AI in decision making: whether the computer will be able to replace the person?

Main article: Artificial intelligence (AI, Artificial intelligence, AI)

Evolution of technologies

Decision making, from the point of view of the ordinary person, is quite often connected with investment of the AI system with the property of subjectivity going to a demonizing of the computer program. An opposite pole – adding to discharge of AI of the systems which are carrying out, in fact, calculation tasks. However, purely calculation and intellectual tasks are separated by rather unsteady edge. Let's tell, neuronets, by and large, are not classical manifestations of AI.

Really, many concepts "mixed up": some call traditional methods of machine learning, including use of neural networks, artificial intelligence. All these tasks are integrated, anyway, by one: for their solution the saved-up data are necessary, notices Alexey Vyskrebentsev, the head of the center of examination of solutions of Foresight company.

The border separating classical solutions of a business intelligence (BI, Business Intelligence), and the advanced applied systems of analytics Big Data is so indistinct.

The expert offers the following list of the IT means popular today helping the person with decision making:

• Regression analysis – forecasting on the basis of external factors: demand forecasting, forecasting of term of "survival" of the equipment, etc.
• Cluster analysis – separation of group of observations into the classes similar in the characteristics. It is often used for customer segmentation for the purpose of differentiation of sales strategy and service.
• Classification – reference of observation to a certain class on a basis not only numerical, but also categorial signs. Apply forecasting of a response to a marketing campaign, forecasting of outflow, forecasting of failure, product quality forecasting, etc.
• The analysis of associations – for traditional check analytics or the analysis of a consumer basket (what goods most often buy together). Proceeding from it, for example, diversity of goods on the hall for ensuring passability, the layout of goods, planning of marketing campaigns usually develop different strategies.
• Sequence "if And, then B": as a rule, it is used for the choice of the best solution or action and also formation of the offer for the client.
• The recommendatory systems which are used for promotion of goods, content.
• Optimization methods - use for production planning, business optimization, optimization marketing mix and also for solving of tasks, connected with the optimal choice of an operation mode, proceeding from a set of the limiting variables (the budget, logistics, raw structure, sales channels, etc.).

In terms of penetration depth of intellectual IT in decision making processes three options of replacement of human mind are shown by computer:

• Replace the person on routine transactions. This sphere is occupied by any digital assistants and means of robotization of business processes (RPA) unloading the person from routine cares and also automation systems of support services: program robots cope with the most part of questions, and only in uncommon situations there is a switching to the "living" employee.
• Replace the person where in real life he will not be able to work, for example, at dangerous productions, in the harmful environment or unavailable territories.
• Replace the person in those tasks with which the computer program copes better than the person. First of all, it is about collecting and analytics of large volumes of data.

According to specialists of Accenture, intelligent tools turn account managers into "superheroes":

Banks anyway very need clever and inventive employees who are ready to increase efficiency of the work by means of technologies. Knowledge of the industry, empathy and skills of negotiating will be always important, but they will amplify more often artificial intelligence technologies, allowing to save scarce time of personal managers, for example, at the expense of the scoring solutions which are built in CRM platforms, the personalized recommendations and hints, commercial analytics".

Achievements of today

Fast processing of large volumes of data with optimization

Agorafreight is a logistic startup which using specialists of "Reksoft" created the universal aggregator of logistics services of all means of transport, including sea, automobile and air transportation, the global scale which contains millions of rates for different transportations today. Tens of the companies worldwide directly enter the offers into the system, online calculations work for transportations between China, Vietnam, South Korea, the USA, India, the states of the European Union and Russia. The IT system automatically determines possible routes and the cost of transportation by different types of transport, the customer needs to select an optimal variant.

Transport – one of those fields of economy where algorithms run the show more and more. For example, we already implemented together with Siemens our solution on the basis of peripheral calculations in one transport global company which is engaged in shipping. Economy of fuel for container carriers made from 2% to 19% depending on run duration, is confident Roman Gots, the director of the department of Big Data and security of Atos in the Russian Federation.

