Big data means big questions

First, it’s worth understanding what is hidden behind a trendy phrase ‘big data’. In a professional environment, this is the name for information arrays, the daily replenishment of which can exceed 150 gigabytes. Therefore, classical databases and conventional analysis tools are powerless here. It is noteworthy that until 2011, such arrays were studied only for scientific research and the preparation of statistics. However, by the beginning of 2012, urgent systematisation of the increased volumes of information was required. 

So, first of all, they are needed for practical use. For example, analysts at large marketplaces study purchase history, seasonality, and even the behavioural characteristics of users to determine to whom and which product to show. In public administration, analysis of data sets helps make informed decisions in healthcare, economics, and law enforcement. Due to data analytics, manufacturing industry is getting rid of the opacity of processes and introducing so-called predictive production, allowing factories to foretell the demand and save resources. Even professional sports have not remained uninvolved, football clubs, for example, use big data to select promising players. 

Despite the potential of big data, it also has a downside that creates serious obstacles for specialists. The first and, perhaps, main problem lies in the very nature of information. Its arrays are extremely heterogeneous, which greatly complicates their processing for any statistical conclusions. In practice, this leads to a sad pattern: the more parameters are taken into account in forecasts, the larger the snowball of errors and inaccuracies accumulated during the analysis becomes.

The issue of cybersecurity is hardly less urgent. Any processes related to the storage and processing of Big Data increase the vulnerability of systems. The more information is stored in one place, the more attractive such a database becomes for attackers, which means the risk of cyber attacks and large-scale leaks of confidential data increases. 

This issue is quite relevant for international discussion today, rector of the Belarusian State University of Informatics and Radio Electronics Vadim Bohush highlighted.

«The dynamics of data production today are described by an exponential relationship,» he informed. «That means, those volumes that ten years ago could still be somehow processed using human potential are simply impossible to process today without the use of specialised algorithms, software and equipment.»

There were also raised the questions of the use of big data technologies in marketing and e-commerce at the conference, and the use of these tools to optimise IT solutions, business and production processes. In addition, modern technologies and analytics tools were discussed, as well as the specifics of implementing big data in such critical areas as medicine and education.

Exoskeletons and autopilots

According to statistics from the International Labor Organisation, there are about 317 million workplace accidents occur annually in the world, and traditional methods of preventing them like safety rules, checklists and scheduled equipment inspections, are not always effective. At the same time, artificial intelligence trained on big data is capable of analysing the condition of equipment at a speed inaccessible to humans, noticing the slightest malfunctions long before the critical point.

In practice, this looks like a transition from scheduled preventive maintenance to maintenance based on actual condition. A striking example is the experience of the Minsk Automobile Plant.

«Our young specialists have created software products that allow us to systematise data received from a vehicle,» director of the scientific and technical centre and chief designer of the Minsk Automobile Plant Andrej Savchits revealed. «In the case when the moment of malfunction has not even occurred yet, but operating conditions are approaching critical points, a warning about the need for maintenance appears.

There are a number of other options for using the power of artificial intelligence in production. One of them is smart devices. Exoskeletons with AI, for example, will be able to reduce the load on the spine, and biometric bracelets will be able to monitor the heart rate and anxiety level of workers. The use of AI in the autopilot system seems very promising as well.

The creation of a truly intelligent and maximally safe hardware and software complex that guides a vehicle along a certain specified trajectory largely depends on how well the car navigates the road and recognises road signs and traffic lights. Students from BSUIR took on the task of solving this recognition problem: in collaboration with MAZ automobile plant, they managed to create a simple neural network model capable of running on the power of the car itself in real time.

According to Andrej Savchits, tests have shown that the system consistently detects traffic lights in photographs with difficult weather conditions and even with the strong flares. At the same time, the data processing speed remains high, and the localisation accuracy is good enough for safe manoeuvring. It is true that according to traffic rules, only a person can drive a vehicle in Belarus, which means that it is not yet possible to use the autopilot system in our country. The technology is nonetheless promising and deserves further development.

AI helping doctors

In Belarus, research in the field of artificial intelligence has been ongoing for a long time, and the key role here belongs to the Joint Institute of Informatics Problems and the National Academy of Sciences as a whole. For example, with the help of AI, Belarusian scientists are developing new medicines, identifying molecular compounds necessary for producing fertilisers or promising and innovative materials.

Head of the laboratory for analysis of biomedical images of the Joint Institute of Informatics Problems of the National Academy of Sciences of Belarus Eduard Snezhko told how AI can help doctors. According to him, neural networks are already quite good at analysing medical images, clinical and laboratory data. As a result, diagnostics becomes more unified, independent of the human factor, and routine tasks of doctors are automated.

However, the problems have not yet been solved. The most serious drawback of neural networks is the potential instability of their work, due to the very principle of AI operation. Thus, changing just a few pixels in an image, for example, a computed tomography or X-ray, can cause a well-trained and working neural network to make a mistake in a diagnosis.

Accordingly, if a neural network, even the most intelligent one, can be deceived, its testimony should initially be treated with special caution and attention. Not to mention the possible falsification of medical data for legal proceedings using AI, and the introduction of false images into hospital information systems by cyber attackers. According to the head of the laboratory for analysis of biomedical images of the Joint Institute of Informatics Problems of the National Academy of Sciences of Belarus, domestic specialists are already working on solving these and other problems.

On legal grounds

It turns out that the widespread use and implementation of artificial intelligence also has disadvantages. Of these, the danger is posed not only by disinformation and the illegal use of personal data, but also by manipulation of public opinion, possible monopolisation of the actively developing neural network market, and more.

Director General of the Joint Institute of Informatics Problems of the National Academy of Sciences of Belarus Sergej Kruglikov commented how Belarus is searching for a balance between innovation and security. He specified that everything related to the creation, implementation and everyday use of artificial intelligence requires clear and understandable rules enshrined in legislation.

Belarus does not yet have a separate law that would comprehensively regulate this area. Still, this does not mean that it is some shadow sphere; on the contrary, a number of strategic documents have already directly stated: the use of AI in various industries meets national interests and is among the priorities of the state’s innovative development. The terms from the world of artificial intelligence are found more and more often in the industry acts, for example, regarding education or electric transport.

«The world does not stand still, and our country is no exception,» Sergej Kruglikov emphasised. «Artificial intelligence issues are receiving the most serious attention today.»

This was confirmed by the model law «On Artificial Intelligence Technologies», adopted on April 18, 2025, by the Interparliamentary Assembly of the CIS Member States. Developed by Belarusian scientists, the document describes the possibilities of government regulation, principles of risk management, rights and responsibilities of developers and users of AI. Special attention was paid to the processing, storage and protection of data. The model law was sent to all Commonwealth countries and recommended for use in national legislation.

The decision to develop a corresponding law in our country has not yet been made, but the movement in this direction is obvious, according to Sergej Kruglikov. A working group has been formed at the National Academy of Sciences of Belarus to create a draft legislation on artificial intelligence; it included 45 people from 30 organisations. Experts have already prepared two documents: a draft glossary of basic AI terms and a draft concept for the future law «On Artificial Intelligence».

Amina Nazarova

Photos from open sources