Pi Network is exploring the capability of Pi’s global network of Nodes to support decentralized AI training and computing tasks. Pi Nodes have long been designed to do more than securing a ledger. Although the Pi blockchain benefits from the decentralized nature of many distributed nodes around the world, Pi’s ledger itself is energy efficient and does not require the full computational resources of such a large distributed Node network. That creates a pool of unused computing capacity that can enable new utilities such as decentralized cloud computing.
Third-parties requiring computing power for AI model training could utilize these unused resources of Pi Node operators, who opt in, and pay the operators in cryptocurrencies. Utilities on Pi Nodes constitute an important component of Pi utility, alongside Pi apps, platform-level utilities and local commerce. These elements provide new approaches to AI infrastructure that allow a distributed system to capture production into its network and enable people to contribute to AI’s production processes directly, and receive compensation through blockchain-based payments in return. OpenMind is developing an operating system and open-source protocol for robots to think, learn, and work together—like Android OS for robots.
To test the feasibility of Pi’s distributed computing, OpenMind developed a container that could request computing tasks from individual computers. The team shared this container with volunteer Pi Node operators who downloaded it to run on their machines. OpenMind then sent tasks through the container requesting the computers to process different images using OpenMind’s AI image recognition model—an example use case where training for image recognition is essential for OpenMind’s robots to interact with the real world. The goal was to discover as many discrete objects as possible within these images, using the volunteer Pi node operators’ computing power to do so.
Leave a Reply