"Smart" monitoring

The Russian brand of Cloudburst clothes together with Verisium company (A startup of business incubator of Kaspersky Lab) began to build in NFC- of a tag articles of clothing. The cloud solution includes the web portal for a brand and mobile application using which it is possible to implement the most different imaginations of the marketing specialist: from authentication of a brand and information on fabric and so forth before receiving personal discounts and bonuses of type of tickets for the closed parties and a premiere or watching the accompanying video, up to a virtual fitting room in a format of augmented reality on the smartphone screen.

"Smart" Cloudburst clothes: each product is supplied with a NFC tag

Analytics of Big Data

Practically each person faces practical work of analytical algorithms Big Data today. IT giant like Facebook, Google, Yandex accumulate data on users, behavior patterns, search queries, etc. for a number of years. Surprises nobody that the conversation on home cuisine in the presence of a smart column or the voice assistant comes to the end with emergence of the corresponding advertizing in an email client. Smart algorithms help to translate texts and, it is necessary to tell, cope with this task better and better.

The corporate sector which does not have access to such volumes of the saved-up data only looks for today the most effective approaches of use of Big Data. So, banks lay great hopes on OpenAPI and opportunities opening with connection of databases of partners.

However, as Nikita Blinov notes, the founder and SEO of Rubbles company, the practical significant movement in this direction will begin only when banks see real commercial perspective.

the Key source of revenues of banks – percent on the credits. A basic process which provides this source, - scoring therefore in scoring systems the artificial intelligence is implemented long ago. The business process following on importance - personalisation of cross-sellings, there smart IT solutions work too. Well, and all the rest – is much less important, in terms of revenues of bank therefore speak about it much, but really do not hurry to be engaged in such systems,

Besides, still not all tasks connected with technological support of analytics of Big Data are solved.

Today the set of various libraries for different programming languages which simplify development in the field of BigData is created. But it is only tools, and here as for data then not everything is so smooth. Often it is necessary to spend a lot of time to receive a suitable data set for the researches, notes Alexander Spiridonov, the head of laboratory of information and network security of Kriptonit company.

Many data sets are closed, others have no rather qualitative marking, the expert continues.

It is an essential restraining factor for development of Big Data in different areas as for a qualitative or innovation research often it is necessary to receive an own data set that takes a lot of time and resources. The marking of data is often carried out in the manual mode if high quality is required. At large volumes of data it creates vital issues for researchers. You should not forget also about hardware resources. Despite good optimization of tools, a task of Big Data require big computing powers which are available to not everyone, summarizes Alexander Spiridonov.

According to Nikita Blinov, data analysis – it is normal a story about optimization of the existing processes therefore the answer to a question is crucial for making decision on implementation of intellectual software: what to measure this improvement in? It is money? Or time of employees? Or something else? How many additional value for business it is possible to receive as a result?

Proceeding from such problem definition, the expert believes that modern solutions of the analysis of Big Data not bad give in to replication:

It is necessary to understand only within what business process it occurs. So, the companies can work in the different industries, but if business processes at them are similar, for example, it is about scheduling, then transfer of the application analytical solution will be rather simple.

In line with such approach the company created a software platform of Rubbles which gives the chance to deploy at the enterprise an intelligent analytical solution many times quicker, than to create it from scratch, even in those fields of activity in which the company did not execute projects yet.

Predictive analytics for industrial enterprises

The intelligent systems belonging to the class of predictive analytics received special popularity in the transport industry. Many commercial solutions which give the chance on the basis of data on last and current status of a transport object rather precisely to calculate its characteristics in the future are presented at the market.

"Such concerns as Mercedes and BMW use in the cars the built-in modems with the international SIM card long ago that on a regular basis to collect diagnostic data from controllers and nodes of the car, - Sergey Kulikovsky, the CEO of Polymatica company says.-The producer receives enormous base of indications and statistics of failures and can reveal patterns in indications which with big confidence figure can lead to failure. The car itself prompts to the driver when it is necessary to change oil or to replace brake shoes". Important, the expert notes that he does it not on the preset program of maintenance, and on predictions of constantly developing model. This model is developed by the producer of the car and continuously specified on the basis of growing to base of failures and indications.

