OpenMined is an encrypted artificial intelligence blockchain network. OpenMined assures to offer encrypted, decentralized artificial intelligence.
What is OpenMined?
OpenMined is an artificial intelligence project. It is based on blockchain concept. This project involves encryption and decentralization for building protocols. These protocols are fully open source.
OpenMined is an open source community. This community aims to build technology to facilitate the decentralized ownership of data and intelligence. The Github page lists a total of 111 people currently involved with the project.
AI systems are increasingly running the world. AI technologies are in the hands of powerful corporation. This will, in turn, lead to an environment where only these corporations can develop AI systems, leading to dangerous centralization of power. So, decentralized platforms like OpenMined are designed to bring back the power into the community. Thus, the ultimate aim of the OpenMined is to decentralise or democratize artificial intelligence development.
OpenMined has approximately 20 ongoing projects. Some of them are as follows:
- OpenMined Unity Application: It applies the PySyft library into a Unity application. Unity is free and portable. It has versatile GPU access. It can also be used for high end graphics gaming consoles like Playstation, XBox and many more.
- PySyft: A library for encrypted, privacy preserving deep learning. It is based on the PyTorch. PySyft is a Python library for secure, private Deep Learning. PySyft decouples private data from model training, using Multi-Party Computation (MPC) within PyTorch.
- MineUI: React application to interface with mine.js
- Capsule: The goal of this library is to allow developers of OpenMined to simulate participating in a network where others have access to secret information that developers do not. It is a service for storing and interacting with secrets, including keys, off the blockchain. In the trivial case, Capsule can generate a private Encryption key and never reveal it to the developer.
- Mine.js: Federated learning client and user data warehouse. In general, the mine script accepts js source and returns all the required call locations as well as the target string. This is a submodule of js-linker and my-filters.
- Sonar: Sonar observes all models being trained and ensures that occupation occurs fairly. It is a smart contract running on an Ethereum Blockchain that holds bounties and stores pointers to AI models on IPFS. It is a decentralized machine learning server hosted on the blockchain.
- PySonar: It is a decentralised machine learning client. There is PySonar2. It is a type inferencer for Python. It performs whole-project interprocedural analysis to infer types.
- CampX: It is an agent training environments that are entirely created using tensor operations. The API is a modified port of PyColab and should be loosely compatible with worlds created for PyColab.
- PyYASHE: It is an experimental library of developing a high performance YASHE implementation. PyYASHE is a Python implementation of the YASHE homomorphic encryption scheme. It is a improved security for a ring-based fully homomorphic scheme.
- PyBV: A Python implementation of the BV homomorphic encryption scheme. API can be imported using the command “import BV”.
- PyAono: PyAono is a python implementation of the homomorphic encryption scheme. A really unique feature supported by this proposal is the support for key switching. All the code is written in PARI library in C++. This code also has an independent header containing a PARI implementation.
- Private AI Resources: Private machine learning progress. This is a curated list of resources related to the research and development of private machine learning.
- Syft-autocomplete: A pseudo-real-world example of how Syft.js would be used in a production environment.
- Openmined Website: OpenMined website is community goal to accurately and beautifully explain the vision of the OpenMined project to all contributors past, present, and future.
- Serverless Website API: A Github statistics fetcher, running on a cron job, with permanent storage to DynamoDB, for the OpenMined community.
The website needs to gather some information rom Github, Slack, and Ghost to display on screen. Due to rate limiting and view count limitations it doesn’t make sense to always ask these services for this information. It is far quicker for us to cache this information elsewhere and retrieve it from there instead.
- OpenMined Ghost Theme: The theme for the OpenMined and Weekly Digs blogs. The default theme for all OpenMined Ghost blogs. This is based heavily on the default Ghost theme Casper.
- Grid: A peer-to-peer network for private, secure, multi-owner artificial intelligence. It is a p2p network of data owners and data scientists who can collectively train AI models using Syft.
- NALU-1: Basic pytorch implementation of NAC/NALU from Neural Arithmetic Logic Units.
- Adapters: Converting external data sources into OpenMined schemes. Adapters convert raw data dumps into a format usable by syft.
- OpenMined UI: The OpenMined UI library is to give UI developers one design language in the creation of OpenMined applications. The OpenMined react.js component library.
Finally, OpenMined projects are focused on encrypted and decentralized artificial intelligence. The aim is to build softwares which promote decentralization of data ownership, machine learning and deep learning libraries. These libraries can be used or built upon without the need of being an expert in either field.
The project focuses on three concepts. They are as follows:
- Decentralization: OpenMined is created on a decentralized blockchain platform.
- Data Ownership and Privacy: OpenMined’s protocols use homomorphic encryption to provide data ownership and privacy.
- Artificial Intelligence: With the help of machine learning and deep learning, intelligence is provided.
OpenMined is an open source community. It is actively developing a number of blockchain-based technologies. They are created with the focus on decentralizing and encrypting artificial intelligence. The goal is to create software that promotes the decentralization of data ownership. Also, creating machine learning and deep learning libraries that will be accessible to the open source community.