AI quynhon FPT.jpg
At FPT AI Quy Nhon (Photo: FPT)

Having worked for FPT for 14 years, Chien has been assigned the task of building and developing an AI center in Quy Nhon.

“Before joining FPT, I once worked for Microsoft. One of my largest aspirations is utilizing technologies in reality to improve the living standards of Vietnamese. And with the establishment of AI Center in Quy Nhon, my aspiration has come true,” he said.

Recalling the first days of the center’s operation, he said: “We understood that we were not really good at research, therefore, we needed to cooperate with research institutes in the world and learn from them. We have been cooperating with Mila, the world’s largest AI research institute, and with Google."

“In 2021, the AI center joined forces with Google to implement a project to support African farmers. About 2 million African farmers are using our product,” he added. “In late 2022, Google, which praised the product, gave an award to the center."

Chien takes pride in his staff. “AI, by nature, is mathematics. At FPT AI Quy Nhon, we have many specialists in mathematics."

The Vietnamese tech firm has had scientific articles published in prestigious international publications. Last year, an article on using Computer Vision in screening cancer was published in the world’s largest scientific journal.

With the qualified staff, Chien believes that Quy Nhon AI Center will become one of the leading AI centers in the region by 2030. FPT University specializing in training in AI is located in Quy Nhon, and an AI center is also there. The center is cooperating with some research institutes in the world to build an FPT AI Institute in the city.

“We continue to cooperate with Mila and Google to carry out some research and learn some technologies. We hope that together with the partners, we will set up an AI institute in Quy Nhon,” Chien said.

The strengths

“In terms of Explainable AI, I can say for sure that we are extremely good and we are highly appreciated in the world. 

Explainable AI helps solve a lot of questions, especially in the fields of finance, healthcare and insurance. Doctors and insurers do not accept AI results that are not be explained.

For most people, an AI model is considered a black box. Previously, people just input information into AI model and AI would give answers without explaining why. But now, Explainable AI will give explanations.

Explainable AI also helps reduce costs related to the development and building of one AI model.

“When building an AI model, the cost for programmers to create the model is not high. The costlier work is collecting enough data to train the AI model,” he explained. “Explainable AI can reduce the costs for creating data. Instead of 1 million records to create an AI model with the accuracy rate of 95 percent, we need 1,000 records only."

Over the last three years, in addition to providing AI solutions to clients, FPT AI Quy Nhon developed products with high quality, provided to many markets, including Japan, Canada, the US, France and Germany. 

One of the products Chien is proudest of is an AI app called akaCam. The product has been used at 20 factories, some commercial banks (ACB, TPBank and Techcombank) and some retailers (Phat Retail).

akaCam not only can recognize subjects, but also can analyse their behaviors. At the Vietnam-Australia shrimp plant, it is used to mark the behavior of feeding shrimp (the feeding may have adverse effect to shrimp quality if workers don’t strictly observe the procedures).

akaCam can also be used at households to discover thefts and support old people and children in case of emergency. It is used at shops and bank transaction offices, and automobile manufacturing factories.

In addition, the Vietnamese tech firm has also researched and developed its products in Edging Computing.

“If you want an AI solution to run, you need to have a GPU system (graphics processor -  electronic circuit capable of performing mathematical operations at high speed) that is powerful enough, especially for solutions that need to be processed promptly in real time,” he explained. 

“Deploying an AI solution requires huge infrastructure costs. This is also one of our concerns when developing products,” he said. “Anyway, once we have Explainable AI, we can optimize and ensure high accuracy, thus reducing infrastructure costs a lot”.

Binh Minh