Nvidia Plans Investment in Robot AI Developer RLWRLD

SEOUL -- US-based semiconductor company Nvidia said it will be investing in RLWRLD, a physical artificial intelligence company focused on building foundation models for humanoid and industrial robots able to manipulate objects using their five-fingered hands.

The announcement was made Nov. 9 during the AI Startup Partnership Forum in Seoul in cooperation with the South Korean Ministry of Science and ICT. No details were revealed regarding the amount of investment, valuation or completion of the deal.

According to South Korean business publication MoneyToday citing industry sources, RLWRLD is seeking 200 billion won ($140 million) in funding at a pre-money valuation of 1 trillion won.

Based in San Francisco, Seoul and Tokyo, RLWRLD is working on developing AI models which can help robots execute tasks that need the use of hands, such as inserting parts and fastening screws and handling objects while moving conveyor belts.

As reported by ROBOT24.com, RLWRLD has started developing its RLDX-2 foundation model which uses human motion data and tactile sense to train robots for dexterity in handling objects. The company has also collaborated with CJ Logistics for development and testing of robot AI in warehouse settings.

Nov. 9 saw the announcement by RLWRLD of the release of DexBench, a benchmark developed in partnership with Nvidia that tests robot dexterity in simulated industrial environments.

DexBench contains 15 manipulation tasks in four domains – retail, logistics, assembly, and hotel food and beverage. The benchmark tests the success of the task completion and robots’ capacity to adapt to novel objects and environments.

RLWRLD will release the code, training data and evaluation environments of the benchmark to researchers and developers, while physical testing on robots is still being considered.