledge bases, other procedures are also carried out: generalization and classification of incoming information, hypotheses about the relationships of facts stored in the system’s memory, diverse in type, reliable and plausible conclusion of derived facts, etc. This is another field of activity for a knowledge engineer.
However, referring directly to the topic of the course work, it is important for us to note the massive introduction of computers in all areas of management. This is a question about the ability of a human administrator to understand the decisions that a computer, included in the control system, makes. The control systems of complex technical complexes today are literally "stuffed" with computers interconnected in complex structures. Working at speeds inaccessible to humans, processing a huge amount of various information received from the control object and from other machines, the computer makes decisions that are often incomprehensible to humans. The only way to understand them is to ask the machine the question: why is the solution like this? And the computer is obliged to give the necessary explanations. For this purpose, it must have a special explanation subsystem that allows the computer to "understand" why it made a particular decision. The emergence of explanation subsystems can be seen as the first step towards the "humanization" of technical systems. It is difficult to overestimate the importance of this step. Technical systems have gone too far in their development, it has become too difficult for a person to interact with them, and the consequences of the actions of our smart but soulless assistants can be too dangerous.
The development of works in the field of artificial intelligence and the widespread introduction of intelligent systems into our lives are evidence of a new stage in the path of scientific and technological progress. It is inevitable - and we must be prepared to meet its consequences with full understanding of what is happening. It is not the problem of WHO WHOM and not the fear that THEY will enslave US if we do not take action should determine this new stage in the life of mankind, but the commonwealth WE + THEM, from which humanity will undoubtedly benefit greatly, because it will help us solve problems, with which we alone can not cope.
Machine and human capabilities:examples, practice and analysis
On March 28, 1979, at the Three Mile Island nuclear power plant (Pennsylvania, USA), an alarm sounded in control room No. 2. At first, the operators did not show much concern, since minor accidents at the station were not so rare, but after a few minutes it became clear that something much more serious had happened this time. A tiny valve in the pneumatic system was stuck, and this caused the water circulation in the secondary circuit cooling water system to stop. Moments later, the uranium core of the reactor began to heat up, and despite all the efforts of the operators, the situation only worsened. The safety valve opened and stuck in this position; radioactive water and steam went into the reactor building, and hence into the atmosphere. A huge bubble of hydrogen formed under the roof of the reactor vessel, which could explode at any moment. There was a threat that the uranium fuel itself would begin to melt. Any of these events could lead to radioactive contamination of the entire territory of the piece. Pennsylvania.
For the next few days, plant personnel, along with experts from the Nuclear Regulatory Commission, fought to take control of the reactor, and a frightened world watched this struggle with alarm. The state governor ordered the evacuation of children and pregnant women from the danger zone, and many residents left on their own. Only a week later, the Metropolitan Edison Company, which owned the station, announced that work had begun on the conservation of the shutdown reactor, and life in Pennsylvania gradually began to return to normal. It took several years to clear the "Augean stables" that the reactor building had become.
The commission, which studied the role of the human factor in this incident, came to the following conclusion: “... such an avalanche of information fell upon the operator: display indications, warning signals, printout data, and the like - that it was completely impossible to identify a malfunction and choose the right measures for elimination."
The Presidential Commission agreed with this conclusion, concluding that the blame should be placed on "insufficient attention to the human factor and its role in ensuring the safety of nuclear power plants." The lesson learned from this accident is obvious: until the design of technical systems is thought out in every detail so that everything that happens in them is absolutely clear to the maintenance personnel, until the information is presented in a form that is convenient for perception by the human eye and brain, and not machine, any malfunction in the automated system can render it completely unmanageable.
