"What Vietnam shouldn't do is that it should not build another big model," Curtis S. Chin, chair of senior fellows and senior advisor for global markets at the U.S.-based Milken Institute, said at the GStar Summit in Ho Chi Minh City on May 29. "Leverage what Google has, what Meta has, what OpenAI has, and then build applications on top of that, customized to fit Vietnam."

Jay Kim, chairman and CEO of South Korean healthcare company Boryung and a board member of healthcare and space company Axiom Space, said the same principle applies beyond AI. Using the space industry as an example, he said many Asian countries, including South Korea, had long pursued a strategy of catching up with technological leaders in the U.S., Europe, and Japan.

"But nowadays with these new technologies enabling us to do so many things, we have to choose a path that is relevant for humanity, rather than simply trying to catch up. So don't pursue the fast followers strategy, just do something that you think matters to humanity."

A fast-follower strategy refers to companies that wait for market leaders to test products before quickly adopting, improving, and scaling them.

Guest speakers at the GStar Summit in Ho Chi Minh City on May 29, 2026. Photo by VnExpress/Bao Lam

Dr. Bui Hai Hung, vice president of technology at Qualcomm AI Research, former CEO of VinAI, and a former Google DeepMind researcher, said his team reached a similar conclusion while developing AI technology in Vietnam.

Before its acquisition by Qualcomm, VinAI invested heavily in advanced AI models such as GAN-VAE, a technology that combines two AI techniques to generate realistic images and audio and other data while maintaining greater control over the results, and regularly presented research at leading AI conferences.

However, the company quickly realized it could not compete directly with organizations that possess vast GPU resources, and shifted toward optimizing smaller AI models that retained most of the capabilities of larger systems, he said.

Running AI directly on devices such as smartphones is better suited for handling private data and reduces resource consumption, protects privacy, and minimizes latency, he said. He added that Qualcomm AI Research continues to pursue that approach, with potential applications in wearable devices, robots, smart glasses and other systems that require near-instant responses.

Yi Tay, a scientist at Google DeepMind, said AI is entering a new phase after years of emphasis on building ever-larger models. Researchers once believed intelligence would continue improving simply by increasing model size, data volume, and computing power, but the rise of AI agents has shown that scaling alone is not enough, he said.

Instead of building models from scratch, companies can use open-source systems and fine-tune them for specific tasks or develop layers that connect multiple AI components, he said, adding that the greatest impact would come from applying AI to real-world problems.

Dr. Luong Minh Thang, research director at Google DeepMind and co-founder of the nonprofit AI organization New Turing Institute (NTI), said the world is moving from the "data era" to the "experience era." As a result, Vietnamese businesses should place greater emphasis on user experience, he said.

"What AI would need next would be about building the experience, building the simulation. For example, how do you simulate a world so that AI can navigate in the real world? How do you simulate virtual cells so you can validate drugs?"

Wendy Nguyen, co-founder of Pacific Gateway Partners, said Vietnam's greatest advantage lies in its people rather than technology. "If you look at the Vietnamese people, we are hungry, we are thirsty and we need to keep that."

She recalled a conversation with a senior executive, who had observed that people in many countries become less driven as their lives improve. But in Vietnam, people have maintained a strong desire to advance though living standards have risen, she said. "That resilience, the thirst we have for years, is pushing us forward."

She also expressed concern that the global AI startup ecosystem may be overheating. In the past, founders often struggled to raise funding, but today, some startups can attract billions of dollars based largely on an idea and a name, she said. Investors in Vietnam should be more cautious and consider whether the AI sector is expanding too rapidly, she warned.

Dr. Luong Minh Thang, research director at Google DeepMind and co-founder of the nonprofit AI organization New Turing Institute. Photo by VnExpress/Bao Lam

Dr. Ed H. Chi, vice president of research at Google DeepMind, said developing talent requires building new skills from an early age, particularly the ability to work effectively with AI systems. The most important skill in the future may not be answering questions, but learning the skill of "asking really good questions," he said.

Chatbot models today are highly effective at answering questions but are "terrible at posing questions," he said. "I think the next iteration of the technology will be asking if these models are really good at asking questions."

He said the same principle applies to businesses, where success often depends on identifying the right problems to solve. The industry was approaching a point where it had developed a highly capable AI "engine" and should now focus on how that technology can benefit society, he pointed out.

"I think we need to start to ask the question like how do we use this engine right for humanity."

GStar 2026 is part of an annual series of AI events organized by NTI, formerly known as VietAI, since 2018 to connect Vietnam's AI community with global experts. Previous editions featured leading figures in AI, including Sam Altman, CEO of OpenAI, and Christopher Manning, director of Stanford University's AI Laboratory, at an event in 2023, and Jeff Dean, chief scientist at Google, in 2024.