Built to Last: AI Models and the Matrix AI Marketplace

Matrix AI Network
Jul 10 · 4 min read

A look at Matrix’s approach to AI models.

The three core pillars of artificial intelligence are data, computing power and AI models. Following the release of the Matrix 2.0 Green Paper, the team is publishing a series of articles delving deeper into key aspects of Matrix 2.0. Part 1of this series introduced the relationship between data, computing power and AI models. Part 2 introduced a few prominent data and computing power issues. This article examines AI model attribution and privacy problems as well as Matrix’s plan to solve them.

If we envision data and computing power as forming the foundation of artificial intelligence, AI models are what enable human/AI interactions. This is because AI models transform cold hard data into something concrete — applications that are directly useable by people. Nevertheless, AI model development faces its own set of challenges.

AI Model Challenges

The development of AI is largely a matter of our ability to scale the application of AI models. In the current environment, two major issues pose a threat to AI model development. First is the ever-increasing difficulty of training AI models. This is largely due to high barriers to entry in the form of high computing power requirements and unfavorable hardware and software conditions. Second is a poorly motivated developer community following years of ineffective intellectual property rights (IPR) protection.

To address these two issues, the Matrix AI Network is developing an AI model marketplace featuring distributed training and AI model attribution made possible by the Matrix 2.0 blockchain-powered multi-dimensional big data platform. Using this marketplace, AI developers can receive proper IPR protection as well as timely feedback to continuously and conveniently optimize their AI models. Strong, proven AI models encourage more users to participate in the training and development of AI models and applications.

Training Challenges: Distributed Training

A decentralized blockchain system will assist developers in creating and training more AI models by redirecting the Matrix AI Network’s surplus computing power. Matrix’s consensus mechanism only requires 32 mining nodes to perform mining activities at any given moment. The surplus computing power of the remaining mining Masternodes is reserved to train AI models and perform other AI-related tasks. This gives data scientists access to computing power in quantities usually reserved for only the largest multinational corporations and academic institutions.

Training Challenges: Attribution

Disappointed and frustrated by poor AI model imitations and ineffective IPR protection, young data scientists often fail or abandon AI model development. To motivate the next generation of developers, the Matrix AI Network uses blockchain to track and attribute AI models to their creators and offer full-and-accurate IPR protection.

Matrix AI Marketplace

Accurate attribution is a necessary and fundamental attribute enabling the establishment of the Matrix AI Marketplace. Each time an AI model is accessed by a user or application, a transaction will be recorded on the Matrix blockchain. The Matrix AI Network blockchain’s immutable ledger ensures that data scientists will be fairly compensated each time their model is accessed. Data scientists are also provided with a complete status record of a model’s usage. This provides valuable feedback to the model creator and allows for convenient future refinements.

Matrix 2.0

Matrix 2.0 combines Matrix 1.0’s opensource blockchain, a data chain, cloud-fog-terminal storage, and a blockchain operating system. It is being designed to support the following features:

  1. Big data transactions with improved data usage efficiency

At the time same, the Matrix 2.0 blockchain platform will also introduce a Secure Multi-Party Computation (SMC) framework that allows multiple parties to participate in computation, while masking inputs and preserving the independence and computational accuracy.

Blockchain operating system

Traditional cloud computing services comprise several computers distributed across several different geographical locations, unified through a network management system. They have largely been centralized affairs. In contrast to these traditional centralized distributed computing systems, the Matrix AI Network operating system is a decentralized solution that features five main functions to ensure convenient and flexible on-demand service.

  1. Orchestration of tasks and distributed compute

The Matrix 2.0 operating system gives users and developers access to a range of services including oracle services, multi-chain management, middleware solutions, smart contract creation, DApp development, and a range of plug-ins.

All parties engaging on Matrix 2.0 receive rewards for their contributions. Data providers are rewarded when their data is used; Computing power providers are rewarded when their computing power is used; Data scientists are rewarded when their AI models are used. Matrix 2.0 is a decentralized AI marketplace.


Matrix AI Network

Written by

Matrix AI Network leverages the latest AI technology to deliver on the promise of blockchain. Website: www.matrix.io Telegram: t.me/matrixainetwork