In a time when data is both gold and a source of power, how can Taizhi Cloud leverage the strengths and services of AIHPC to help industries embrace artificial intelligence and accelerate their ability to meet various challenges?
Artificial intelligence (AI) and high-performance computing (HPC) are already well-known technologies in recent years, but what exactly is the strategy behind combining the two?
"Imagine that AI will generate a massive amount of data in the future," Wu Han-chang, general manager of Taiwan Smart Cloud (Taiwan Smart Cloud), vividly described the picture. However, in order for this data to be effective, it will need the assistance of computing power and the connection of different application areas. "Therefore, we need to build a broad channel and tools to accelerate the transmission speed between the two places."
The "channel" Wu Hanzhang mentioned relies on the 5G infrastructure of its partner, Taiwan Mobile, while the accelerated transmission is achieved through NVIDIA's GPU computing power. With the assistance of this computing power and the high bandwidth and low latency of 5G, even with massive amounts of AI data, there are no concerns about transmission difficulties. This prevents data from becoming isolated and instead provides smarter guidance for research and applications. So, how does Taizhi Cloud possess the capability to build AIHPC?
Backed by supercomputers, it is impeccable in both speed and energy efficiency.
"Our AI service is based on a supercomputer as its core design," Wu Hanzhang said confidently. The supercomputer he referred to is none other than "Taiwania 2," a collaborative project between ASUS, Taiwan Mobile, the National Center for Power Research (NCPR), and other partners. Besides ranking 20th in the world's Top 500 highest-performance computing mainframes, it consumes only 798 kW of power while performing 9 gigaflops (9 PFLOPS) of computation, making it the most energy-efficient high-speed computing mainframe in Taiwan's history. "We believe that in this era of promoting ESG, we can save customers more than half of their carbon footprint," Wu Hanzhang said.
With the Taiwan Sugi-2 as a powerful computing backbone, the provision of services by Taiwan Smart Cloud is greatly enhanced. In medical institutions, Taiwan Smart Cloud can solve the problems of huge pathological images, time-consuming AI model training, and the need for large amounts of memory and GPUs. Results that used to take 10 days can now be obtained in just 1.5 days with the help of Taiwan Smart Cloud AIHPC, accelerating and improving the quality of medical care. In the process of building smart cities, Taiwan Smart Cloud can also leverage its advantages in providing hybrid cloud services, ensuring the security of smart cities is not compromised by meeting the needs of low-latency communication and efficient image transmission between the cloud and the ground.
Therefore, Wu Hanzhang stated that Taiwan Smart Cloud's local advantages align with the government's plan to promote AI in Taiwan, aiming to accelerate the implementation of application services. The adoption of a cloud architecture aims to overcome geographical limitations and facilitate the deployment of hybrid cloud solutions by integrating with customers' local infrastructure or other public cloud deployments. To further promote AI applications across industries, Wu Hanzhang also revealed that the "AI Factory" project is actively underway, having already partnered with over 150 startups in fields such as precision medicine, natural language processing, and autonomous vehicles. The goal is not only to reach over 250 startups by the end of the year but also to leverage ASUS resources to create better products through Taiwan Smart Cloud's services.
With broad applications, AI can penetrate into various fields.
In addition to applications such as smart healthcare and smart manufacturing, the platform built by Taizhi Cloud can support customers in various large-scale and massive data processing, accelerate the model training of smart industry applications, and enable general enterprises to achieve digital transformation while embracing the use of digital tools.
To this end, Wu Hanzhang pointed out that TAIJUI offers various solutions, including the OneAI one-stop machine learning development collaboration platform, the TWCC hybrid cloud architecture, and DAS Analytics Master, enabling AI applications to penetrate deeper into various fields. Wu Hanzhang stated that TAIJUI can provide customers with more complete solutions, helping to reduce costs that previously required time, equipment, and even manpower, lowering the barrier to entry into artificial intelligence, and accelerating the pace of industrial transformation.
Taking OneAI's one-stop machine learning development collaboration platform as an example, Taizhi Cloud understood that enterprises might encounter problems such as insufficient IT manpower or computing power when developing machine learning models. Therefore, it proactively built a basic model for customers to use. Lu Daocheng, OCIS Manager of ASUS Computer, shared that their main technology is developing image recognition solutions, but how to quickly improve accuracy has become a headache in the technology development process.
"Fortunately, Taizhi Cloud provided a solution that allowed us to accelerate model training," said Lu Daocheng. After importing the image data, the team only needed to adjust the parameters on the platform for each result, which reduced the development of the original model, the construction of infrastructure, and the use of manpower. These were the main reasons why the team chose to use Taizhi Cloud services, which accelerated the development process for users.
But what if the desired model isn't available on Taizhi Cloud? Haifu Smart addresses the challenge of enabling brand clients to more accurately identify valuable consumers in the cookieless era. CEO and co-founder Hao-Hsien Chu revealed that they have also noticed a new challenge in the industry: a major problem in the cookieless era is that brand clients become unfamiliar with how to market their products and services to consumers. Faced with a large amount of past data, how can they help brand clients find valuable consumers?
Therefore, Haifu Smart leveraged the expertise of Taizhi Cloud to assist in training an AI model tailored to Haifu Smart's needs. Through this new AI model, when brand clients encounter unfamiliar consumers, they can use the AI model to determine the consumer profile and promote products and services in a way that the consumer prefers. Zhu Haoxian stated that Taizhi Cloud provided Haifu Smart with the supercomputer used to train the AI model. "The supercomputer, combined with our existing massive amounts of data and industry expertise, can generate an AI model that accurately outlines the consumer profile in a short period of time, reducing marketing costs for brand clients and increasing sales."
Zhu Haoxian further explained that training a large-scale AI model is costly, and without a professional partner, a lot of resources, time, and money would be wasted, and it would be impossible to create a valuable AI model afterward. Taizhi Cloud became the perfect partner in this situation. Its diversified AI tools, combined with supercomputers, allowed Haifu Wisdom to focus on professionally organizing data, while entrusting the AI model calculations to Taizhi Cloud, jointly creating a solution to meet the challenges of the cookieless era.
Be a top-notch teammate for businesses, and cultivate the next national guardian, "Shan".
"What we hope to provide is assistance to enterprises in the most difficult stage of starting from 0 to 1," said Wu Hanzhang. He believes that digital transformation may be one of the possible outcomes of embracing AI, but with the help of high computing power, data can create value for the industry, generating great benefits from research to application scenarios. "And Taizhi Cloud also has strong resources to assist in the 1 to N stage." After overcoming the most difficult two major challenges, Wu Hanzhang believes that Taizhi Cloud can still be a partner for enterprises to meet the challenges of the market.
Looking to the future, Wu Hanzhang believes that with supercomputers as its strongest backing, Taiwan Cloud should also shoulder the responsibility of popularizing and democratizing AI. While actively researching management and business models, it should also continue to be a great partner in the industry as the market embraces the new era of metaverse, Web3.0, and even blockchain. "I hope we can become the next guardian deity 'Shan'," he said.
Source:Digital Age