With the advent of the era of ultra-large artificial intelligence (AI), a new order is being created.”
Jungwoo Ha Naver’s AI Lab director made the remarks at the Asia Economy “Get Ready Web 3.0” forum held at the Pangyo Gravity Hotel on the 27th, the opening of the first ultra-large AI ecosystem. Director Ha explained, “If you use ultra-large AI that can be learned and used for general use, you can experiment with various services cost-effectively without creating an AI model for each service.”
“Self-directed learning” and “transformers” were cited as technology keywords that opened the era of ultra-large AI. Self-guided learning is a learning technique that does not have to provide the correct answer to the data. In the past, it took a lot of time and money to give answers every time they learned data, but self-governing learning does not have to be so, so large-scale data can be learned. Transformer, which can process a large amount of data such as text, voice, and video, can stack layers of AI models, enabling ultra-large AI. Director Ha explained, “We expected qualitative problems if the amount of data increases, but according to a recent study released by Meta, the model that learned a lot of data is less biased and much more stable.”
The characteristic of ultra-large AI is that it is possible to upgrade performance without changing the characteristics of previously learned models. In the past, problem-specific AI was created by creating AI that can be applied in common and learning specific data. In this case, the characteristics of the data or model learned in the past have changed. On the other hand, ultra-large AI can quickly become a service at a low cost and time because it can upgrade its performance without changing. This means that it has become a “remaining business” for companies. This is why many companies focus on investing in ultra-large AI, Ha said. “We can quickly test various applications with a little effort.”
As it evolved into a super-large AI, there was also an era in which AI and humans collaborated. When developers write down the functions to be implemented in writing, AI writes the appropriate source code. Developers can focus on designing the entire program to reduce errors and increase productivity. Director Ha said, “When you enter text, AI makes it an image or video,” and added, “As AI continues to develop into a content creation tool, it is time for creators to consider using it.”
Naver is expanding its ecosystem through the super-large AI ‘Hyperclova’. It applied related technologies to Naver shopping and search, and also introduced a service that helps elderly people in single-person households talk through “Clova Care Call.” On the other hand, the threshold for ultra-large AI technology was lowered through ‘Clova Studio’. Clova Studio is an open platform that can examine the applicability of AI and use it for actual services without development-related expertise. Director Ha said, “AI is also creating innovative functions as parameters grow like semiconductors,” adding, “Start-ups have created an ecosystem that creates applications without technical barriers based on NoCode AI.”