Since recent time the prognostics actively is implemented in the railway industry where prevents crashes of the rolling stock and infrastructure facilities - methods of empirical modeling, statistical and parametrical analysis, machine learning, neural networks and some other technologies are for this purpose used. All these technologies allow to carry out intellectual data processing, arriving from the onboard measuring systems installed on objects tells Nina Kvaratskheliya, the leading business analyst of Center 2M company.

Actions for digitalization of the equipment – connection of digital sensors, interfaces and data collection systems – give the chance to apply automated systems of monitoring of technical condition of the equipment.

Despite the complexity, they work in the mode of "real time", signaling about danger or switching-off the equipment "upon - the expert emphasizes.

The Center 2M company designed and developed the intelligent system of predictive analytics intended for solving of tasks of monitoring and the forecast of technical condition of electric motors. A system successfully passed preliminary tests, having shown rather high accuracy (more than 85%) of detection of prenegative statuses of electric motors even on rather small amount of data. The high efficiency is shown also by other intelligent systems implemented in the railway industry.

The smart train at a profile exhibition in Shcherbinka

So, after the implementing solution "Smart Locomotive" in Lokotekh, the largest service the holding, failures on the line decreased by 32%, and time for diagnostics of the locomotive was reduced from 4 to 10 o'clock minutes.

Structure of the solution "Smart Locomotive" Clover Group

In Clover Group company which developed this intelligent system of diagnostics and the forecast of technical condition of the equipment of locomotives, told that for the period trial operation a system processed these 2 million (!) hours of operation of locomotives, more than 120 thousand incidents in operation of the locomotive equipment are automatically revealed.

Clever assistant to the professional analyst

The systems of artificial intelligence - incredibly powerful tool for the solution of analytical problems of a certain type, but it is effective in hands of the qualified specialist having not only theoretical knowledge, but also wide practical experience in use of this tool. For example, the error put at a design stage of neural network can lead to the considerable time delays in the solution of an assigned task caused by the long cycle of training preceding its detection tells Sergey Germanovich, the technical director of the company Polymatica.

In the field of the advanced tools focused on professional analysts - the approaches to their creation. Let's tell, at the global level tools of the analytical Palantir Technologies platform can serve as a sample. Activity of the company of the same name is surrounded with a mystery cloud: they it is known for developments, so-called, dual purpose, and different funds, including, the In-Q-Tel fund created in 1999 which performs in the USA mediatorial function between CIA of the USA and the scientific organizations are involved in its financing. However after a few years ago the Palanti tools became available to commercial clients, real details of this software became known.

In general, this software package is engaged in the fact that it very significantly facilitates really, monumental task on sifting of huge arrays with badly comparable data. Software of Palantir it is capable quickly, in the mode of parallel processing to comb all available databases and to reveal information fragments connected with each other. The program systematizes all this and offers analytics in the type convenient for viewing and drawing up advanced analytical queries.

It should be noted that in a basis of the Palantir platform the idea not of pure automation of pass of data arrays, and smart synthesis where the analysis methods relying on the approximated model of thinking of the person, and the powerful algorithmic engine capable to scan at the same time several databases at very thin level of granulation are integrated is put. Actually the engine is capable to take information from huge databases and to submit it so that the analyst could "cut layers" and browse results of sifting by almost infinite number of methods.

The expanded architecture of the analytical systems on the Palantir Technologies platform uses the object approach to the description of the processed data implemented based on the semantic mechanism of ontologies. At the same time dynamic ontologies can be adapted to new data domain. In a system it is implemented several search mechanisms using ontologies for simplification of requests and search of the words, close including phonetic mechanisms for effective search in speech data.

Sounds beautifully. But in practice for obtaining reasonable result the specialist should understand that under a cowl at such system and why it processed data in one way or another. So having handed such system to the inhabitant, you do not receive the standing result, and such system will seem to the specialist too narrow and nonflexible. But approach - to learn to integrate diverse data - it is quite perspective, comments Alexander Spiridonov.

Where in intellectual work will the machine replace the person?

For major problems the whole staff of experts who are engaged in solution development is employed. But also ready-made smart solutions do not do too without group of analysts who are trained on use of this tool, further exploit him, though plunge into subject not so deeply as solution designers in the first case.

