Genesis AI is a startup aimed at building foundational models for running all kinds of robots, and has emerged from stealth with a huge $105 million seed round co-led by Eclipse Ventures and Khosla Ventures.
Founded last December by Zhou Xian (pictured above, left), he received his PhD in Robotics from Carnegie Mellon University and founded Théophile Gervet (pictured above), a former research scientist at the French AI Lab Mistral.
Large language models are trained on vast datasets of text, whereas robotics AI models need to be trained on data in the physical world. However, getting real-world data is a costly and time-consuming effort.
To overcome this, Genesis is turning to synthetic data. This is generated using a unique physics engine that can accurately model the physical world.
Genesis’ Synthetic Data Engine was born from an academic project led by Xian in collaboration with researchers from 18 universities. Several participants in that project participated in Genesis and comprised the current staff of over 20 researchers specializing in robotics, machine learning and graphics.
Genesis claims that its unique simulation engine allows models to be developed faster. This is a clear advantage over competitors that rely on Nvidia’s software.
Other companies working on developing general-purpose AI models for robots include physical intelligence, which raised a $400 million round, and Skild AI, which was valued at $4 billion earlier this year.
“It’s a big unknown. Does anyone have a big robotics basic model that generalizes across tasks?
“Of all the teams we’ve seen, we like it [Genesis’] “An approach to chasing the Robotics Foundation model,” she added.
Genesis develops synthetic data and builds basic models in two offices, Silicon Valley and Paris.
As the next milestone, Genesis plans to release its model to the Robotics community by the end of the year.
