Naver Corp. and Kakao Corp., Korea’s two tech giants, are postponing releases of their latest generative artificial intelligence (AI) models to fine-tune their performance, as they are conscious of the better-than-expected Korean-language performance at AI services on offer from global behemoths like Google LLC and Microsoft Corp.

The local tech giants seem to be choosing to keep working on enhancing their AI models instead of committing to a hasty release out of belief that their generative AI models, when combined with their other services, will win market-dominant positions thanks to their strong dominance and user base already built-in through their web portal and messaging app services.

According to industry sources on Wednesday, Naver and Kakao both have decided to delay the launch of their upgraded generative AI models and vertical AI services. Earlier this year, the two companies were expecting to release their own AI models that could compete with global AI service sometime in the first half. However, market expectations of what their AI models should offer have now been adjusted following a series of new AI services from global tech titans.

“Both Naver and Kakao ultimately aim to create personal digital agents that can provide an AI assistant that helps with scheduling and asset management,” said another unnamed industry insider. “The delays are probably based on their strategic decisions to release a better-functioning AI model and services, what they dream of, rather than having a hasty release just to meet a launch target,” the insider added



Naver is seemingly the most avid in this latest trend. Leaving its glory as the nation’s No.1 web portal behind, the company is working on a full-scale transformation of its services, including changes in platform user interfaces (UI) and user experiences (UX), in line with the new era’s paradigm shift. Hyper Clova X, an upgraded large language model (LLM) AI of its previous Hyper Clova, will be released sometime in July or August. The release of the company’s upgraded search chatbot based on Hyper Clova X has also been pushed back to the third quarter. Naver plans to run limited internal and beta tests for revisions before rolling out the chatbot service, dubbed Search GPT for now.

“The company has a pretty clear goal: to develop an LLM that can satisfy its business-to-business customers. It is working on the sophistication of its LLM to the extent of emergency operations to meet that internal goal,” said an industry insider who has knowledge of the company’s situation. Following a recent website re-design for the desktop version of its web portal, the company is now planning to revamp its mobile version, including the mobile homepage, integrated search functions, as well as the UI and UX.

Kakao’s approach is more open-ended. It will wait until its hyperscale AI has satisfactory performance before any release, with no specific timeframe set in place. Kakao Brain Corp., the company’s AI subsidiary, will unveil KoGPT 2.0, an upgraded version of its LLM, KoGPT, as early as sometime in the third quarter this year, pushing back the initial plan to release it in the first half of this year. An interactive AI chatbot based on its LLM, called KoChat GPT for now, is planned for release within the year. An AI-based total service that enables a wide range of services, such as reservations, payments and commerce, all within the Kakao environment, is also part of the IT giant’s grand plans.

“We are debating whether to release KoGPT 2.0 and KoChat GPT simultaneously, or to release the LLM first and then launch the service. We plan to showcase a roadmap for our generative AI business later this year,” said one official from the company. The release of the upgraded version of Kalo, Kakao Brain’s AI image generator, has also been postponed to early next month, from this month.

By Ko Min-suh and Chang Iou-chung

[ⓒ Pulse by Maeil Business Newspaper & mk.co.kr, All rights reserved]