It is possible to assume that the smart computer system can charge highly specialized tasks, and it will cope with them not worse than the person analyst? Yes it is. In banks, retail, industry and smart IT solutions which became these intellectual assistants to the specialist person work on transport today. It is about such solutions which belong to the class of the decision making support systems (DMSS) at the level of average management, Nikita Blinov notes and gives examples. The marketing specialist of bank which implements the promo-program in one click receives ready selection to the choice of clients with the necessary parameters. The manager of shop who develops the program of discounts for these or those goods receives a demand forecast on these goods.

Analytical software, of course, does not replace the marketing specialist, but saves to it and his colleagues a lot of time. And besides quality of the solution proposed by machine - at high expert level,

Actually it is that today the computer systems focused on decision support by the person quite successfully perform intellectual works which have routine character for the person.

Smart tools of the person analyst

Analytical solutions of the previous generation were able to make well calculations, but only the person made decisions on the basis of those data which provided it software. Systems created on the basis of artificial intelligence are capable to create some solution, the offer of the action plan.

With it the person is helped by the advanced tools of the professional analyst which combine different methods and models of data processing today.

For example, application of cascades of artificial neural networks (perseptron) for solving of tasks of multidimensional classification in banking sector, insurance business and medicine (when the number of input parameters is excessively high) yields the best result, than traditional mathematical methods. And use of neural networks of Hopfield yields excellent result in problems of search of anomalies and hidden patterns in large volumes of data.

But, as a rule, all current client tasks can be solved by traditional methods of numerical analytics", - notices Germanovich.

Solutions of tomorrow

If the model in the course of the work is able independently to select the necessary data, to interpret them and to use for self-training, it will be serious help to all specialists in data analysis and, perhaps, the first step by an era of strong AI. In the long term it is possible to assume capability to innovative thinking, transfer of experience from absolutely different areas and creation of new knowledge.

The basic moment in all this story – ability of the machine to create new knowledge.

Generation by the smart system of new useful data

In 2018 IBM Research together with Symrise, the company working at the perfumery market was announced by solution AI of Philyra which will develop new aromas for spirits. Of course, the beautiful case directly connected with problems of digital modeling of chemical technologies. In the similar direction very perspective segment of the market called by digital pharmacology moves.

In February of this year a team of scientists from Skoltech, MSU and IEFB of RAS completed the project of digital modeling of new medicine using supercomputer molecular modeling on technical resources of MSU. New short neuroactive peptide – substance which will help with treatment of a clinical depression by the unique technique developed in Russia was developed.

Scientists-physicians from the Russian company Insilico Medicine look for medicine for fight against aging. With development of new drugs they are helped by machine learning technologies. In this industry technologies of ten-year prescription are often applied today, Sergey Nikolenko, the director of science of Neuromation company, the IT partner of the medical company while using machine learning it is possible to generate different options complains and to quickly select the most suitable for further testing in laboratory.

Today it is known it is insignificant a small part of molecules. For search manually there will not be enough life of all our Universe even if we will attract all people living on the planet. Therefore for the solution of this very important task we need both the most powerful supercomputers, and the AI new methods and the new theory would not prevent,

Medical diagnostics

The companies "TechLAB" and "Netrika" in February completed in Leningrad Region implementation of a subsystem of "Onkopomoshch". Its main purpose - effective routing of onkopatsiyent and also control of terms, volumes and results of the medical aid rendered to them. In translation into normal Russian it means that the smart computer system which locates extensive base of clinical data and clinical records of patients with malignant new growths or suspicions on them will provide recommendations about their treatment taking into account an individual clinical picture, to provide succession of treatment and also to control observance of the ordered treatment route.

So looks digital onkopasport the patient

In oncology it is extremely important to begin treatment as soon as possible: the forecast of a disease and effectiveness of treatment depends on it, explains Igor Bashkov, the head of development of the Health care direction of Netrika Group.

Recognition of new threats of information security

The intensity of emotions at the front of cybersecurity increases: the number of cyber attacks from year to year promptly grows. According to Positive Technologies, in 2019 the company recorded more than one and a half thousand attacks that is 19% more, than last year.

Categories of the victims of cyber attacks in 2019 among legal entities. Source: Relevant cyber-threats: results of 2019, Positive Technologies, March, 2020.

