Genesis AI Raises $105M to Develop AI Models for Robotics

Genesis AI Secures $105 Million Seed Funding

Genesis AI, a newly launched startup focused on creating a core model for diverse robotic applications, has exited stealth mode. The company has successfully raised $105 million in a seed round co-led by Eclipse and Khosla Ventures.

Founding and Vision

Established last December, Genesis AI was founded by Zhou Xian, a robotics PhD graduate from Carnegie Mellon University, and Théophile Gervet, previously a research scientist at the French AI lab Mistral.

The startup’s objective is to construct a versatile model capable of automating numerous repetitive tasks for robots. These tasks span various sectors, including laboratory operations and domestic chores.

The Challenge of Robotics Data

While large language models excel with text-based data, robotic AI requires training on information pertaining to the physical environment. Gathering this real-world data presents significant challenges in terms of both cost and time investment.

Genesis’ Solution: Synthetic Data

To address this hurdle, Genesis AI is leveraging synthetic data. This data is generated through a unique physics engine designed to accurately simulate the complexities of the physical world.

Origins and Expertise

The synthetic data engine was initially developed as an academic project spearheaded by Xian, in collaboration with researchers from 18 universities.

A substantial number of participants from this original project have now joined Genesis, forming a team of over 20 specialists in robotics, machine learning, and computer graphics.

Competitive Advantage

Genesis AI asserts that its proprietary simulation engine facilitates faster model development. This provides a notable advantage compared to competitors who depend on Nvidia’s software solutions.

Industry Landscape

Several other companies are also actively developing general-purpose AI models for robotics. These include Physical Intelligence, which secured $400 million in funding, and Skild AI, recently valued at $4 billion.

Investor Perspective

Kanu Gulati, a partner at Khosla Ventures, emphasized the uncertainty surrounding the development of a broadly applicable robotics foundation model. She stated, “It’s a big unknown: Will anybody have a large robotics foundation model that will generalize across tasks? That’s a bet we want to go after.”

Gulati further added, “Of all the teams we have seen, we like [Genesis’] approach for going after robotics foundation models.”

Global Presence and Future Plans

Genesis AI is currently developing its synthetic data and foundational model across two locations: Silicon Valley and Paris.

The company intends to release its model to the broader robotics community before the end of the current year, marking a significant milestone in its development.