Kakao Speeds up ChatGPT Competition… Command→Create video creation AI
Kakao-Korea University development ‘Flame’
Virtual human operates according to commands
Announcement of the world's most advanced AI conference
Video creation, meta and Google also developed
Kakao unveiled a new artificial intelligence (AI) technology that turns text into video. While foreshadowing that they will participate in the AI competition between Big Techs, which has intensified with ChatGPT, they are trying to increase their competitiveness by securing new high-level technologies.
According to the information technology (IT) industry on the 14th, Kakao Brain, an AI-specialized subsidiary of Kakao, held a meeting on the 9th of this month (local time) at the world's most prestigious AI society, the American Academy of Artificial Intelligence (AAAI). An oral session was presented on a jointly developed text to motion (text to motion) AI 'FLAME'. Oral presentations are made by selecting papers that have been highly evaluated by AAAI.
Frame utilizes super-giant AI (AI with greatly increased computational power based on large-scale infrastructure), and when a user inputs a command, the corresponding action is implemented through a virtual human. For example, if you enter “a person who kicks with the right foot,” the virtual human on the screen actually lifts his right foot and kicks. The company explained that even when complex sentences are entered, it creates high-quality movements. Kim Il-doo, CEO of Kakao Brain, said, "Flame will make a great contribution to work automation and work efficiency improvement in areas where motion graphics production is required, such as games, animations, and movies."
Since text-to-video (text to video) AI, which includes Frame, is a technology that has not yet been commercialized worldwide, Kakao Brain can also try to preoccupy the market. In the second half of last year, Meta released its own AI called 'Make-A-Video' and Google called 'Imagen Video'. Professor Choi, who jointly developed Flame, said, “Many companies are focusing on text-based video creation,” and “we will create differentiated research results.”
Following Korean model 'KoGPT' and text-image multi-modal (a model that learns text and images simultaneously) 'Karlo', Kakao Brain is expected to further enhance its competitiveness in super-giant AI through the development of Flame. . In particular, as a latecomer to the competition, it seems to be trying to compete in the multimedia field such as images and videos, rather than a language model that is largely pushed aside by ChatGPT overseas and HyperClova (Naver) in Korea. B^DISCOVER, an image creation application using Calo, was commercialized in October last year, and is about to release a paid version for professionals in the first quarter of this year. Kahlo is also introduced in the creation of KakaoTalk profiles and wallpapers. KoGPT will be commercialized within this year as a vertical (specialized field) service such as an AI assistant and marketing tool.
