AsianScientist (Sep. 22, 2021) – It goes without saying that your morning commute in Hanoi, Vietnam, would be vastly different from what it would be in Houston, Texas. There were over 276 million vehicles on US roads in 2019, a figure that works out to 842 cars per 1,000 people. In contrast, motorbikes dominate the roads in Vietnam, which has just 23 cars per 1,000 people—nearly 37 times less.
Yet when it comes to training autonomous vehicles (AV), the vast majority of data comes from the US and Europe. VinFast, a Vietnamese automotive startup with its sights set on the global market, is hoping to use supercomputers to bridge the difference.
“When I came back to Vietnam, I had to relearn how to drive here—the traffic conditions are very different from the US,” said Dr. Hung Bui, an artificial intelligence (AI) researcher formerly with Google DeepMind and Adobe Research, and currently the director of VinAI, the AI research arm of VinFast’s parent company, Vingroup.
“After a while I got the hang of it, but it got me thinking a machine probably will do an even better job—Vietnam’s driving conditions provide the ultimate challenge for systems trying to reach Level 5 autonomy,” he said, referring to fully autonomous cars that will not even require steering wheels.
Helping VinFast along is the country’s most powerful AI supercomputer: VinAI’s NVIDIA DGX SuperPOD. Branded as the world’s first cloud-native supercomputer, the DGX SuperPOD will be used to retrain VinAI’s driving perception system as new data arrives every 24 hours, Hung said.
Supercharging R&D
World-leading AV systems built by private companies in Southeast Asia are just one of the many ways that high performance computing could impact the region. In light of the ongoing COVID-19 pandemic, for example, one of the most pressing needs is in drug development and healthcare. In Japan, the Fugaku supercomputer has been used to model the behavior of virus particles spreading among people.
“High performance computing provides researchers with tremendous power to address many grand scientific challenges,” said Dr. Rossen Apostolov, executive director of BioExcel, a Centre of Excellence for Computational Biomolecular Research at the PDC Centre for High Performance Computing, KTH Royal Institute of Technology in Stockholm, Sweden.
“Supercomputers allow screening of thousands of biomolecules in a matter of days for developing a cure or a vaccine against the virus. Similarly, modeling the propagation of a viral infection through a population allows for timely and effective measures to reducing its spread,” he added.
Using supercomputers to better understand the coronavirus comes amid growing HPC activity in the hemisphere. June 2021 will mark a year since Asia returned to the leading position in the TOP500 supercomputer rankings, with Fugaku—built by Fujitsu and research center RIKEN—bringing Japan back to No. 1 for the first time since 2012.
Aside from US machines, the intervening decade has also seen Chinese systems—Sunway TaihuLight and Tianhe-2A—leading the pack, resulting in a greater Asian presence in the field of high performance computing.
The Association of Southeast Asian Nations (ASEAN) is represented in the ranking with Singapore’s four HPC systems in the November 2020 list. While the TOP500 is only one way of taking the pulse of developments in HPC, Asia’s growing role is becoming manifest in a number of national and international initiatives.