In addition to the ongoing ‘chip war’, the world is witnessing another technology phenomenon. In fact, AI has existed for a long time and people have been using it without taking notice of it. For example, people use apps that help make them look more beautiful when taking pictures; or they receive suggestions on music pieces, videos or websites.

Generative AI has extremely powerful capabilities which can be utilized in many fields. All technology firms want to have Generative AI and use it as a leverage to create a competitive gap over their rivals.

Tech firms are rushing to create Generative AI of their own. And in the battle, labor force, data and GPU (graphics processing units) are the major factors.

The question is where Vietnam should be in the AI war, whether it should become a player, or stand outside and just get benefits from the players.

Generative AI products that people are experiencing, such as ChatGPT, Bing and Gemini all have large language models (LLM) with hundreds of billion of parameters. The models are created from huge volumes of data.

In order to train, update and operate LLMs, businesses have to spend big amounts of money. To operate ChatGPT alone, OpenAI has to spend $700,000 each day. Meanwhile, to create a new version, businesses would need about $100 million.

Only large firms are capable of training and operating large models. Meanwhile, other firms focus on building apps based on Generative AI and tune smaller models to serve very specific works.

At present, Vietnamese language is still not well supported by LLMs, including exclusive and open-source models. This partially restricts Generative AI apps in Vietnam.

Hung said he believes that Vietnam should gather strength to build a good data set in Vietnamese and provide it publicly, so that all those who want to train language models can use it.

The next thing that needs to be done is building a data set assessing the Vietnamese language processing capability of Generative AI models, so that businesses and app developers can choose suitable models for their apps.

If Vietnam can do both of these, Vietnam and language model trainers will ‘win-win’. 

“At this moment, I think that Vietnam should focus on apps rather than training a LLM,” he said.

Research about Generative AI was first conducted in 2017, but Generative AI only began catching higher attention five years later when ChatGPT was born. 

According to Bill Gates, there are two fields that Generative AI can create the most influence – healthcare and education. This could be where Vietnam should start when developing Vietnamese Generative AI apps.

Asked if Vietnamese AI developers have opportunities to compete with international giants in the Vietnamese market, and if Vietnam’s AI can reach out to the world, Hung was optimistic.

Hung believes that if Vietnamese developers can create a data set of Vietnamese language good enough to fine tune and assess LLMs, this will bring benefits to all.

Whether AI can generate big value also depends on the creation of apps as well, not just models. There could be many other Generative AI apps, not just chatbot and virtual assistants.

In general, any individual and institution that can understand what the market needs and can give good solutions will have opportunities, no matter whether they are domestic or foreign invested.

Many programs and forums have been held recently in Vietnam to promote the development of AI. 

Trong Dat