AI All Startup News in Korea (Korean Startup News)

NAVER kicks off ‘Deview 2023’.. “Unveiling ‘HyperClova X’, the hyperscale AI in July”

2023-02-27 4 min read

NAVER kicks off ‘Deview 2023’.. “Unveiling ‘HyperClova X’, the hyperscale AI in July”

Reading Time: 4 minutes

NAVER’s largest developer conference in Korea, DEVIEW 2023, kicked off today. At DEVIEW, NAVER shared with participants the company’s advanced technologies in various fields that it is researching and developing for the global market, and reiterated its commitment to advancing technology together through communication.

The event began with a keynote by Kim Yoo-won, CEO of NAVER CLOUD, who spoke about how AI, such as ChatGPT, is heralding huge changes in the world. Kim said that NAVER CLOUD will leapfrog into a hyperscale technology company that leads the AI transformation through synergy between ultra-scale AI technology and cloud-SaaS business.

In addition, Kim announced that NAVER CLOUD will unveil HyperClova X, a hyper-scale AI that can be optimized to meet customer needs, in July. As the technology paradigm shift to super-scale AI such as Generative AI is rapidly progressing, Team NAVER, which represents the Korean IT industry, plans to lead the commercialization of super-scale AI and make it available to everyone, thereby creating a domestic super-scale AI ecosystem that is globally competitive.

“Team NAVER is ready to join the tide of global change by combining global-level AI technology and capabilities,” said Mr. Kim. “HyperCLOVA X is a super-scale AI upgrade that allows customers to combine their own data with HyperCLOVA to instantly provide responses that meet user needs. We expect that anyone, from individual services to specific companies or countries, will be able to create AI products optimized for their own purposes and create new business opportunities to create greater value.”

Since the advent of ChatGPT, there has been growing interest in super-scale AI services around the world, but commercialization has been challenging due to language limitations, data reliability, and high costs. Therefore, it is necessary to invest heavily in AI that best understands specific languages and cultures and fits the country’s industry and environment, as well as the technology and data for it. 

NAVER, which has already proactively invested in its own infrastructure and technology, plans to work with partners to improve the competitiveness of the domestic AI industry.

Following Kim, NAVER Cloud’s Head of Hyperscale AI Technology, Nakho Sung, explained the competitiveness of HyperClova X, NAVER Cloud’s CTO, Yongjae Kwak, explained the infrastructure essential for Hyperscale AI, and finally, Yongbeom Kim, Chief Scientist of NAVER Search US, continued the keynote by explaining the SearchGPT project.

Dr. Nakho Sung, Head of Hyperscale AI Technology, explained the competitiveness and strengths of HyperCLOVA X, which secures both functionality and stability. “For more specialized and advanced hyper-scale AI services, it is necessary to learn from data that is close to the user, whether it is an individual or an enterprise,” he said, emphasizing that the demo version of HyperCLOVA X is a real-world example.

“As HyperClovaX is an AI that learns 6,500 times more Korean than ‘ChatGPT’ and has been improved to express the appearance of AI that users want, even a small amount of data can be combined with the customer’s data to build a super-scale AI product optimized for that area, such as a specific service or enterprise,” he said. “We have also upgraded it to immediately provide responses that meet user needs through various interfaces.”

Next, CTO Yongjae Kwak introduced AI semiconductor solutions and hyperscale IDCs, which are essential for super-scale AI services. “Considering that super-scale AI will become more advanced in the future, an integrated approach to software, hardware, and operating environment is required,” he said.

Specifically, CTO Kwak said, “Together with Samsung Electronics, we are creating a lightweight AI semiconductor solution that has all the functions required for LLM (large-scale language model) computation, learning, and inference, but has a model size that is 1/10 the size of existing GPUs and more than 4 times the power efficiency.” “For each of the hyperscale IDCs to be opened this year, we have secured stable power capacity and cooling technology to efficiently provide AI infrastructure, and if this AI semiconductor solution is operated in each Sejong, it will be possible to operate more efficiently and reliably with an operating scale that is 4 times larger than existing GPUs.”

In the last part of the keynote, Dr. Yongbeom Kim, Chief Scientist of NAVER Search US, explained NAVER’s next-generation search technology project, SearchGPT. The SearchGPT project uses HyperCLOVA as the backbone of OCEAN, a search large-scale language model (Search LLM) specialized for NAVER Search. Based on this, the technology models NAVER’s 20 years of user search flow data and guides users to the best path to achieve their search goals, enabling NAVER to better understand search intent and results and generate answers that emphasize trustworthiness.

“We are focusing on three things: trustworthy information, connected to NAVER services, and multimodal for effective information provision,” he said. “We will generate information with accuracy and reliability through cross-validation and user feedback, and provide information optimized for users’ search intentions through organic linkage with NAVER services such as shopping, pay, maps, etc. in an intuitive form using images, videos, and voice.” “We expect to provide users with a more differentiated search experience through the ‘SearchGPT project,'” he concluded the keynote.

[Startup News in Korea]


투자자와 창업팀의 간극을 메울 수 있는 컨텐츠에 관심이 많은 초기 스타트업 투자자이자 와우테일(wowtale) 기자입니다. Linkedin Facebook투자실적
Leave a comment

댓글 남기기