In 1975, the Dutch steel company Estelle Hugovens installed a new, highly automated hot rolling mill at its plant on the seafront near Amsterdam. Anticipating the gigantic increase in labor productivity due to the introduction of advanced technology, the management of the enterprise was shocked to find that, in fact, output had declined. Consultants from the British Steel Corporation were invited to help, who, in the report on the results of the study, indicated that the main reason was the improper organization of the interaction of operators with the machine. The New Scientist described it this way: “Operators lost confidence in themselves so much that in some cases they simply left the control panel unattended. In addition, the operators did not always fully understand the control theory underlying the program of the control computer, and this prompted them, if possible, to "withdraw" from control until obvious problems were discovered. But due to the fact that they intervened in the process with a great delay, the average productivity was lower than in plants using traditional methods of management. Thus, automation entailed a decrease in productivity and at the same time further removed operators from control processes.”
The problem was compounded by the fact that in the new design of the rolling mill, the strip is hidden from view throughout the entire rolling path, which did not allow operators to even visually follow the process. In their report, the consultants, in particular, unreservedly insisted that operators should be brought closer to the technological process, and information displays should help people understand the meaning of the decisions that automation makes, and not just report on the progress of the process.
Next example? air traffic management, which is of equal concern to both passengers and controllers around the world. It has become all too common for planes to nearly collide in flight, not to mention electronic failures that leave controllers helpless for precious seconds, if not minutes. According to the University of Illinois Research Coordination Laboratory, computer-assisted air traffic control in America is becoming so complex that operators can sometimes find it difficult to understand what is happening. As for the prospects for the future, there are two opposing views on what should be the management systems that will replace the existing ones. Some experts call for more and more automation, believing that this will eliminate the uncertainties associated with the presence of a person; others believe that humans and machines should be partners of sorts in this common cause. But no matter what path the further development of control systems takes, situations are always possible where human intervention is required. And if the creators of the system do not take care in advance that a person can understand how the system works, then his intervention will most likely be very minor and will occur with a great delay.
The military issue cannot be ignored.
During the eight months of 1979-1980. The US military received three false alarms warning of an "attack" by Soviet missiles. All signals came from the control center of the North American Air Force, hidden in the depths of the mountain in pcs. Colorado. The first false alarm was simply the result of an operator's mistake, who had carelessly planted a tape of training information into the system. The second time one of the components of the system failed: the integrated circuit failed. The third signal turned out to be intentional - it was an attempt to reproduce the conditions of the second alarm for verification purposes.
Fortunately, a few minutes after these false alarms, the all-clear was given, but the nervous overstrain caused by them was not forgotten. It is quite clear that a system that can literally lead to the end of the world must be done this way. so that the possibility of misunderstanding in the relationship between man and machine is completely excluded.
The conclusion that follows from these stories is obvious: as technical systems become more and more complex, they become more and more difficult to understand, and therefore, to control. This is especially true of computing systems, which, even if intended to do the simplest things, must be very complex. We strive to ensure that they can solve problems of practical importance, and thereby increase their complexity to a level that is beyond the ability of a person or even a group of people to understand. That time has already come. As we have just shown, large computing programs and operating systems grow to a scale where neither their creators nor users are able to cope with them.
If computing systems continue to develop along the same path as now, when more and more functions are assigned to their already not very reliable architecture, then there is no doubt that the computers of the 90s will become completely unusable: uncontrollable and frightening - sort of helpers of the world "evil spirit". Human society, which is already heavily dependent on such machines, will face a crisis of monstrous proportions. Computing machines - as they exist now - have in a sense already reached the limit of their capabilities. Today, the main task is no longer to bring their performance to the maximum, to extract everything possible from machine resources. On the contrary, their work should be based on a completely different idea - the idea of anthropocentrism. In order for us to understand how machines work, we need to learn how to organize them in the image and likeness of the work of the human brain.
We can take this sinister story further by imagining our future as it has been described by science fiction writers, starting with Samuel Butler: a world in which machines have taken over. This idea is generally dismissed by technical experts as absurd. But is it really that absurd? Take, for example, computers that are already used in managing the life of our cities. Its functions include not only the tasks of the central administration, but also public services, maintenance of order in the city, education, banks, air traffic, traffic control, problems of building and planning organizations. And there comes a moment when the corresponding computer networks begin to directly address each other - initially for the simplest reasons. If, for example, in one system it is decided to dig up the road, then the garbage trucks need to change the route of movement. If someone orders a plane ticket, the airline must check whether he is entitled to use the presented credit card.
