During the AI Journey session at GITEX Global 2024, the biggest tech and startup show on the planet, in Dubai, speakers from Sber, the AIRI Institute, and Speech Technology Center talked about AI technology developments in Russia and abroad.
In their presentations, the experts discussed how multimodality can improve the efficiency of large language models (LLMs), what steps are needed to develop a sovereign language model, and why it is important for young people to develop critical thinking when working with AI.
Semyon Budyonny, Candidate of Physical and Mathematical Sciences, head of Advanced AI Technologies Development at Sberbank and scientific advisor at the AIRI Institute, reported on Sber’s flagship Generative Artificial Intelligence (GenAI) solutions, the GigaChat neural network model, and the Kandinsky image and video generation neural network. The speaker also focused on Artificial General Intelligence (AGI), which is the next potential key milestone of machine intelligence development after GenAI with the intermediate milestone of Interactive AI, and outlined the characteristics of AGI development and its stages of technological maturity according to its cognitive abilities.
Dmitry Bugaichenko, Candidate of Physical and Mathematical Sciences, head of Artificial Intelligence Development in B2C at Sberbank, highlighted the issue of synergy between LLM and classical approaches to the development of AI solutions. Based on the examples of industry leaders and Sberbank, the speaker suggested considering the opportunities that this opens up, including new scenarios of interaction between recommendation systems and humans.
Diana Zlobina, head of development for Sberbank’s GigaChat service, called for innovation-driven solutions to be used to narrow the technological gap rather than widen it. She emphasized that AI tools can change the world for the better if used correctly. Sber is striving to implement technologies and foster best practices, including AI ethics, to set an example for Russian and international businesses.
Ivan Oseledets, Doctor of Physical and Mathematical Sciences, Professor of the Russian Academy of Sciences, Director General of the AIRI Institute, Professor at Skoltech, emphasized that the education system should embrace the new technological revolution and adapt to work with AI tools. By using AI services in the classroom–ChatGPT, Kandinsky, GigaChat, and similar models–teachers help students master new technologies. When designing educational programs, emphasis should be placed on developing students’ critical thinking skills so that they do not shift responsibility to AI assistants but rather learn how to use the new tools effectively.
Andrey Kuznetsov, Candidate of Technical Sciences, director of FusionBrain Lab at the AIRI Institute, said that language models have already become the standard for effective solutions for science and business, but the world around us is not limited to language. Multimodality can bring language models to a new level. Architectures based on it allow for solving a wide range of tasks familiar to every person at the intersection of text, images, audio, video, and other types of data. The speaker focused on current trends and future demanded tasks for multimodal models research.
Eugene Zubakov, Chief Commercial Officer at Speech Technology Center, noted that LLMs can have a significant impact on the global economy. A topical issue for most foreign countries is the creation of a sovereign large language model. For this purpose, it is vital to take into account all local norms and standards, identify needs, choose an architecture, collect data, process it, and, as a result, train the model. Evaluation, testing, and support of the artificial intelligence are equally important steps, as this is the only way to make sure that the model is effective and meets the requirements.
AI Journey is one of the largest international conferences on artificial intelligence, bringing together leading experts, scientists, and business representatives from around the world to discuss the most trending topics in the field. This year’s main event will be held in Moscow for the ninth time. Speakers will address current issues that will determine the next steps in AI development: AGI, LLM, CV, quantum technologies, GenAI, AI in ESG, and much more. This year, the AIJ track will be included in the program of leading international industry-specific conferences in Uzbekistan, India, and the UAE.
Press Release received on Mail
