[NE Coffee Chat] iChemist Targets Synthetic Data Market with 3D Digital Twin

Generating Synthetic Data with 3D Modeling
No Inbreeding Concerns Unlike AI-Generated Content
Reduced Consulting Costs Ease Financial Burden

In the era of artificial intelligence (AI), competitiveness depends on who can secure higher quality data. However, there are limitations to training AI using only data collected from reality. Elon Musk, CEO of Tesla, stated, "The total amount of knowledge accumulated by humanity has already been exhausted in the AI training process," and proposed 'synthetic data' as a solution. This refers to data artificially generated through computer programs. It also includes images and videos created by generative AI. However, synthetic data created by AI carries the problem that learning from data with the same pattern can degrade AI prediction performance.

To overcome these limitations, iChemist developed 'CEN,' a synthetic data generator applying 3D digital twin technology.

Jung Min-wook, CEO of iChemist, explained in an interview with Asia Economy on the 25th, "When using images created by existing generative AI as training data, labeling work is necessary, but with CEN, data refinement is automatically performed the moment images are rendered based on 3D models." While existing generative AI creates new images by transforming existing data, CEN directly implements the scenes users want through 3D modeling. For example, if one wants to generate port images, instead of simply synthesizing existing images, even moss attached to rocks is precisely realized in 3D.

Jung Min-wook, CEO of iChemist, is participating in CES 2024, the world's largest electronics and IT exhibition, introducing the company's solutions. Photo by iChemist

The advantage of this technology is that it can generate data not found in existing big data. CEO Jung said, "The Agency for Defense Development requested about 1,000 images of the K2 tank at night, but we could not find K2 images in existing data," adding, "Using 3D digital twin technology, we directly generated 10 data items per second."

Currently, global companies such as NVIDIA and Unity also operate 3D synthetic data generation platforms. However, according to CEO Jung, these platforms are systems for 3D experts, making the entry barrier high for data engineers. He explained, "To efficiently utilize overseas platforms, a deep understanding of 3D technology is required, but CEN is designed so that data engineers can use it more intuitively."

iChemist focused on lowering the entry barrier for data production platforms while also reducing costs. CEO Jung said, "The annual usage fee for NVIDIA's 'Omniverse' platform alone reaches about 20 million KRW," adding, "Usually, synthetic data platforms incur high consulting costs, but CEN reduces cost burdens by allowing customers to perform the verification process after data generation themselves."

iChemist aims first to expand its market share in the domestic synthetic data market. As of last year, the domestic data market size was 1.4 trillion KRW, with the synthetic data ratio estimated to have increased to 40%. CEO Jung said, "Starting this year, we plan to expand B2G (business-to-government) operations to further increase sales," adding, "Our long-term goal is to collaborate with global information technology (IT) companies to directly provide the CEN platform."

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