Artificial Intelligence Journey (AIJ), one of the world’s key AI conferences, took place on Nov 23-24 in Moscow.

The two-day event is annually organized by Sber - Russia’s largest bank and a leading global financial institution with strong dedication to IT.

For the seventh time AIJ united leading AI researchers and developers, adopters of AI-powered solutions, and enthusiasts of this key life-changing technology of the 21st century affecting everyone. In just two days, the conference garnered over 50 million views from 180+ countries, once again emerging as the world’s biggest scientific event in AI.

The participants of AIJ heard presentations by leading world-class scientists, researchers and experts from top companies in the field of AI/ML application. This year saw the launch of a new track called ‘AI Journey Science’, which focused on the development of science and interaction between the scientific community, government and business.

AI Journey has also hosted a panel discussion entitled “Future of AI in BRICS and Shanghai Cooperation Organisation”, welcoming speakers from the China, South Africa, India, Brazil, and Uzbekistan. Some of these experts had worked on the UNESCO Recommendation on the Ethics of Artificial Intelligence. The panelists shared their views on the next steps that should be taken to outline the general approach to AI management in BRICS and SCO nations.

Keynote discussion saw highest level guests

The keynote discussion “AI technology for economic growth” took place on the final day of the event and was attended by President of Russia Vladimir Putin and moderated by CEO and Chairman of the Sberbank Executive Board Herman Gref.

Herman Gref believes that this year’s key trend is customization or a very simple use of industrial artificial intelligence. AI has grown to become a technology available even to small companies. In the last two years, Russia has made serious progress, both across its regions and at the level of its government.

According to Vladimir Putin, Russian AI solutions in many areas hold leading positions on a global scale. For instance, Russia is in the Top 10 countries by level of public services digitalization. The value of AI breakthroughs is colossal: their outcomes enable the country’s sovereignty, security, autonomy, ability to achieve industrial, social, and economic development goals. The Russian President thanked Sber for assuming the responsibility of coordinating activities in this domain.

The foundation for an accelerated AI development had been laid, Vladimir Putin observed. However, now it was imperative to ensure mass AI adoption and data-driven decision-making. The Russian President instructed to amend relevant regulations to lift existing barriers and maximize the speed of development, testing, and commissioning of high-potential AI technology in different economic and social spheres.

New solutions presented at AIJ

Every year groundbreaking state-of-the-art solutions presented at AIJ. Two most important presentations of this year included:

Kandinsky 2.0

Sber has presented Kandinsky 2.0 – Russia’s first multilingual diffusion model that generates images based on text descriptions with two billion metrics at the AI Journey.

This neural network was developed and trained by the Sber AI research team supported by Artificial Intelligence Research Institute (AIRI) based on a consolidated Sber AI and SberDevices dataset of one billion text-image pairs.

Kandinsky 2.0 is enhanced with the increasingly popular diffusion approach, as models of this type of architecture, unlike transformers, yield good results in almost all multimedia content (images, videos, 3D, and audio) text-based generation tasks.

The model can process requests in 101 languages with equal speed and quality. These include the widely used Russian and English, and rarer languages, e.g., Mongolian, alike. It will understand the task, even if the request contains words from different languages.

MedBench

Sber scientists have presented at AIJ the first open Russian-language platform for solving medical tasks with AI. Given the specifics of the subject-matter area and the scarcity of relevant data, this is a key project with potential to benefit the Russian-speaking research community.

Three automated natural language processing tasks are already available at MedBench: logic and knowledge, Q&A modelling, and understanding the meaning of medical texts. For each tasks, Sber and the Siberian State Medical University have prepared assessors’ answers and performed solution quality assessments. The platform maintains a public leaderboard with ranked best solutions.