Three years ago, the privacy computing AI network PlatON released its project white paper. After more than three years being direct active in the market, on September 14, PlatON released the white paper 2.0 on social media, which regarded as an oath of applicating private AI network technology.

PlatON 2.0 established the three phase goals formally,which are building a decentralized private computing network, an AI (artificial intelligence) market, and an AI collaboration network based on the three major technologies of blockchain, privacy computing and AI.

PlatON2.0 Layering and Progression of Privacy AI Network

PlatON White Paper 2.0 discloses the design architecture of its common chain in detail, which adopts a three-layer network architecture, namely the consensus layer, the privacy protection computing network, and the AI network.

Layer1: Consensus network

The consensus network is a P2P (peer-to-peer) blockchain network composed of nodes. The nodes of the blockchain are connected through the P2P protocol. The consensus protocol can be negotiated without trust from external part, which can achieve a certain sense of Serverless.

For example, in the field of intelligent transportation, AI is the “brain” behind driverless vehicles. These autonomous vehicles need to trust each other so that they can cooperate to accomplish their common goals. The AI system has no mechanism to ensure that these autonomous vehicles reach a consensus in a credible way. Therefore, they need a trusted third party or a consensus network to help the subjects in the AI system complete their tasks. However, various data leaks have shown that third parties may expose the public to security and privacy issues.

Certainly, the blockchain network cannot solve all problems. Its disadvantages are low efficiency and data transparency.

  1. On the blockchain, we can execute smart contracts. However, due to the limitations of performance and transaction cost, the blockchain cannot support smart contracts to execute too complex calculation logic and we can only access the data stored on the chain. Besides, the data is also limited. So, the AI model training cannot just be completed on Layer1.
  2. On the blockchain, each participant will acquire a complete copy of the data, and all transaction data is open and transparent. Obviously, the native blockchain technology does not have the ability to protect privacy. A privacy computing protocol based on homomorphic encryption, zero-knowledge proof, TEE and other technologies needs superimposing on the consensus network so that it can protect the privacy of data and computing on the chain.

Layer2: Privacy Computing Network

The privacy computing network above the Consensus network layer is the most important layer of PlatON. This layer includes privacy computing algorithms and data storage protocols to build an open computing power trading market for AI networks and the future, providing essential data processing functions for PlatON and applications built on PlatON. The privacy computing network allows users to contribute computing power and connects users who have data computing needs. Specifically, it performs data calculations for users by the privacy computation and provides incentives for computing power providers by PlatON native tokens.

The data in the privacy computing network is generally stored at local, and the security and privacy of the data are ensured by technologies such as MPC(secure multi-party computation) and federated learning. It makes the data is available and invisible so that subjects prefer to share sensitive data (such as consumption and health information). Over time, the market will accumulate more and more high-quality data. AI experts will be motivated to create and share better AI models.

Layer3: collaborative AI network

Compared with the more robust combination between Layer1 and Layer2, Layer3 is a relatively independent network. The ultimate goal of this layer is to form an autonomous AI network. AI models can be trained by using the datasets and computing resources on the privacy computing network. We can deploy AI models in the AI network and form an AI service market by AI agents serving externally. Through multi-agent systems and other technologies, we can run AI agents to communicate, collaborate, and create more innovative AI services.

For example, the Fetch project is committed to building AEA(Autonomous Economic Agents) and making them cooperate in an organized way. AEA are software entities that can perform actions without external stimuli. They can search for and interact with other autonomous economic agents intelligently.

At present, many projects are trying to combine blockchain, privacy computing, and AI. Some combine privacy computing and blockchain to enhance blockchain privacy protection and computing capabilities, and some combine blockchain and AI to form an AI market. Some use the decentralized blockchain to build computing power and data markets. However, they can only meet part of the demand of privacy AI. Only a few projects, such as Fetch, SingularityNET, and PlatON, are dedicated to building an AI ecosystem.

Among them, the goal of PlatON is to build a privacy protection computing network and an AI collaboration network. The main applications are AI training and services and autonomous agents.

Based on plenty of complex development work and the three-layer network architecture design, PlatON has the advantages of decentralization, security, privacy, efficiency, flexibility, etc. It can support more flexible and extensive application scenarios include finance, medicine, smart city, and IoT.

Financial applications

With private computing technology, operators, Internet platforms, insurance institutions, and other multi-channel data will open more control-type private data to strengthen collaboration with banks. The confidential manner will support the integration and business. Banks will realize the whole process monitor of pre-loan, in-loan, and pro- to improve the timeliness of control. Through the intersection of privacy and the joint privacy query, it is easy to obtain the customer’s comprehensive credit risk without revealing any customer ID and private information. That forms joint insurance, prevention, and control.

Pharmaceutical applications

As an AI infrastructure, PlatON provides a reliable collaboration environment for hospitals, pharmaceutical enterprises, and various scientific research institutions. It connects activities, research fields, operation modes, and data flows from different fields to achieve a large-scale data aggregation effect so that we can make the best of pharmaceutical datasets, such as clinical trials, pharmaceuticals, electrical health records and patient genomics data, etc. PlatON will accelerate the discovery and R & D process of new drugs when a data analysis and mining system of various technical stages such as pharmacogenomics, disease omics, collateral pharmacology, egg structure simulation, etc. is built.

Disclaimer: The views, suggestions, and opinions expressed here are the sole responsibility of the experts. No USA Times Media  journalist was involved in the writing and production of this article.