The important feature of a present status of cybersecurity threats is in what the events happening to the companies and the organizations is possible to two strongly differing classes.

The clerk who shirked a training on cybersecurity opened the letter allegedly from the client of his firm and activated macroes in Word or Excel. A confidential information was laid out in the database, access to which was incorrectly configured. Forgot to organize backup, and organization activity is paralyzed by the encoder. Not always it is even possible to call such "attacks" the attacks because they happen in a random way, and mailing of malware - is automated long ago, tells Georgy Lagoda, the deputy CEO of Programmny Produkt Group.

The second – much more dangerous. According to data of the research Positive Technologies, the target (APT) attacks in 2019 was significantly more, than mass - their share made 60%. At the same time malefactors began to use actively methods of machine learning and artificial intelligence that gives them the chance to perform target APT with rather low cost value.

Often malefactors use not declared possibilities of software, a tab, open office interfaces, vulnerabilities of zero day, apply non-standard logic of operation of devices and programs when, for example, the source of suspicious traffic can mask under stationary phones using the corresponding ranges of MAC-and the IP addresses of the company of the developer of these phones. It is extremely difficult to be protected from such attacks, and without skills of experts, timely reaction, even intuition and an intuition, not to manage here, explains Alexander Spiridonov from Kriptonit company.

It is obvious that in these conditions capability of an analytical system to generate fruitful hypotheses on the basis of the available data should find broad application in the field of the information security (IS). Participants of the cybersecurity market say that became in an essential part to create the agenda of the Russian market of cybersecurity those players of the domestic market who live and act in a paradigm 2.0, i.e. are guided not only by reflection of the attacks, but also by pro-active identification of threats.

Example of the advanced solution such – the SIEM systems intended for identification of suspicious incidents of cybersecurity. Their task - to reveal from hundreds of millions of events which are taking place every second, suspicious patterns, and to define what combinations require immediate intervention of the person.

At the same time these combinations can have absolutely unevident nature, to be stretched in time, to seem not connected among themselves, at least, before successful cyber attack. Only the machine, just can cope with the similar analysis because it is about fast processing of enormous amount of data. From that how advanced the used algorithms are, the efficiency of such systems depends, tells Georgy Lagoda.

For example, the intelligent system of monitoring of events of security in digital infrastructure of RuSIEM allows to create the uniform center of collection of information about the events in network, collecting data from the different systems and devices: network equipment, servers, workstations, information security tools and to business applications and services.

Structure of a SIEM system

SIEM systems can generalize data. They work by in advance created rules, relying on numerous data sources. But if the malefactor manages to bypass these rules, consider that the attack took place successfully. Why? It is the second question. Perhaps, there is no necessary rule of operation, or the insufficient quantity of data sources is connected to a system, or it somewhere is accidental it was disconnected, put out of action, timely it is not updated.

The excessive delay of services of the company can even cause the successful attack: not smoothly running communication between departments of security, network sales managers, administrators. Often in case of an emergency situation, the expert, on the organization of a conference notices, waiting of profile specialists can leave many times more time, than on response to security incident.

SIEM systems are important and protect from the ordinary attacks, but to hope that they equally well will protect from APT, would be naive. Advanced solutions do not exist because of specifics of the processed data, infrastructure and the SIEM settings in the different organizations now, speaks Alexander Spiridonov.

As an example of very advanced cybersecurity solutions it is possible to call, for example, the product IoT Inspector developed by SEC Consult Services company. It is designed to reveal vulnerable devices of Internet of Things in digital infrastructure of the companies.

Owing to very low level of development of shell programs, IoT-devices often are one of the weakest points, and even direct threat. Besides, if the network is big, then to find in it some long ago forgotten router which still works, can be an uncommon task, explains Georgy Lagoda.

IoT Inspector is created just for the solution of similar problems.

One more very promoted product - Cybertrap, the advanced solution for deception of cyber-malefactors: they are enticed on controlled Wednesday and there monitor all their actions, despite of their most sophisticated attempts it is reserved.

Management of complex systems: smart city

Since 2019 within the national projects "Digital Economy" and "Housing and Urban Environment" in our country the Smart City program is implemented. It is already possible to see the first results of the first pilot projects. In particular, unusual versions of usual elements of the urban environment appeared.

