Aim Intelligence, LG Electronics, and Openmind Jointly Unveil AI Robot 'Video2Robot' Solution

Robots can now learn safely in advance. AI safety startup Aim Intelligence has unveiled the 'Video2Robot' pipeline, opening a new chapter in robot safety research through collaboration with LG Electronics and Openmind, a company specializing in robot OS. This innovation, under the concept of 'Physical AI Safety,' aims to dramatically enhance safety by proactively verifying potential risks that robots may encounter in reality during the simulation phase. The open-source based 'Video2Robot' serves as a core engine, meticulously converting natural human movements from video into robot learning data. A highlight of this technology is its utilization of generative AI. It allows for the free creation of challenging scenarios in a virtual space, such as falls, collisions, and loss of balance – situations that are difficult or even dangerous to reproduce in real environments – thereby enabling the acquisition of a large volume of 'failure data'. This allows robots to undergo 'preemptive learning' for numerous risky situations before actual accidents occur, effectively resolving the limitations and immense cost issues associated with traditional physical testing. Aim Intelligence, which has focused on multimodal AI safety research, has successfully expanded its software-centric capabilities into the realm of solving physical robot safety problems through this 'Video2Robot' achievement. Immediately after its public release, it garnered over 400 stars on GitHub and more than 200,000 views on social media, demonstrating global interest. Moving forward, the company plans to build a more comprehensive 'Physical AI Safety' framework to contribute to robots integrating more safely into our lives.