Kakao launches Korea's first multimodal AI

Kakao has launched its multimodal AI model, Kanana, capable of processing text, voice, and images simultaneously.

Kanana is Korea’s first large language model (LLM) designed to handle diverse input types.

Users can interact with it using combinations of text, audio, and image queries. It responds in either text or voice, depending on the context.

Kanana features speech emotion recognition and analyzes non-verbal cues like intonation and voice patterns.

It has been trained on large Korean datasets, allowing it to understand regional dialects and translate them into standard Korean.

Kakao says the model performs well on both Korean and English benchmarks, with strong results in Korean tasks.

🔗 Source: The Korea Times

🧠 Food for thought

1️⃣ South Korea’s strategic AI push following the “AlphaGo shock”

Kakao’s Kanana-o represents the latest advancement in South Korea’s determined national AI strategy that began in earnest after 2016’s “AlphaGo shock,” when Google’s AI defeated Korean Go champion Lee Sedol.

This defeat prompted the South Korean government to announce an $860 million fund specifically for AI development, recognizing a critical technology gap with global leaders 1.

This pattern of government-led technological advancement has deep roots in South Korea’s economic development model, where the state has historically established research institutions and funding programs that laid groundwork for future innovations 2.

The competition between Kakao and Naver in developing Korean-language AI models reflects this national priority, with Naver pursuing a “Sovereign AI” approach customized by industry and culture, while Kakao focuses on practical, service-oriented AI applications 3.

South Korea’s cultural acceptance of technology as solution-oriented rather than threatening has created fertile ground for AI adoption, with the country previously ranking as the most innovative economy for four consecutive years as of 2017 4.

2️⃣ Multimodal AI represents the next frontier in human-computer interaction

Kanana-o’s ability to simultaneously process text, voice, and images positions it within a rapidly expanding global multimodal AI market, projected to grow from $893.5 million in 2023 to $10.55 billion by 2031 at a CAGR of 36.2% 5.