Turning now to less nightmarish but more pressing issues, let's look at the deep economic stagnation, the high unemployment rate, and the crises of confidence that have increasingly plagued the world in recent years. All these phenomena, which actually take place, are at first glance completely inexplicable. Let's start with the problem of economic growth, or rather the lack of it. In fact, the productive capital of the industrialized countries is not shrinking. However, due to the continuous progress of science and technology, it is constantly being transformed. What is the nature of this change? The invested capital brings higher profits. Factory workers can now produce more in a day than they could thirty years ago. A farmer may cut more hay than needed to justify renting a mower. The day is not far off when self-driving mowers will appear.
Moreover, scientific and technological development does not just happen at a constant speed: no matter how we evaluate its pace, it is obvious that they are steadily increasing. Why, then, are we not getting richer at the same rate? Even making allowance for the losses associated with the reorganization of work in certain industries, humanity as a whole should be in a significant gain. Apparently, a certain force is acting, blocking that cornucopia from which, it seemed, blessings should now be pouring on all of us.
It seems that we are all united in our regrets about this. But different people have different attitudes about which part of this process should be stigmatized. Some are absolutely sure that this is the fault of the trade unions, which are in a secret conspiracy with an invisible network of subversive elements and terrorists from all over the world who achieve their political goals. For others, the culprits are to be found in the offices of giant corporations and banks, possibly allied with a secret network of transnational monopolies and cartels, led by one or two "evil dwarfs" from Zurich, pursuing their own political goals. There is also a third "school of thought", perhaps not as subject to passions as the previous two, but even more delusional, which believes that technology itself is to blame for everything. It is not uncommon for an angry shopper to take it out on a non-functioning vending machine until it is completely unable to operate.
Although, perhaps, such an anti-technical position is not so delusional. This idea can at least be discussed, since the examples given earlier do show that our technical achievements are somewhat similar to a non-working automaton.
We will have to make a small digression into the depths of history in order to find out whether there was any stable and at the same time evolving process throughout its entire length? Such a process is not difficult to find - it was the impetus for the development of agriculture. And millennium after millennium, our ancestors did not seem to notice that this process was continuously going in the same direction, until in the XIX-XX centuries. we have not reached the last stage of acceleration. Such a steady process was a gradual, albeit painful, with many failures and stops, the growth of man's understanding of the world around him, the growth of his ability to manage this world.
Today, with the help of computer technology, we are trying to learn how to solve complex problems that cannot yet be solved on a computer - problems that cannot be solved “head on”, finding the answer to the question in a finite number of steps by simple calculations. However, it happens that, although the problem itself is very difficult, the inverse problem is solved much easier. For example, calculating the square root is very difficult, but the square of a number is very easy. It is possible that the student will find it more economical to calculate the squares of all the numbers that he may be asked about and fill out a huge table of results (only writing them in reverse: first the squares, ordered by magnitude, and for them the bases, perhaps with some interpolation to fill passes). Then, if you need to find out some square root, you can just look at the table. But this method has one big drawback - the result is completely inexplicable to the user.
The question arises, is it not better not to invent this kind of stupid reference systems, the very existence of which humiliates a person, because they neglect his judgments and understanding. It is interesting that for the first time this argument was brought by Plato more than 2300 years ago. In his Phaedra, Socrates tells the story of the Egyptian god Toga, who came to the king of the gods Tamuz with the words: "My lord, I have invented a witty thing called writing, it will improve both the wisdom and the memory of the Egyptians."
In response, Tamuz stated that, on the contrary, writing is a low-grade substitute for memory and understanding. “The one who acquires it will stop training his memory and become forgetful, he will rely on writing, hoping that these icons will remind him of something, instead of relying on his internal reserves.”
