Physicl Raises $10M To Enhance Robot Training

Physicl, a new venture founded by Alex de Vigan, has secured $10 million in seed funding led by Inovia Capital. The company aims to solve training problems for robots with more realistic simulations. De Vigan's previous business, nfinite, was impacted by the rise of generative AI.

De Vigan spent the last decade building nfinite, which created 3D models of goods for e-commerce. However, with the arrival of transformer-based AI models, the company shifted its focus to generative imagery. De Vigan saw an opportunity to repurpose the technology for robot training.

Researchers have been struggling to train robots using video and remote operations due to the expense and limited returns. Simulations in digital worlds are a potential solution, but creating faithful representations of real-world physics has proven challenging. Physicl's platform may address this issue.

Physicl is entering a crowded market of startups providing data and tooling for foundation model builders and robot operators. The company plans to focus on post-training models on difficult edge cases. Its product, Studio, offers 10,000 3D environments and 100,000 assets for generating synthetic data.

Studio allows users to create scenes for training robots on specific problems, such as discerning objects in cluttered environments. SpAItial, a world model builder, has collaborated with Physicl to improve the fidelity of its products. Physicl's goal is to provide tooling for ongoing synthetic data generation, enabling customers to feed in vast amounts of data and generate high-quality 3D representations.

Physicl's platform may enable customers to auto-generate data on the fly, depending on the training state. This could significantly reduce the time and effort required for data curation. The company's technology has the potential to make robots more capable and useful in various tasks. This article is based on reporting by TechCrunch.