The main directions use of ICT for city development. Source: research "Indicators of the Smart Cities of NIITS 2017"

For example, in the town of Satka in Chelyabinsk region which became one of pilot regions of "The smart city" on one of streets roadbed status sensors which allow to estimate quality of a covering in real time are installed. Sensors analyze, for example, the level of moisture and ice and also content of deicing mixes. The collected data allow to control work of road services and to timely make decisions on repair.

The Schwabe holding developed the smart traffic light equipped with the laser block which projects in air a visual protecting signal for drivers. From outside visually it looks as ranks of parallel lines which from distance of 100 - 200 meters warn about location of road "zebra". The signal is well visible in the dark and fog.

In some cities hi-tech trash cans appeared: they are supplied with the sensors displaying filling level. Garbage trucks will save fuel, approaching only the filled tanks.

Just in recent days the functionality of a capital system video surveillances which develops within the project "Safe city Moscow" extended. The chief of Head department MINISTRY OF INTERNAL AFFAIRS on To Moscow Oleg Baranov says that a system successfully helps employees MINISTRY OF INTERNAL AFFAIRS to detect the citizens breaking the quarantine mode of century To Moscow.

The calls facing the industry. What disturbs smart programs to undertake more intellectual work?

Restrictions of the systems of predictive production analytics

Modern technologies of industrial Internet of Things (IIoT, Industrial Internet of Things) give the chance to connect a large number of transport objects to monitoring systems. It promotes the best training of neural networks that increases depth and forecast accuracy. The systems of predictive analytics integrated with Internet of Things increase efficiency of the equipment, reduce operating costs and prevent accidents on transport.

However at implementation of such systems of the company face rather complex technology problems. First of all, weak equipment of technical objects digital metering equipment concerns them. Also problems at data transmission in the centers of real-time processing are observed.

Besides, as marks out Ning Kwaratskheliya, a basis of all intelligent system of monitoring is the mathematical core which contains "digital images" of the operational equipment during the work in the different modes.

Creation of such digital portraits in itself is a difficult scientific task and requires considerable time for modeling of technology processes and accumulation of statistical data, emphasizes Kvaratskheliya.

However, benefits from use of intelligent systems of monitoring of technical condition of the equipment block difficulties of their creation, the expert believes, that systems draw the attention of experts to "suspicious" deviations in advance, allow to avoid accident and the related serious material damage".

Calls of the world of information security

The call of this moment is that, despite existence in our market of the cybersecurity systems conforming to the highest modern requirements deficit of solutions is all the same felt.

Matter is not that existing solutions, products and services are not capable to satisfy the growing requirements. Just there are no universal solutions "from a box" which at once will cover all current and future demands. What in the cybersecurity environment it is accepted to call "a landscape of threats", i.e. the general situation in the field of cyber security, - a thing extremely changeable. Perhaps, it "predictably unpredictable" Wednesday, explains Georgy Lagoda.

Result of such permanent unpredictability - a constant "a race of arms" of cybercriminals and cyberdefenders which will never end.

Behind examples especially far it is not necessary to go - all remember a story with exploits, "implants" and we designate the malware, efforts of unknown Shadow Brokers developed by Equation Group, and "flowed away". The most dangerous of these exploits were shown afterwards in global epidemic of the encoder of WannaCry. It is unlikely someone can foresee when the next "black swan" of similar sense and what effects it will have comes up, tells Lagoda.

But one is obvious: "light forces" should have not less intelligent and powerful tools, than their opponents.

Need for emergence of new developments is not connected directly with quality and the level of development of existing. Just calls change. Nobody cancels a basic antivirus, for example, but some cannot cost them now, describes the current situation Georgy Lagoda.

If to speak about the future, then it would be desirable to see gradual reduction of share of participation of experts in maintenance of appropriate level of security and investigation of incidents.

Barriers to implementation of the large-scale AI systems in medicine

This combination of two factors: world motion speed around us, and the direction of this movement causing melancholy towards complete overregulation.

