[The Era of Supergiant AI] Ha Jung-woo, Head of Naver AI Lab, "A New Order Has Opened with Supergiant AI"

Asia Economy 'Get Ready Web 3.0, The 1st Mega AI Ecosystem Opening' Forum

Ha Jung-woo, head of the Naver AI Lab, is giving a lecture on "The Present and Future of the Super Large-Scale AI Ecosystem" at the "Get Ready Web3.0, The 1st Super Large-Scale Artificial Intelligence (AI) Ecosystem Opening" forum hosted by Asia Economy on the 27th at Gravity Seoul Pangyo in Seongnam, Gyeonggi Province.
Seongnam ? Photo by Kim Hyun-min kimhyun81@

[Asia Economy Reporter Yuri Choi] "A new order is being created as the era of ultra-large-scale artificial intelligence (AI) arrives."

On the 27th, Ha Jung-woo, head of Naver AI Lab, said this at the Asia Economy 'Get Ready Web 3.0, The 1st Ultra-large AI Ecosystem Opening' forum held at the Gravity Hotel in Pangyo. Ha explained, "By utilizing ultra-large AI that can learn autonomously and be used universally, it is possible to experiment with various services cost-effectively without creating AI models for each individual service."

He identified 'self-supervised learning' and 'transformer' as the key technologies that ushered in the era of ultra-large AI. Self-supervised learning is a learning technique that does not require providing answers for the data. Previously, every time data was trained, answers had to be provided, which was time-consuming and costly, but self-supervised learning eliminates this need, enabling training on large-scale data. The 'transformer,' capable of processing large amounts of data such as text, voice, and video, allows stacking layers of AI models, making ultra-large AI possible. Ha said, "It was expected that increasing data volume would cause qualitative problems, but according to recent research released by Meta, models trained on large amounts of data have less bias and are much more stable."

The characteristic of ultra-large AI is that it can enhance performance without changing the features of previously trained models. In the past, AI that could be commonly applied was created, and then specific data was trained to develop problem-specialized AI. In such cases, the features of previously trained data or models changed. In contrast, ultra-large AI can enhance performance without changes, enabling rapid service deployment with less cost and time. This means a 'profitable business' from a corporate perspective. Ha emphasized, "This is why many companies are focusing their investments on ultra-large AI," adding, "With minimal effort, various applications can be quickly tested."

As AI evolves into ultra-large AI, an era of collaboration between AI and humans has also begun. When developers write the functions they want to implement in text, AI generates the corresponding source code. Developers can focus on overall program design, reducing errors and increasing productivity. Ha said, "When text is input, AI creates images or videos accordingly," diagnosing, "As AI continues to develop as a content creation tool, it is time for creators to consider how to utilize it."

Naver is expanding its ecosystem through the ultra-large AI 'HyperCLOVA.' Technologies have been applied to Naver Shopping and search, and a service called 'CLOVA Care Call' was introduced to provide companionship for elderly people living alone. On the other hand, through 'CLOVA Studio,' the barrier to ultra-large AI technology has been lowered. CLOVA Studio is an open platform that allows exploring AI application possibilities and utilizing them in actual services without specialized development knowledge. Ha said, "Like semiconductors, AI is creating innovative functions as parameters grow larger," adding, "An ecosystem has emerged where startups can create applications without technical barriers based on no-code AI."

This content was produced with the assistance of AI translation services.