En
Register Log in

List of publications on a keyword: «large language models»

  • Hardware architectures for local learning of large language models: an overview of scalable solutions

    Conference Paper
    Current issues of integrating science and practice in the context of transforming socio-economic systems
    Authors:
    Копыркин Николай Андреевич, Anna D. Sudakova
    Work direction:
    Развитие информационных технологий в условиях трансформации социально-экономических систем
    Abstract:
    The article discusses which hardware accelerators are advisable to use if you need to train large language models on your own servers, rather than in the cloud. The main types of accelerators are analyzed — GPU, ASIC, FPGA and others. The cost of ownership of different solutions for local use when installing 256 accelerators is compared separately. As a result, it is concluded that the choice of equipment should take into account not only the speed of calculations, but also the costs of electricity, cooling and software compatibility.
    Keywords:
    large language models, hardware accelerators, scalable architectures, total cost of ownership, local deployment
    Go 

  • AI Assistants as Tutors: Ethical Boundaries for Using Neural Networks to Personalize Programming Education

    Conference Paper
    Education, innovation, research as a resource for community development
    Authors:
    Inna G. Iglinskaia, Anastasiia V. Petrakova
    Work direction:
    Использование нестандартных решений для повышения качества, доступности и эффективности образования
    Abstract:
    The article discusses the current problem of integrating artificial intelligence (AI), in particular, large language models (LLM), into the educational process of training specialists in the specialty 02/09/07 "Information Systems and Programming". The purpose of the study is to analyze the potential of AI assistants as personalized tutors to develop programming skills and identify the associated ethical risks and pedagogical limitations. Based on the analysis of modern literature and the author's practical experience, key dilemmas have been identified: the reliability of the generated code and explanations, the risk of intellectual dependence, issues of academic integrity and digital inequality. The article offers a model for the balanced use of AI assistants based on the principles of pedagogical support, the development of critical thinking and the formation of digital hygiene among students. It is concluded that AI is a powerful tool for improving the accessibility and effectivenes
    Keywords:
    pedagogical design, tutoring, ChatGPT, programming, AI ethics, academic integrity, artificial intelligence in education, learning personalization, large language models (LLM), and professional education
    Go 

  • The principle of operation of recommender systems and their impact on people using YouTube as an example

    Conference Paper
    Relevant issues of management, economics and economic security
    Authors:
    Mark I. Tagirov, Aleksandr A. Ushakov, Olesia V. Tereshchenko
    Work direction:
    Менеджмент и его разновидности, диверсификация, маркетинг, ценообразование и логистика
    Abstract:
    This article analyzes YouTube's recommendation system and explores its architecture, which is based on a two-stage machine learning model, taking into account the implementation of self-learning methods, and its evolution in 2025. The article focuses on identifying and systematizing the risks associated with content personalization, such as the formation of "information bubbles," the impact on user behavior, and the promotion of radical content. Recommendations for mitigating the identified negative effects on users and developers are presented, based on current scientific publications and statistics.
    Keywords:
    artificial intelligence, machine learning, personalization, algorithms, addiction, YouTube, recommender systems, information bubble, large language models
    Go