Naver and Kakao have revised their business strategies to move from using only their own artificial intelligence (AI) technologies to incorporating third-party large language models (LLM). This shift is seen as an effort to accelerate the enhancement of their businesses or services by integrating externally advanced LLMs.

According to industry sources on the 10th, both Naver and Kakao have decided to modify their previous strategy of relying solely on their own LLMs. Instead, they announced that they will adopt a two-track approach, utilizing multiple LLMs as needed.

Until now, Naver had developed its own LLM, "HyperCLOVA X," while Kakao had developed "Kanana."

However, among these changes, Naver announced its revised AI strategy during its 4th-quarter earnings conference call on the 7th, drawing attention to CEO Choi Soo-yeon’s participation in the "AI Action Summit." The summit, which gathers leaders from global AI companies, offers an opportunity for discussions on AI business matters with relevant stakeholders.

In the recent conference call, CEO Choi Soo-yeon stated, "We have kept the possibility open for collaboration with various external LLMs, including those from global big tech companies." She added, "Although there have been no cases of using third-party LLMs in service offerings yet, we are relatively flexible and are considering the introduction of external LLMs."

She further added, "While competitors are expanding AI capabilities indefinitely, HyperCLOVA X aims to efficiently build model capabilities and speed optimized for Naver services."

Based on CEO Choi’s statement, the IT industry suggests two possible directions. One is that HyperCLOVA X will be utilized for the integration of Naver services, while external LLMs will be used to enhance internal services. The other possibility is that Naver could separate the use of LLMs for areas that require a closed model, such as B2B business, and areas that need to adapt quickly to changes, such as B2C business.

On the 4th, Kakao held a press conference with Sam Altman, CEO of OpenAI, and announced plans to integrate OpenAI’s ChatGPT into its own LLM, "Kanana."

Kakao CEO Jeong Shin-a explained the partnership in a keynote presentation, stating, "By collaborating with OpenAI, a company with global technological competitiveness, we aim to provide innovative customer experiences and lead the popularization of AI services."

Kakao has decided to utilize OpenAI’s latest AI technology APIs for its major services, including KakaoTalk and Kanana. Furthermore, in order to accelerate its transition to becoming an "AI native company," Kakao has also decided to introduce ChatGPT Enterprise.

Kanana is an AI agent service that not only provides contextual responses in one-on-one conversations but also understands and engages in group chats, helping users build and strengthen relationships. Kakao plans to further enhance this with OpenAI's technology to offer users the highest level of AI experience.

Kakao had already set up an "AI orchestration" strategy to use external LLMs and has been preparing for its implementation. In the second half of last year, Kakao formed a separate AI orchestration team to develop a strategy for leveraging global big tech LLMs.

Kakao's AI orchestration strategy is seen as a way to complement its relatively lagging AI capabilities. Last year, Kakao decided to abandon its self-developed LLM "KoGPT," which had been delayed for release, and instead declared a complete restart of its AI efforts.

CEO Jeong Shin-a also stated, "We are now in a game where the focus has shifted from the competition between language models to providing meaningful services that users can actually use. Kakao is currently considering AI solutions that can be integrated into our services."

There are concerns from some quarters regarding Naver and Kakao’s two-track strategy of utilizing external LLMs. An industry insider noted, "The concept of 'sovereign AI,' which refers to AI sovereignty, is gaining attention." They added, "I hope that the two-track strategy of relying on external LLMs does not lead to neglecting the development of their own LLMs."

ChatGPT를 사용하여 번역한 기사입니다.