Kakao Corp., South Korea’s premier mobile platform operator, has made a significant leap in the field of artificial intelligence by developing a multimodal large language model (MLLM) called Honeybee. This latest innovation, with advanced image-to-text technology, places Kakao alongside global Big Tech firms in AI development.
During a strategic meeting on artificial intelligence hosted by the Ministry of Science and ICT, Kakao’s CEO nominee, Chung Sina, announced the completion of Honeybee. Developed by Kakao Brain Corp., an AI research affiliate, Honeybee is a groundbreaking tool capable of interpreting and reasoning across multiple modes including text, images, video, audio, and coding.
In a collaborative and open-source spirit, Kakao has made Honeybee’s inference source code available on GitHub. This move allows developers worldwide to access and utilize the model in their AI research, fostering innovation and knowledge sharing in the AI community.
Honeybee operates on a similar level to Google DeepMind’s Gemini and OpenAI’s GPT-4. These are multimodal LLMs known for their extensive multitask language understanding capabilities. Honeybee’s design enables it to comprehend both images and text, offering answers to complex questions that often surpass human performance.
For instance, when asked to analyze a photograph and determine the number of victories by a basketball player, Honeybee can generate precise answers, demonstrating its advanced AI capabilities.
Kakao has been actively involved in AI development, having previously introduced KoGPT, a Korean LLM, in 2021, and its subsequent upgrade, KoGPT 2.0. Based on OpenAI’s GPT-3, KoGPT is recognized as the most potent Korean language chatbot, primarily trained on Korean text.
The recent government-private sector AI meeting, attended by leaders of major Korean tech companies, highlighted the need for government support in advancing AI technology. The meeting underscored the importance of AI in shaping societal thinking and behavior, emphasizing the collaborative effort needed between public and private sectors for AI-based national growth and innovation.