Socrates quotes Ammon as denouncing the vicious idea that “one can convey or gain clear and distinct knowledge about a subject through writing, or that written words can do more than simply remind the reader of what he already knows.” In other words, a person may decide that wisdom is in writing, when in reality the wisdom must be in the person himself. “One might suppose,” adds Socrates, “that the written words understand what they say, but if they are asked again what is meant by such and such, they will again and again give the same answer. ".
In other words, Socrates, as it were, complains about the fact that writing will not be able to pass the famous Alan Turing test (according to this test, a machine can prove that it has intelligence if it manages to convince a person talking to it through a teleprinter that his interlocutor - human). Indeed, if the machine could explain what it contained, then one could assume that in some sense it "understood", thus demonstrating its intelligence. Like writing, future reference systems with trillions of bits of memory will not pass the Turing test. But, like writing, such systems certainly have a right to exist, and will help change the world. Is it good or bad? Until we get to the bottom of Socrates' claims to this new problem, these gigantic help systems will be only partly a boon, and often a big nuisance. Recall that such databases contain only basic elementary facts relating to a particular issue, and do not include any understanding, conclusions, judgments, classification concepts, and the like.
In order for any creatures - human or machine - to communicate with each other, they must have the same mindset. Since we cannot change the mindset of people, we will have to change it in machines. We need to completely redesign everything that programs do when solving a problem, not just the way they interact with the user. The way information is stored in a program, i.e. the way of presenting the solution of the problem must be understandable to a person and described by concepts already familiar to him. Expert systems based on inference rules are specifically designed to deal with human concepts, both when they are obtained from experts in the field, and when they are explained to the user. This is not bad for a start, but there is still a lot to be done to establish communication between man and machine in the language of concepts.
If similar ideas are used to solve problems of automation of production processes or in other control systems, then we will call such automation “soft”. The need for it is constantly growing, which makes it possible to at least partially neutralize the excessive complexity associated with rigid automation. The most pressing social need now is not to expand the process of automation, but to humanize it. Of course, for simple or medium complexity tasks, the "opacity" of control systems is not so dangerous, and therefore we put up with it for a long time. Let's say that the resource allocation program does it better than the project manager. In that case, why should he be curious about how she does it, or challenge her decisions, if he gets what he wants? Let it be a "black box" to the extent that it is set by the program.
However, there are other applications of information systems where the ability to "look inside the box" is essential. So far, there are few of them, since the processes of information processing have yet to be deeply embedded in more and more complex and responsible areas of human activity. Complexity and responsibility are two independent characteristics of systems that lead us to insist that the program work within "human limits". Some problems are so difficult that it is simply impossible to solve them without an intellectual partnership between man and machine. Others concern matters of life and death, or the very possibility of managing the economy.
With soft automation, the system already at the design stage is adjusted to the human mindset. If, looking into the future, we imagine hordes of robots working together in our factories, the question inevitably arises: “How will communication be carried out between them? By wire, using infrared radiation or radio signals, or through some other channels inaccessible to humans? Of course, it would be better to make this connection using a synthesized voice, because this would allow the person on duty to hear what is happening, and, as practice has shown, this is quite possible.
Pros and cons of usingartificial intelligence in management
The trend towards automating factories and machines has been around for a long time. Except for some special purpose, no one thinks anymore about making bolts on a conventional lathe, where the turner has to watch the movement of the cutter and adjust it manually. At present, the production of bolts in large quantities without serious human intervention is a common task of an ordinary screw-cutting machine. Although this machine does not specifically use either a feedback process or a vacuum tube, this machine achieves almost the same goals. Feedback and the vacuum tube made possible not the sporadic construction of separate automatic mechanisms, but the general policy of creating automatic mechanisms of the most diverse types. In solving this problem, the principles of such devices were reinforced by our theoretical study of communication, which fully takes into account the possibilities of communication between machine and machine. It is this confluence of circumstances that makes the new age of automation possible today.
