From
Send to

Kakao to join AI battle with Korean-language KoGPT

Feb. 10, 2023 - 15:37 By Jie Ye-eun

Kakao's Pangyo headquarters in Gyeonggi Province (Kakao)

Kakao, the operator of South Korea’s No. 1 messenger app KakaoTalk, said Friday that it plans to expand its vertical Korean language-based artificial intelligence services based on an AI language model called KoGPT this year.

Kakao CEO Hong Eun-taek said in its earnings conference that his group would seek to compete against global AI services with a specialized Korean language service.

“The rise of hyperscale AI models like ChatGPT is both an opportunity and a crisis for Kakao,” he said.

Global tech companies with ample capital and technologies have a "definite competitive edge" over smaller companies in the race for hyperscale AI model development because the technology is unobtainable with innovative ideas alone.

“Rather than competing against big global companies in the same field, we want to focus on a vertical AI service with a Korean-language AI model (developed by Kakao Brain)," he said.

Vertical AI refers to AI that is highly optimized and applied to a specific problem in a certain industry. On the other hand, horizontal AI allows companies and startups in particular to quickly adapt modules to fit individual clients’ needs.

A Kakao official told The Korea Herald that the fields KoGPT would specialize in had not yet been decided.

But the CEO said that the company was looking to move swiftly to introduce its AI-powered vertical service later this year and improve its AI capability.

Kakao Brain, an AI technology research and development subsidiary of Kakao, introduced its Korean-language AI model in October last year. The company has also released various AI-related services.

On Thursday, Kakao’s AI unit unveiled Karlo, an AI image model that can understand users’ input and provides images in a variety of styles and colors.

Kakao's move comes after California-based startup OpenAI's release of ChatGPT, which is being hailed as a game-changer for various industries, with its ability to learn and adapt quickly to new information.

However, Kakao officials have not yet specified plans about whether it will apply the KoGPT model to its search engine Daum as Microsoft has done for its search engine Bing.

Kakao’s crosstown rival Naver also announced its plan to launch a new AI search service called SearchGPT in the first half of this year.

During the earnings call, Kakao Chief Investment Officer Bae Jae-hyun also highlighted KoGPT’s “high level of cost-effectiveness” as an advantage which would use fewer parameters than competing AI models to achieve the same level of performance.

When asked about the amount of investment put into the KoGPT service, the CIO said: “We will not make a significant increase in the level of investment (on KoGPT) from our previous investments on other AI projects made through Kakao Brain and Kakao Enterprise.”

Before Friday’s earnings call, the tech behemoth released its 2022 earnings report before the opening on the Korea Exchange.

Kakao said its 2022 operating profit fell 2.4 percent from the same period a year before to 508.5 billion won ($402 million). The company marked its first negative growth in four years, recording 8.2 percent from the previous year's 9.7 percent.

The company's net profit came in at 1.02 trillion won on a consolidated basis in 2022, down 38 percent from the previous year’s profit of 1.64 trillion won. Its annual sales, however, rose 15.8 percent on-year to 7.1 trillion won in the same year.

For the October-December period, Kakao posted 100.4 billion won in operating profit, down 5.8 percent from a year prior. The figure was slightly higher than the consensus of 96 billion won of local analysts provided by market intelligence FnGuide.

The company’s sales lost 0.6 percent to 1.77 trillion won, while it posted a net loss of 539.3 billion won over the cited period, falling into the red from the same period last year’s profit of 224.1 billion.

Kakao attributed its poor performance last year to slow growth connected to a global economic slowdown, as well as the compensation fees and aftereffects of the massive service disruptions caused by a data center fire in October.