VinBigdata successfully develops AI generation technology.

VinBigdata announced the successful development of a Vietnamese big language model on August 21st, creating a foundation for mastering AI generative technology.

Large Language Models (LLMs) are models trained using deep learning techniques on massive datasets of text or images. These models are capable of understanding knowledge, generating text independently, and performing various natural language processing tasks. They are considered key to developing generative AI technology – capable of creating new content and ideas in various forms (text, images, audio, etc.).

With the successful development of a large-scale Vietnamese language model, VinBigdata will integrate technology to make VinBase (a comprehensive multi-cognitive artificial intelligence platform) the leading AI generation platform in Vietnam, while also providing solutions developed on this technology such as Generative AI chatbots, callbots, and the next-generation ViVi virtual assistant... This technology enhances the naturalness of machine communication and helps users search and synthesize information more quickly and easily than before.

Professor Vu Ha Van - Scientific Director of VinBigdata Company. Photo: VinBigdata — Professor Vu Ha Van - Scientific Director of VinBigdata Company. Photo: *VinBigdata*

Professor Vu Ha Van, Director of Science at VinBigdata, stated that several large corporations worldwide have successfully researched and launched products based on big language models, such as OpenAI with ChatGPT and Google with Bard. In Vietnam, VinBigdata is being developed by Vingroup to build a Vietnamese big language model. According to Professor Van, this model focuses on solving three core problems: improving accuracy, reducing infrastructure costs, and ensuring security.

"Instead of needing around 175 billion parameters like ChatGPT, VinBigdata can create a large-scale language model with a few billion parameters while still being able to generate highly authentic texts, focusing on Vietnamese data and knowledge," the unit's leader said.

Mastering technology, developing from scratch, and building a large-scale Vietnamese language model are considered crucial steps for VinBigdata in integrating AI generation technology into its ecosystem of products and services in the market. Currently, the company is initially applying this new technology to its VinBase KB (VinBase Knowledge Base Portal) product line. This product is capable of retrieving information and automatically generating answers based on data collected from massive datasets within its knowledge base.

Representatives from VinBigdata shared this information at the AI generation launch event. Photo: VinBigdata — Representatives from VinBigdata shared this information at the AI generation launch event. Photo: *VinBigdata*

Vingroup is expected to launch two main product lines this December: VinBase 2.0 and the ViGPT application. VinBase 2.0 is an AI platform that generates multi-cognitive responses with solutions for businesses and government agencies. Meanwhile, the ViGPT application, introduced by the company as "the Vietnamese version of ChatGPT," will be open for community access and testing. With ViGPT, users can ask and answer questions about specific Vietnamese information (regulations, legal documents) or local information (history, literature, scenic spots, local specialties).

VinBigdata Joint Stock Company has a database system of up to 3,500 Terabytes. The system contains hundreds of thousands of hours of voice data, images, and information that have been cleaned, processed, classified, and used for AI training and R&D infrastructure. This infrastructure includes dozens of Nvidia DGX A100 server clusters. The company also boasts a team of Vietnamese professors, scientists, and technology experts from around the world.

Hoai Phuong

Source link