RLWRLD Secures $14.8M to Develop Foundational AI Model for Advanced Robotics

RLWRLD raises $14.8M to build a foundational model for robotics | TechCrunch

As robotics has advanced, industry has steadily adopted more robots to automate away many kinds of grunt work. More than 540,000 new industrial robots were installed worldwide in 2023, taking the number of total industrial robots active to above 4 million, per IFR. Industrial robots typically excel at repetitive tasks, but they find it challenging to perform precise tasks, handle delicate materials, and adjust to changing conditions -- a robot in a restaurant's kitchen would get in the way more than be helpful, for example. That is why many industrial processes are still manual. South Korean startup RLWRLD aims to solve this problem with a foundational AI model that it has built specifically for robotics by combining large language models with traditional robotics software. The company says this model will enable robots to make quick and agile movements and perform some amount of "logical reasoning" as well. "Using RLWRLD's foundation model, processes that require a lot of manual work can be completely automated by learning and copying human expertise, making work environments more efficient," Jung-Hee Ryu, founder and CEO of RLWRLD, said in an exclusive interview with TechCrunch. The startup is now coming out of stealth with 21 billion KRW (about $14.8 million) in seed funding. The round was led by venture capital firm Hashed, and Mirae Asset Venture Investment and Global Brain also invested. Notably, RLWRLD has attracted a long list of big strategic investors -- Ana Group, PKSHA, Mitsui Chemical, Shimadzu and KDDI from Japan; LG Electronics and SK Telecom from Korea; and Amber Manufacturing from India. RLWRLD said the seed funding will be used to fund proof-of-concept projects with its strategic investors; secure computing infrastructure like GPUs, purchase robots, and devices to collect extensive data; and hire top research talent. The startup will also use the new money to develop advanced hand movements involving five-fingers -- a capability that's not yet been demonstrated by its competitors like Tesla, Figure AI and 1X, Ryu said. Ryu said RLWRLD is also working with its strategic investors to explore ways to automate different human-centric workflows using its AI model. They are together preparing a humanoid-based autonomous action demonstration, scheduled for later this year, Ryu said. In addition, the company is working to develop a platform that can support various kinds of robots, including industrial, collaborative, autonomous mobile robots and humanoids. Founded in 2024, RLWRLD is Ryu's third startup. His second startup, Olaworks, was acquired by Intel in 2012, and eventually became Intel's Korea R&D center within its computer vision division. And in 2015, he founded a startup accelerator, Future Play, that focuses on deep tech companies. When asked what inspired him to start a new company again, Ryu said he noticed how quickly AI startups were increasing in number in the U.S, Europe and China, while comparable AI startups in Korea and Japan were relatively absent. He spoke with over 30 AI professors from Korea and Japan about their challenges -- everything from the lack of infrastructure like data and GPUs, and the obstacles that discouraged them to launch a venture -- and the opportunities available. "I determined that it would be strategically beneficial to prioritize robotics foundation models (RFM) over the technologically saturated field of LLMs, capitalizing on Korea and Japan's notable global strengths in manufacturing," he said. Soon afterwards, he brought on board six professors from top-ranked institutions in South Korea, including KAIST, SNU, and POSTECH, along with their research teams, to launch RLWRLD. RLWRLD isn't alone in tackling this problem. Startups like Skild AI and Physical Intelligence are building similar foundational models for robotics, as are larger firms like Tesla, Google DeepMind, and Nvidia. But Ryu believes his startup has a good start, as it already has the AI and robotics experts it needs to develop foundational models for robotics, as well as humanoid robots with high degree of freedom (DoF). "Additionally, [such companies] typically rely on low-DoF robots such as two-fingered grippers. RLWRLD has already secured a high-DoF reference robot, and therefore expects superior performance outcomes," he said. Ryu also said that thanks to its strategic investors, RLWRLD can quickly gather valuable data from manufacturing sites located nearby. In 2024, a report indicated that Japan and South Korea collectively accounted for 9.2% of worldwide manufacturing production. RLWRLD aims to generate revenue as early as this year through proof-of-concept (PoC) projects and collaboration demonstrations with strategic partners. The startup's long-term goal is to cater to factories, logistics centers, and retail stores, and even robots that can be used in domestic environments to help with household chores. In the meantime, the priority is to target industrials since they are willing to pay the most and have strong demand for automation.

South Korean RLWRLD Secures $14.8 Mn to Advance Robotics AI Model | AIM Media House

The funding round, led by venture capital firm Hashed, also included investments from Mirae Asset Venture Investment and Global Brain. The startup's strategic partners include LG Electronics, SK Telecom, Mitsui Chemical, Shimadzu and KDDI from Japan and Amber Manufacturing from India. RLWRLD aims to address limitations in industrial robotics by combining large language models with traditional robotics software. Founder and CEO Jung-Hee Ryu explained, "Using RLWRLD's foundation model, processes that require a lot of manual work can be completely automated by learning and copying human expertise, making work environments more efficient." The funds will support proof-of-concept projects with strategic investors, infrastructure acquisition like GPUs and robots, and the development of advanced five-fingered hand movements. The company also plans to demonstrate humanoid-based autonomous actions later this year while working on a platform to support various types of robots, including industrial and collaborative models. RLWRLD was founded in 2024 by Ryu, who previously launched Olaworks (acquired by Intel) and FuturePlay, a deep tech accelerator. RLWRLD also aims to generate revenue this year through collaboration demonstrations and projects targeting industrial sectors with high demand for automation. Looking at building AI for robotics, other companies in the US have also achieved developments. This began in February this year, when Figure AI announced Helix, a vision-language-action model that allowed humanoid robots to perform complex tasks using natural language. Watch the robot pick up virtually any object by natural prompting, like 'Pick up the desert item' here. Following this, in March, Google DeepMind built the next generation of humanoid robots with its new AI model, Gemini Robotics, which enables Apptronik's Apollo to understand natural language and perform real-world tasks in environments it has not seen before. In the same month, NVIDIA unveiled Isaac GR00T N1, its first generalist foundation model for humanoid robots, and made it open-sourced. GR00T N1 enables robots to reason, plan, and act in complex environments, combining 'slow thinking' with reasoning and perception and 'fast thinking' with precise and real-time actions. The giant also introduced Newton, an open-source physics engine for robotics simulation developed in collaboration with Google DeepMind and Disney Research.