The industrial technology now in existence includes the entirety of the results of the first industrial revolution, along with many of the inventions that we now regard as the forerunners of the second industrial revolution. What the exact boundaries between these two revolutions might be is too early to say. In its potentiality, the vacuum tube definitely belongs to an industrial revolution distinct from the energy age; and yet it is only at the present time that the true significance of the invention of the vacuum tube is sufficiently understood to place the present century in a new, second industrial revolution.
Let's paint a picture of a more perfect age - the age of automation. Consider, for example, what the car factory of the future will look like, and in particular the assembly line, which is the part of the car factory that uses the most amount of human labor, the sequence of operations will be controlled by a device like a modern high-speed computer. It is possible to reduce all mathematics to the performance of a series of purely logical problems. If such a piece of mathematics is embodied in a machine, then that machine will be a computing device in the usual sense. However, such a computer, in addition to solving ordinary mathematical problems, will be able to solve the logical problem of distributing through channels a number of orders regarding mathematical operations. Therefore, such a device will contain, just as modern high-speed computers do contain, at least one large node, which is designed to perform purely logical operations.
The instructions for such a machine - I am here also talking about current practice - are given by a device that we call a programming coil. Orders given to the machine can be sent to it by a program coil, the nature and extent of the instruction of which is completely predetermined. It is also possible that real contingencies that the machine encounters in carrying out its tasks may be transferred as the basis for further regulation to a new control tape created by the machine itself, or to a modification of the old control tape.
It might be thought that the present high cost of computing machines precludes their use in industrial processes, and, moreover, that the sensitivity of operation required in their design and the variability of their functions preclude mass production methods in the construction of these machines. None of these statements are correct. First, the huge computers currently used for very complex mathematical work cost about hundreds of thousands of dollars. Even this price would not be out of reach for a control machine in a really large plant, but it is still too expensive.
Modern computing machines are developing so rapidly that virtually every designed machine is a new model. In other words, most of these apparently exorbitant costs go towards paying for new design work and the production of new parts, which require very highly skilled labor and the most expensive conditions. If, therefore, the price and model of one of these computers were fixed, and if this model were used by dozens, then it is highly doubtful that its price would be more than a sum of the order of tens of thousands of dollars. Such a smaller machine, not suitable for solving the most difficult computational problems, but nevertheless quite suitable for running a plant, would probably cost no more than a few thousand dollars in any kind of moderate-scale production.
Consider now the problem of mass production of computers. If for mass production the only favorable opportunity would be mass production of standard machines, then it is quite clear that for a considerable period the best we could hope for is production on a moderate scale. However, in each machine, the details are mostly repeated quite often. This applies equally to the storage device, and to the logical apparatus, and to the arithmetic unit. Thus, the production of only a few dozen machines is potentially mass production of parts and has the economic advantages of mass production.
Yet it would seem that the sensitivity of the machine should mean the need to create a special new model for each individual job. This is also incorrect. Even with a rough similarity in the type of mathematical and logical operations that are required of the mathematical and logical nodes of the machine, the overall performance of the machine of its tasks is regulated by the program coil, or at least the original program coil. The manufacture of a program coil of such a machine is a very difficult task for a highly qualified specialist; however, this is a job that is done once and for all, and when the machine is modified for the purposes of a new industrial assembly, it needs only to be partially repeated. Thus the cost of such a skilled technician will be spread over a huge amount of output and will not really be a factor in the use of the machine.
The computing device is the center of an automatic factory, but it will never represent the whole factory. On the other hand, it receives its detailed instructions from elements that are of the nature of the sense organs, such as photocells, capacitors for determining the thickness of paper rolls, thermometers, hydrogen concentration meters, and from the general types of apparatus currently created by instrument manufacturers. firms for manual control of production processes. These devices are already arranged in such a way that they transmit indications to individual posts by means of electricity. In order to enable these devices to transmit their information to an automatic high-speed computer, all that is needed is a reading device that converts the position or scale into the form of serial numbers. Such a device already exists and presents no great difficulty either in principle or in structural details. The problem of the sense organ is not new, and it has already been effectively solved.
