PrismaX is a platform designed to address physical AI and robotics industry challenges of data, models, and scaling. | Source: PrismaX
Robotics developers need reliable data to build artificial intelligence models for training robots. PrismaX, which is building a robotics teleoperations platform, this week launched after completing an $11 million funding round.
“Despite steady inflows from venture capital, the physical AI and robotics industry has been unable to create an incentive structure that yields high-quality and affordable data,” the company claimed. PrismaX said it is developing a standard for fair use in the industry in which the data that powers models generates revenue, which is then returned to the communities that create the data.
“Scalability of visual datasets is hindering breakthrough advancements in robotics,” said Bayley Wang, co-founder and CEO of PrismaX. “That is why we are creating the ecosystem that will produce a dataset of diverse scenarios, highly generalizable compared to centralized data, which benefits from improved scalability and reduces data-collection bias.”
“Through our decentralized incentive structure, PrismaX will create the path towards fully autonomous robots,” he added. “Our platform will allow people to work with AI rather than be replaced by it. As the industry advances, it is important to remember that human capital is required to build the models necessary to continue scaling.”
PrismaX was founded this year by Wang and Chyna Qu, who bring together experience in both robotics and decentralized technologies. The San Francisco-based startup is building and scaling foundational models to power advances in physical generative AI.
Three pillars to support data flywheel
PrismaX said it focuses on the following three foundational elements of the robotics industry:
- Data: The company plans to build protocols and mechanisms to validate and incentivize visual data, enabling robotics datasets to scale to the same levels as text data and to improve accuracy and reliability.
- Teleoperation: PrismaX aims to define a uniform standard for teleoperation, providing turnkey access to operators, payments, and software so robotics companies can focus on the things that make them different.
- Models: Through collaboration with leading AI teams, the company aims to build models to power increasingly autonomous robots. This would amplify the impact of PrismaX’s network by allowing operators to replace multiple physical workers, all while increasing the quality and quantity of the data collected.
The company claimed that these pillars will create a “data flywheel” in which large-scale data leads to better foundation models. This, in turn, would boost teleoperation efficiency, which then fuels more real-world data collection, thereby completing the loop.
Physical AI and robotics firms are currently sinking time and money into efforts to scale teleoperations teams with little breakthrough advances, PrismaX said. By developing a platform with robust teleoperation standards, the company said it will help physical AI businesses to scale quickly and reduce the costly redundancies occurring as a result of data being gathered in silos across the industry.
Now accepting session submissions!
PrismaX plans to build out fleet
“Humanoid robots are largely stuck in the lab, but PrismaX believes that, conceptually, they can do much more,” it said. “They could be used for tasks such as folding laundry, assembling fast-food orders, or administering drugs in hospitals.”
a16z CSX led PrismaX’s funding round. The Stanford Blockchain Builder Fund, Symbolic, Volt Capital, Virtuals Protocol, and other angel investors also participated. The completion of this fundraising coincided with PrismaX’s official launch at a16z’s CSX Demo Day on June 3.
PrismaX plans to use the investment to build out its fleet, hone its teleop standards, and expand its data-collection portal. It said it will give robotics firms access to a community of teleoperators who can help scale their visual datasets and build smarter machines faster.
“In the short term, PrismaX is looking to attract AI enthusiasts drawn to novel data-collection challenges and their potential impact,” stated the company. “These participants will be rewarded for their contributions and form the foundational models invaluable to robotics firms seeking data for training runs.”
The post PrismaX launches with $11M to scale virtual datasets for robotics foundation models appeared first on The Robot Report.