Vietnam needs to accelerate its pace in the race for Artificial Intelligence.

The above statement was made by Mr. Nguyen Kim Anh, Chief Product Officer (CPO) of VinBigdata, in an interview with VietNamNet regarding the Vietnamese big language model that VinBigdata announced the successful development of on August 21st. VietNamNet would like to share the content of this interview with our readers.

Could you tell us about the current state of AI application in Vietnam?

Mr. Nguyen Kim Anh: AI is one of the most popular keywords in the world in recent years. In Vietnam, we have also discussed this topic in many different forums. However, the reality shows that the deployment of AI in Vietnam still faces many challenges such as limitations in resources, infrastructure, and specialized personnel…

According to the 2022 Artificial Intelligence (AI) Readiness Index report, Vietnam ranked 55th globally, while Singapore ranked 2nd, Malaysia 29th, and Thailand 31st. This indicates that Vietnam still needs to be more proactive in promoting AI research and application.

I believe we have many advantages in leapfrogging because Vietnam possesses a young workforce with a very strong foundation in mathematics and engineering. This is a significant advantage that we need to seize. With the support and cooperation from the government and businesses, I believe we will quickly grasp the potential for developing the AI industry to contribute to the socio -economic development of the country.

Why did VinBigdata decide to develop the recently launched Vietnamese-language big data model?

Mr. Nguyen Kim Anh: The large language model is a solid foundation for the creation of AI-generated applications, such as Open AI's ChatGPT or Google's Bard. However, for products from foreign countries, Vietnamese is not among the core languages. This affects the quality of content delivered to users to some extent.

If you ask for detailed information that is specific and characteristic of Vietnamese people, the likelihood of errors is quite high. More seriously, if users use this tool to look up information and update news, it could lead to negative consequences in the future, such as distorted basic knowledge among a segment of the Vietnamese population, especially young people.

Large-scale language models provide a solid foundation for the creation of AI-generated products.

Therefore, building a large-scale Vietnamese language model is essential for developing products that best suit and serve the needs of Vietnamese people. From its inception, VinBigdata has identified "data" as the core and most important factor in researching and launching AI products to serve the market.

This is also why we were able to pioneer the early completion of building a large-scale Vietnamese language model. Instead of using 175 billion parameters like ChatGPT, we only needed a few billion parameters. This is not only significant in terms of application but also has social implications, as more and more people are using AI generation as a tool for learning, working, and searching for information.

With the recently announced Vietnamese big language model, how will VinBigdata apply it, sir?

Mr. Nguyen Kim Anh: Having complete control over the technology, developing from scratch, and successfully building a large-scale Vietnamese language model is considered a significant step forward, helping VinBigdata integrate AI generation technology into the ecosystem of products and services it offers to the market.

We have outlined a specific roadmap for putting research findings into practice. Specifically, VinBigdata will introduce two versions of the application to consumers: first for businesses and then for the community.

With its enterprise version, VinBigdata will integrate technology to make VinBase (a comprehensive multi-cognitive artificial intelligence platform) the first AI generation platform in Vietnam, while also providing development solutions based on AI generation technology such as Generative AI chatbots, callbots, and the next-generation ViVi virtual assistant…

This technology will help increase the naturalness of machine communication, while also supporting users in searching for and synthesizing information much more quickly and simply than before.

VinBigdata has initially successfully applied this new technology to its VinBase KB (VinBase Knowledge Base Portal) product line. The product is capable of retrieving information and automatically generating answers based on data collected from massive datasets within its knowledge base.

Successfully building a large-scale Vietnamese language model is considered a significant step forward, enabling VinBigdata to integrate AI generation technology into its ecosystem of products and services already offered to the market.

According to our plan, at the end of December 2023, VinBigdata will launch ViGPT – “Vietnamese version of ChatGPT”, open to the community for completely free access and testing. With ViGPT, users can ask and answer questions about information specific to Vietnam (regulations, legal documents, etc.), or information about specific characteristics (history, geography, literature, etc.).

This is a project of great significance to the community and society, and is expected to become a useful tool, providing strong support for Vietnamese people in their studies and work.

Does VinBigdata intend to share this Vietnamese big language model with interested parties?

Mr. Nguyen Kim Anh: The successful development of the Vietnamese big language model is not only significant for VinBigdata but also for society and the science and technology community in Vietnam. The VinBase platform integrates AI generation, allowing businesses to fully utilize this core technology for their own branded products.

With this model, VinBigdata hopes that Vietnamese businesses will have more opportunities to access AI technology that generates or inherits the achievements from the Vietnamese big language model that we have developed.

Furthermore, with ViGPT, Vietnamese people will also have their own "Vietnamese version of ChatGPT," allowing them to easily search for information quickly, accurately, and reliably.

The Vietnamese big language model was introduced by VinBigdata at an event in Ho Chi Minh City.

Security and copyright are crucial factors in developing generative AI. With the recently launched Vietnamese big language model, how has VinBigdata addressed this issue, sir?

Mr. Nguyen Kim Anh: As a research unit specializing in big data and artificial intelligence, VinBigdata always respects data security by building rigorous processes and strictly adhering to data security regulations. Specifically, to ensure user rights, VinBigdata strives to create transparency in the process of data collection and use, especially with data owned by individuals.

Simultaneously, we proactively build multi-layered security measures, rigorous data lifecycle control processes, and appropriate data access control (who has the right to access and use which data?) to prevent information leaks. Globally and in Vietnam, there are regulations and standards regarding data security, and VinBigdata always adheres to these principles.

Furthermore, copyright issues have become a rather thorny topic following the rapid development of AI in recent times. Understanding this, VinBigdata focuses on developing products that provide specific, localized information characteristic of Vietnamese people and specific fields, rather than a broad, general approach, thus avoiding misinformation.

Thanks to this initial strategic direction, VinBigdata can confidently create products that provide more accurate knowledge than ChatGPT in specialized fields such as finance and banking, law, healthcare, etc., and with information specific to Vietnam such as history, geography, literature, etc.

According to him, is it time for Vietnam to apply AI in practice?

Mr. Nguyen Kim Anh: I think we need to accelerate even more in the race for artificial intelligence. Because the countries and large corporations around the world have already moved ahead of us by a certain margin.

Currently, Vietnam can be said to have not yet truly implemented AI extensively in practice and is only in the initial stages.

However, I believe that with concerted efforts from all levels, combined with the high-quality AI workforce currently available in Vietnam, we can seize the opportunity to leapfrog ahead!

Source