The control system must contain, in addition to these sense organs, effectors, or components that act on the outside world. Some types of these effectors are already familiar to us, such as motors with control valves, electrical clutches, etc. To reproduce more accurately the functions of the human hand, supplemented by the functions of the human eye, some of these effectors still have to be invented. When machining automotive frames, it is perfectly possible to leave smooth-machined surfaces on metal consoles as reference points. A photoelectric mechanism, powered, for example, by light points, can bring a working tool - whether it be a drill, or a riveting hammer, or whatever tool we need - in close proximity to these surfaces. The final fixation of the position can secure the tool against the reference surfaces and thus establish a tight contact, but not so tight as to cause destruction of these surfaces. This is just one way to get the job done. Any qualified engineer can come up with a dozen more.
Of course, we assume that the instruments acting as sense organs register not only the initial state of the work, but also the result of all previous processes. Thus, the machine can perform feedback operations: either fully mastered operations of a simple type, or operations involving more complex recognition processes, regulated by such a central control as a logical or mathematical device. In other words, the all-encompassing control device will correspond to the animal as a whole with sense organs, effectors and proprioceptors, and not to an isolated brain, the efficiency and practical knowledge of which depends on our intervention, as is the case in a superfast computer.
The rate at which these new devices can be introduced into industry will vary greatly across industries. Automated machines performing roughly the same functions are already widely used in industries with continuous processes, such as in canneries, steel mills, and especially in wire and tinplate factories. They are also known in paper mills, which also work in line. Another area in which automatons are needed is in these kinds of factories, where production is too dangerous for a significant number of workers to risk their lives in operating it, and where an accident can be so serious and costly that its possibility must be foreseen in advance, and not left to the hasty judgment of some person at the scene of the accident. If it is possible to think over the line of behavior in advance, then it can be applied to the program tape, which will control the behavior in accordance with the readings of the device. In other words, such plants must operate under a mode quite similar to the mode of blocking and operation of the switches of a railway checkpoint. Such a regime is already established in refineries, in many other chemical plants and in the handling of the kind of hazardous materials encountered in the operation of atomic energy.
We have already mentioned the assembly line as an area of application for this kind of technique. In the assembly line, as in a chemical plant or a paper mill with continuous processes, it is necessary to carry out a certain statistical control over the quality of the product. This control depends on the testing process. Scientists have now developed these sampling processes by developing techniques called sequential analysis, where sampling is no longer done as a whole, but is a continuous process that occurs along with production. Consequently, those processes that can be performed by a technique so standardized that it can be transferred to a statistician who does not understand the logic behind it can also be performed by a computer. In other words, again except for the higher levels of operation, the machine can also take care of day-to-day statistical control as well as the production process.
Normally, factories have an accounting procedure that is independent of production, but since the data of this accounting comes from the machine or from the assembly line, it can be sent directly to the computer. Other data may be entered into the computer from time to time by a human operator, but most of the clerical work can be done mechanically, and only extraordinary information, such as external correspondence, will be left to people. However, even most external correspondence may be received from correspondents on punched cards, or printed on punched cards by a very low-skilled employee. Starting from this stage, all processes can be performed by the machine. This mechanization can also be applied to a significant part of the library archival fund of an industrial enterprise.
In other words, the machine does not give preference to either physical or office work. Thus, the possible areas into which the new industrial revolution is able to penetrate are very wide and include all low-level decision-making labor, in much the same way that the labor displaced by the machine of the previous industrial revolution included any aspect of human energy. Of course, some professions will not be affected by the new industrial revolution, either because the new control machines are not economical in such insignificant branches of industry that are unable to bear the large capital costs associated with this, or because the work of a number of specialists is so diverse that new program coils will be needed for almost every single job. I cannot imagine automatic machinery such as decision makers being used in grocers or garages, although I can very clearly imagine the use of this equipment by the grocery wholesaler and automobile manufacturer. The agricultural worker, although automatic machines are beginning to take root in his production, is also protected from their complete domination by the size of the land he has to cultivate, by the variability of the crops he has to cultivate, by the special conditions of weather, and the like, with which he has to face. . Where such machines can be used, it is not improbable that decision machines will be used to some extent.
