In recent years, artificial intelligence has thrived on abundant digital content such as text, images, and videos found online. However, the landscape shifts dramatically when discussing robotics. Unlike traditional AI, which focuses on processing information, robotics involves interacting with the physical world\u2014moving objects, manipulating parts, and executing real-time actions. This necessitates a different kind of foundational data, and the next AI race appears to be branching out into new territories.<\/p>\n
Config, a startup located in Seoul and San Jos\u00e9, has garnered significant interest from some of South Korea’s largest corporations, including Samsung, Hyundai, and LG. According to Foley Hoag<\/strong>, which provided legal advisory for Config’s recent seed round, the startup has successfully raised $27 million<\/strong> in funding. This operation, led by Samsung Venture Investment<\/strong>, includes contributions from other key players like ZER01NE Ventures and LG Technology Ventures. The funding has elevated Config’s valuation to over $200 million<\/strong>, bringing its total capital raised to $35 million<\/strong>.<\/p>\n Config aims to establish itself in the value chain similarly to TSMC in the semiconductor industry. Rather than competing with end-users, Config seeks to be the critical supplier of foundational data that other manufacturers require for developing their robotic AI solutions. This model is increasingly appealing to large manufacturers keen on reducing reliance on external vendors.<\/p>\n One of the most significant differentiators for industrial robotics is the nature of data collection. While language models predominantly deal with digital information, robotics data must be collected through physical actions. As Config\u2019s CEO, Minjoon Seo<\/strong>, noted, this necessitates robots, dedicated testing spaces, and human teams to enable effective data gathering. With the demand for more capable robotic machines on the rise, the costs associated with data labeling and collection also escalate, as it involves real-world tasks instead of readily accessible online information.<\/p>\n Config differentiates itself by advocating a unique approach to data utilization. Many teams typically use human motion data and then adapt it for machine learning. In contrast, Config focuses on pre-processing that data to ensure it aligns better with robotic behavior and interaction. This proactive strategy optimizes training efficiency and enhances machine learning outcomes.<\/p>\n Currently, Config employs nearly 300<\/strong>, tirelessly focused on data production. The startup claims to have amassed over 100,000 hours<\/strong> of human movement data, significantly overshadowing competitors, which have gathered merely about 3,000 hours<\/strong>. This impressive scale is a testament to Config’s commitment to data operations.<\/p>\n Looking forward, Config plans to expand its operational footprint in Vietnam<\/strong> and Seoul<\/strong>, aiming to accumulate one million hours<\/strong> of data. The company also projects that its enterprise platform will achieve $10 million<\/strong> in annual recurring revenue by the end of 2027<\/strong>. Additionally, a forthcoming Robot-as-a-Service<\/strong> model is set to allow businesses to leverage Config’s foundational models through the cloud, eliminating the need for integrated hardware.<\/p>\n The efforts and innovations by Config signal promising developments in the robotics field. The emphasis is shifting towards the groundwork that translates human actions into machine learning, demonstrating that successful robotics will hinge not just on the machines themselves but on the comprehensive data and training that underlie them. Although Config is still in its infancy, the backing from heavyweights like Samsung, Hyundai, and LG hints at a future rich with potential.<\/p>\nAspiring to Be the “TSMC” of Robotics Data<\/h3>\n
The Unique Challenges of Robotic Data Collection<\/h3>\n
Innovative Approaches to Data Transformation<\/h3>\n
Scalability and Future Aspirations<\/h3>\n
The Path Ahead<\/h3>\n