JSC DOM.RF tested YandexGPT from Yandex for several months
Posted: Wed Jan 22, 2025 5:45 am
On January 31, at a webinar of the FinTech Association on artificial intelligence, Maria Kulabukhova, head of the fintech department at Yandex.Cloud LLC (Yandex Cloud), reported that companies from the fintech industry can use generative language models for customer support in chatbots, for investment recommendations, news analysis, and for internal information search in knowledge bases.
For business in general, according cyprus whatsapp number database to Maria Kulabukhova, LLM (large language models) can be useful in sales (for answering frequently asked questions, proofreading and editing letters, filling out CRM); in marketing (for writing articles, blog posts, creating advertising creatives); and also for customer support departments and HR departments for writing job postings.
on the platform of the FinTech Association. Grigory Gryaznov, Deputy Director and Head of the Data and New Technologies Department of DOM.RF, shared the results of this pilot: the first task for the generative AI was to monitor news channels and find news that contained information about the risks for DOM.RF counterparties - to do this, the neural network had to sort the news into positive, neutral and negative; as a result, the AI was able to identify the essence of the news and detect negative ones with 90% accuracy.
The second scenario for using generative AI from Yandex in DOM.RF, according to Grigory Gryaznov, was training a banking chatbot: for this, the neural network had to formulate possible questions from users that could be applied to existing answers, that is, to become a kind of intermediary between a person and a bot; as a result, the AI generated more than 5,000 training phrases for 350 bot scenarios.
As Grigory Gryaznov noted, the neural network was only able to be used in those tasks for which external data is sufficient, so the experiment resulted in analyzing the information field and coming up with potential questions for the chat; but the bank is not yet ready to allow a third-party neural network to work with internal sensitive data.
Read also
Russian companies are in no hurry to implement generative AI
Russian companies will need at least two years to implement generative artificial intelligence in business processes. IT market experts agree that the procedure will take about five years.
Alexey Sidoryuk, Advisor to the CEO for Artificial Intelligence at the FinTech Association, said that in 2024, the association will continue to pilot the use of generative AI in banks. Six organizations are participating in the pilot: SberDevices (Giga Chat models), Yandex Cloud (Yandex GPT models), Alfa-Bank, DOM.RF Bank, Gazprombank, and VTB Bank.
According to the FinTech Association, piloting will help achieve the following results: domestic vendors of generative neural networks will be able to determine, based on feedback from fintech market players, what their future products should look like for installation in the customer's digital circuit (on premise) and their optimal financial model; developers will also select areas for improving neural models and determine in which areas in the banking, investment and insurance sectors it is best to apply GenAI.
In August 2023, Alexey Sidoryuk, who was then the director of the Digital Transformation of Industries department at the Digital Economy ANO, noted that, despite Russia's success in the global race for generative neural networks, domestic AI solutions, including language models, lack practical application: companies do not understand how to integrate them into business processes, and the creators - Yandex and Sber - have not yet found the optimal monetization option.
For business in general, according cyprus whatsapp number database to Maria Kulabukhova, LLM (large language models) can be useful in sales (for answering frequently asked questions, proofreading and editing letters, filling out CRM); in marketing (for writing articles, blog posts, creating advertising creatives); and also for customer support departments and HR departments for writing job postings.
on the platform of the FinTech Association. Grigory Gryaznov, Deputy Director and Head of the Data and New Technologies Department of DOM.RF, shared the results of this pilot: the first task for the generative AI was to monitor news channels and find news that contained information about the risks for DOM.RF counterparties - to do this, the neural network had to sort the news into positive, neutral and negative; as a result, the AI was able to identify the essence of the news and detect negative ones with 90% accuracy.
The second scenario for using generative AI from Yandex in DOM.RF, according to Grigory Gryaznov, was training a banking chatbot: for this, the neural network had to formulate possible questions from users that could be applied to existing answers, that is, to become a kind of intermediary between a person and a bot; as a result, the AI generated more than 5,000 training phrases for 350 bot scenarios.
As Grigory Gryaznov noted, the neural network was only able to be used in those tasks for which external data is sufficient, so the experiment resulted in analyzing the information field and coming up with potential questions for the chat; but the bank is not yet ready to allow a third-party neural network to work with internal sensitive data.
Read also
Russian companies are in no hurry to implement generative AI
Russian companies will need at least two years to implement generative artificial intelligence in business processes. IT market experts agree that the procedure will take about five years.
Alexey Sidoryuk, Advisor to the CEO for Artificial Intelligence at the FinTech Association, said that in 2024, the association will continue to pilot the use of generative AI in banks. Six organizations are participating in the pilot: SberDevices (Giga Chat models), Yandex Cloud (Yandex GPT models), Alfa-Bank, DOM.RF Bank, Gazprombank, and VTB Bank.
According to the FinTech Association, piloting will help achieve the following results: domestic vendors of generative neural networks will be able to determine, based on feedback from fintech market players, what their future products should look like for installation in the customer's digital circuit (on premise) and their optimal financial model; developers will also select areas for improving neural models and determine in which areas in the banking, investment and insurance sectors it is best to apply GenAI.
In August 2023, Alexey Sidoryuk, who was then the director of the Digital Transformation of Industries department at the Digital Economy ANO, noted that, despite Russia's success in the global race for generative neural networks, domestic AI solutions, including language models, lack practical application: companies do not understand how to integrate them into business processes, and the creators - Yandex and Sber - have not yet found the optimal monetization option.