‘Gigachat’ Russia’s Sberbank launches Al to compete with ChatGPT
Sberbank has developed a technology named GigaChat to compete with ChatGPT in the artificial intelligence chatbot race. Initially available in an invite-only testing mode, GigaChat stands out with its ability to communicate more intelligently in Russian than other foreign neural networks.
Russia’s ‘Gigachat’: Key Points
- The release of Microsoft-backed OpenAI’s ChatGPT last year has spurred the technology industry to incorporate AI further into users’ hands to reshape how people work and win business.
- Sberbank, Russia’s leading bank, has invested heavily in technology to reduce the country’s dependence on imports, which has become critically important due to export cuts and sanctions over Moscow’s actions in Ukraine by Western nations.
Australian High Commission in India Announces Government Grant for Project in Kargil
What is Gigachat?
- The architecture of GigaChat is based on the NeONKA (Neural Omnimodal Network with Knowledge-Awareness) neural network ensemble model, which involves multiple neural network models, supervised fine-tuning, and reinforcement learning with human feedback.
- Sber has developed a new neural network model that can perform various intellectual tasks, such as supporting a conversation, writing texts, answering fact-based questions, and creating images by integrating the Kandinsky 2.1 model into the ensemble.
- The company plans to release NeONKA 3.5 with 13 billion parameters to accelerate the evolution of the open-source community and AI.
- The model has been optimized to fit high load systems and trained using the Christofari Neo supercomputer.
SberDevices and Sber AI are developing the model with support from AIRI and AI experts to train GigaChat in specific knowledge domains and fine-tune its performance accurately. The new model will be integrated into many Sber products and services, including virtual assistant Salute.
Also Read: AI godfather Geoffrey Hinton quits Google