Sber announced the start of closed testing by invitation of the multimodal neural network GigaChat. System
The GigaChat service architecture is based onneural network ensemble of the NeONKA (NEural Omnimodal Network with Knowledge-Awareness) model. It includes various neural network models that allow AI to solve a wide range of tasks: answer questions, keep up a conversation, or write texts. In addition, the development includes the Kandinsky 2.1 model, which is used to create images.
The model is currently in closed beta mode. To take part in testing, you need to join the closed community in Telegram. Access is via a link.
Representatives of the company say that they workon the implementation of GigaChat into existing Sber products and services. In particular, the Salyut virtual assistant will start using it in the near future, and there are also plans to implement a cloud API for business. In addition, in order to accelerate the development of the open-source community and technologies in the field of artificial intelligence, the company plans to make ru-GPT 3.5, which is part of NeONKA, with 13 billion parameters in the public domain.
After the success of chatbots based on ChatGPT, manyCompanies around the world have entered the AI race. Earlier this year, Google introduced its own Bard AI model, and Chinese tech giants Baidu and Alibaba Group reported limited testing of the models.
Read more:
There was a photo inside the second deepest underwater sinkhole in the world
See what happened to Mercury when it got as close as possible to the Sun
Scientists are ready to recognize the new tree as the oldest in the world