Kakao announced that it will participate in the 'Asia-Pacific Economic Cooperation (APEC) 2025 Korea Global Digital and AI Forum and Corporate Exhibition' to be held on the 5th, introducing two proprietary technologies.
The forum, held as a side event of the APEC Digital and AI Ministerial Meeting, will take place at the Sheraton Grand Hotel in Songdo, Incheon. The corporate exhibition, taking place outside the forum, will feature about 10 corporations, including Kakao and other domestic and international IT companies.
Kakao will showcase its self-developed ▲ lightweight multimodal language model Kanana-1.5-v-3b and ▲ AI guardrail model Kanana at the corporate exhibition booth set up at the venue. Attendees will be able to experience the functions and features of the models through demonstration spaces.
The Kanana-1.5-v-3b, released as open source on the 24th of last month, is a lightweight multimodal language model capable of processing not only text but also image information. It is characterized by its high directive execution performance, accurately understanding the intent of questions, and exceptional Korean and English image comprehension abilities.
The AI guardrail model Kanana Safeguard is based on a language model developed by Kakao and utilizes a self-constructed dataset that reflects the Korean language and culture, providing performance specialized in the Korean language. It consists of three different models and is explained that it can effectively detect harmfulness and risks according to the type of risk.
Kanana Safeguard serves as a technical and institutional mechanism that can defend against risks such as harmful content amid the spread of various generative AI services, and it was released as open source in May, becoming the first of a domestic corporation to contribute to the ecosystem.