[AI Revolution](23) 'Synthetic Data' Company Training AI... C&AI

The performance of artificial intelligence (AI) is determined by the amount of 'study.' The AI of autonomous vehicles must learn from a sufficient amount of video data covering situations that may be encountered on the road. Medical AI also needs to learn the locations and shapes of various lesions to function properly. Insufficient study can lead to accidents or misdiagnoses. This highlights the importance of training data for AI. However, actual data is severely lacking. This is why 'synthetic data' has emerged in the AI era. Synthetic data is virtual data created to train AI. Computer algorithms generate it infinitely by reflecting the characteristics of real data. In this market, the startup CN.AI is regarded as a company with the most advanced experience and technology.

On the 26th, CN.AI announced plans to launch the synthetic data generation platform 'CN Flow.' 'CN Flow' is a one-stop platform that consolidates previously fragmented image generation technologies. Even non-experts can easily and quickly generate synthetic data or image content. It can generate synthetic data for various situations to be used in AI training, and content creators can produce the images they need. Currently, to prepare for the official launch, they are conducting a proof of concept (PoC) for image content generation business with leading domestic companies.

An image of a drone flying in the sky created by selecting a background image and a drone image in 'CN Flow'

원본보기 아이콘

CN.AI was founded in 2019 by CEO Wonseob Lee, a former Samsung Electronics employee. CEO Lee focused on the fact that synthetic data can solve the data shortage problem, which was a bottleneck in AI advancement. AI performance improves with more training data, but acquiring data incurs significant costs. Real data collection and use are restricted due to privacy issues, and labeling the data requires human intervention, often referred to as the '21st-century doll eye sticking.'

CN.AI explained that synthetic data technology, which produces quantitatively and qualitatively enhanced data, can reduce the data costs that account for the largest expense in building AI models. It enables data generation tailored to customer needs and creates new data without issues such as privacy infringement. This is why MIT Technology Review selected synthetic data as one of the top 10 breakthrough technologies last year. The global research firm Gartner predicts that by 2026, synthetic data used for AI training will surpass real data. The global market size is also expected to reach $26.1 billion next year and expand to 580 billion KRW domestically during the same period.

Based on its expertise in synthetic data specialized in image generation, CN.AI has carried out projects across various industries. It has supplied synthetic data to government agencies, educational institutions, and medical institutions, including large corporations. Leveraging its expertise in image data generation, it has also ventured into the generative AI business. CN.AI’s generative AI technology accurately creates images described by users in text and combines multiple existing images to produce new ones. It can also modify existing images to suit user preferences. CEO Lee said, "We will lead the With AI era where everyone uses high-performance AI," adding, "Our goal is to provide an AI-based image and video generation platform based on the CN Flow platform, which is grounded in AI synthetic data."

© The Asia Business Daily(www.asiae.co.kr). All rights reserved.