It brings an example of negative overregulation: order MZ Russian Federation No. 203-n "About the statement of quality criterions of medical care". Physicians will not violate the order of the regulator and, therefore, will appoint to the patient of a research, specified in the instruction even if the senselessness of this or that appointment in a specific case is obvious.

Besides, Mihail Kauffman emphasizes, these criteria are based on the principles of evidential medicine and statistics, but today to them there can not correspond some results of researches received, for example, using sensing technologies of images.

If the AI program claims that in the medical picture with the probability of 82.646% there is no new growth, but who will bear responsibility that it there nevertheless was? Even if developers trained a neuronet at a data set of one million pictures in size whether we can be sure that it is enough? - asks a question Mihail Kauffman.

While in the law there are no obligations to hand over all pictures in work of AI (I heard that so it works in some clinics abroad), it will be work not for the benefit of our health, science and the industry, and for export more likely, – the expert is sure.

Similar story - with the recommendations of medicines, for example, on the basis of indicators of pressure or a warm rhythm.

If the result received using AI does not lay down in an outline of evidential medicine, then the last wins even if AI, really, will detect something uncommon in the studied data, comments Kauffman.

According to him, the main problem which is slowing down implementation of advanced methods AI in medical diagnostics is that "the interests of the patient left consideration long ago, the maximum accomplishment of standards, without thinking about conciseness and relevance of their provisions interests all:

If you thought up a wonderful product which something good does to the patient, but standards from it will not be executed more, than without it whether chief physicians of hospitals and officials of health care will your miracle be interested? Will not be. And doctors of privates or patients will not be asked, not to them to solve what to spend the state resources for.

Calls of "smart city"

In this sense the modular architecture in many meanings is more reliable and more effective than complete. The scalable solutions allowing to be integrated with others, eventually, can provide the best protection, than "boxed" products with the fixed feature set, regardless of that, this set is how broad, speaks Georgy Lagoda.

At the same time, the expert is sure, absolutely universal solutions which will remain like those for many years do not exist.

Denis Serechenko, the director of digital transformation of Huawei Enterprise in the region Eurasia, also believes that each project of the smart city is unique, and its experience practically does not give in to replication.

I Think, today in general it is impossible to say that this or that city reached a status "smart". All of them: Barcelona, Amsterdam, New York, Shenzhen, Beijing, Moscow, etc., - have certain elements of the of "city mind". However business cases which give the chance to replicate the best experience, no.

For today it so. For video analytics and application of smart algorithms or application of 5G networks for video analytics some basic level is necessary. All this is possible when in the city there is already a video surveillance at least in some type. If we speak about Russia, then so far providing the cities with video surveillance leaves much to be desired, -agrees Roman Gots.

Georgy Lagoda considers that in present conditions the choice of a complex of the modern IT solutions implemented based on the latest technologies, such as BigData and IoT and, such which in the best way are suitable for solving of tasks of the specific region, for example, smart parkings, monitoring of a passenger traffic, the booking systems, etc. will be optimal strategy.

Denis Serechenko, in turn, is sure that creation of the smart city is based on two bases: volume of public finance and necessary ICT infrastructure. The financing source factor in this case, really, is extraordinary important: IT solution is difficult, intelligently capacious, so, expensive, but co-investors from the environment of private business do not queue for such projects – the social sphere is not represented too profitable case.

The main barriers or problems interfering implementation of the AI systems today. Source: "Market in Russia", IDC, 2019.

Summing up all revealed barriers and restrictions constraining today wider use of the AI systems for decision making in complex practical projects experts selected three key factors:

The main driving force are the investors ready not only to sponsor development, but also to create cross-disciplinary groups, to provide a resource of "technicians" and "subject teachers" in startups, heat-sink such ideas and practices in the project pools, emphasizes Mihail Kauffman.

2. Legal. Legal regulation of AI technologies regarding a possibility of decision-making by machine without participation of the person. The discussion on a hot topic of responsibility for the made decision is so far very far from end.
3. Organizational. Development of technologies of decision making using intelligent systems and their mass practical implementations asks preparation of the corresponding personnel: for development, operation and development of the smart systems.

You See Also

• Artificial intelligence (market of Russia)
• Artificial intelligence (world market)

• Big Data
• Big Data in Russia
• Big Data world market
• Data Mining