Frodobots raises $6M for Solana-based robotics network
“If we were to be successful, we would be on par with the Teslas and the DeepMinds of the world,” Frodobots’ co-founder said
Frodobots and Adobe stock modified by Blockworks
This is a segment from the Lightspeed newsletter. To read full editions, subscribe.
Robotics and crowdsourced data startup Frodobots has raised $6 million in seed funding to create BitRobot, which will be a Solana-based embodied AI research network.
BitRobot will aim to connect different subgroups in robotics research using a shared token and network to solve resource allocation and data scarcity. Protocol VC led the funding round, which also drew participation from Big Brain Holdings, Solana Ventures, Virtuals Protocol and Solana co-founders Anatoly Yakovenko and Raj Gokal, among others. Frodobots previously also raised $2 million in pre-seed funding.
Frodobots started out building sidewalk robots that slowly trawled global city streets to build giant crowdsourced datasets including things like audio, video and GPS data. The project aims to add token incentivization to become part of the decentralized physical infrastructure (DePIN) movement.
Over the past several months, however, Frodobots’ ambitions have widened: The lab’s co-founder Michael Chung Yeung Cho said the goal is now to launch BitRobot and “solve embodied AI.” Frodobots will be a subnet in the BitRobot network.
“If we were to be successful, we would be on par with the Teslas and the DeepMinds of the world,” Cho said, adding that the new project would help generate a “series of really performant foundational robotics models that can be deployed across all kinds of robots.”
The problem, Cho and his co-founder Jonathan Victor explained, is that rich robotics data is scarce, and robot developers are working on subsets of a bigger shared problem. Large language models have the entire internet to train on, Victor said, but most robotics businesses outside of Tesla’s self-driving unit are “starting effectively from zero.”
That’s partly because while a new AI model, for instance, could be benchmarked digitally in an hour, robotics faces the bottleneck of having to test products in the real world. You can’t know if a self-driving car really works until you drive it for a trillion miles, Cho said.
BitRobot will create a shared network of “subnets” so that pooled resources and data can more efficiently create robotics data. Humans, fleets of humanoid robots and Frodobots’ sidewalk bots are example subnets, according to a document shared with Blockworks.
With BitRobot, massive amounts of resources could be shifted between researchers based on need, and a DePIN token reward system could align the economic incentives. There is also cross-embodiment learning, which suggests that AI models trained on one type of robot — such as a robotic arm — can effectively transfer learned skills to entirely different robotic systems, like drones or wheeled robots. This works because many robotic tasks share core design needs.
A white paper containing more details on BitRobot, as well as the project’s tokenomics, will be released in a few weeks.
Get the news in your inbox. Explore Blockworks newsletters:
- The Breakdown: Decoding crypto and the markets. Daily.
- 0xResearch: Alpha in your inbox. Think like an analyst.