Of course, the introduction of these new devices and the time frame during which they can be expected to be implemented are mainly economic issues, the consideration of which is not the purpose of the course work. Barring some violent political change or another major war, it will take ten to twenty years for the new machines to take their rightful place.
A very important issue is the analysis of the consequences - economic and social.
In the first place, we can expect a sharp drop and final cessation of the demand for this kind of factory labor, which performs exclusively monotonous chores. Ultimately, the elimination of extremely uninteresting monotonous lesson tasks can be beneficial and serve as a source of leisure necessary for the all-round cultural development of a person. But it can also lead to the same low value and pernicious results in the field of culture, which for the most part have been obtained from radio and cinema.
Be that as it may, the transitional period for the introduction of these new means, especially if it occurs instantly, as can be expected in the event of a new war, will result in an immediate transitional period of a disastrous crisis. There is a lot of experience showing how industrialists feel about new industrial potential. All their propaganda boils down to the fact that the introduction of new technology should not be considered as the business of the government, but should be left to every entrepreneur who wants to invest in this technology. We also know that industrialists are hard to restrain when it comes to extracting from industry all the profits that can be extracted from it, in order to leave society to be content with crumbs.
Under these conditions, the industry will be filled with new mechanisms only to the extent that it is obvious that they will bring immediate profit, regardless of the future damage that they can cause. We will be witnessing a process along the same lines as the Atomic Energy Process, in which the use of atomic energy to build bombs has jeopardized the very urgent prospects for the future use of atomic energy to replace our oil and coal reserves, which centuries if not in decades, completely depleted. Note that the production of atomic bombs does not compete with energy-producing firms.
Similar Documents
Can artificial intelligence at this level of development of technology and technology surpass human intelligence. Can a person upon contact recognize artificial intelligence. The main opportunities for the practical application of artificial intelligence.
presentation, added 03/04/2013
Artificial intelligence is a scientific direction associated with machine modeling of human intellectual functions. Features of artificial intelligence Development of artificial intelligence, promising directions in its research and modeling.
abstract, added 11/18/2010
Agent-based approach to the study of artificial intelligence. Reasoning modeling, natural language processing, machine learning, robotics, speech recognition. Modern artificial intelligence. Performing a Turing test.
test, added 03/10/2015
The history of the development of artificial intelligence in foreign countries, in Russia and in the Republic of Kazakhstan. Development of a project for the effective implementation and adaptation of artificial intelligence in human society. Integration of the artificial into the natural.
scientific work, added 12/23/2014
The essence of the term "artificial intelligence"; history of its development. Science and technology of creating intelligent machines and computer programs. The challenge of using computers to understand human intelligence. Analysis, synthesis and understanding of texts.
thesis, added 06/17/2013
The phenomenon of thinking. Creation of artificial intelligence. Mechanical, electronic, cybernetic, neural approach. The advent of the perceptron. Artificial intelligence is an example of the integration of many scientific fields.
abstract, added 05/20/2003
Areas of human activity in which artificial intelligence can be applied. Solving the problems of artificial intelligence in computer science using the design of knowledge bases and expert systems. Automatic theorem proving.
term paper, added 08/29/2013
Historical overview of the development of work in the field of artificial intelligence. Creation of algorithmic and software for computers, which allows solving intellectual problems no worse than a person. From logic games to medical diagnostics.
abstract, added 10/26/2009
Studying the problem of artificial intelligence. The process of processing information in the human brain. Deciphering the brain codes of the phenomena of subjective reality. Natural intelligence as a fact that has a subjective reality with the principle of invariance.
abstract, added 04.12.2011
Components and architecture of an intelligent agent, its addition with learning tools. Various approaches to the creation of artificial intelligence, prospects for its development. Ethical and moral implications of the development of intelligent machines and programs.
