Kakao Brain Opens 'Koyo', Large-Scale AI Development Resource, to External Companies

[Asia Economy Reporter Kang Nahum] Kakao Brain is opening the largest image-text dataset in Korea to AI development companies to expand the base of large-scale artificial intelligence (AI) research and development.

Kakao Brain announced on the 30th that it will release the dataset "Coyo," consisting of approximately 740 million image-text pairs, on its official website on the 31st. Datasets are one of the key elements necessary for large-scale AI models to produce precise results. To achieve high performance, many companies manually pair images and texts to collect datasets, which involves high costs and significant time investment.

When developing "Coyo," Kakao Brain effectively reduced the costs and time involved by automatically collecting image-text pairs online using proprietary technology, while simultaneously selecting high-quality data to achieve superior performance. The quality of the data was confirmed through cross-validation by reproducing large-scale AI models released by leading global AI companies.

"Coyo" was applied in the development of Kakao Brain's previously released large-scale AI image generation model "RQ-Transformer" and AI artist "Karlo." The "RQ-Transformer," released in April, was comprehensively recognized for its technological excellence and was given the opportunity to present its paper at the world-renowned academic conference CVPR 2022 held in June. Recently, through collaborations with contemporary artist Go Sang-woo and Samsung Electronics' "Galaxy Book Art Project," the potential of the AI artist "Karlo" was demonstrated.

Kakao Brain plans to release an additional large-scale AI model utilizing the "Coyo" dataset in the first half of next year. Kim Il-doo, CEO of Kakao Brain, said, "The release of this dataset will be an important foundation and milestone to accelerate the development of large-scale AI technology," adding, "Going forward, Kakao Brain will continue to lead technological leadership in various fields, collaborate with the AI community, and contribute to the development of the AI ecosystem."

This content was produced with the assistance of AI translation services.