2. File Servers
“Push” Pub-Sub
PIM’S
Databases
Web sites
Content Portals Search Engines
Enterprise Portals
Evolution of the Web
P2P File-Sharding
Groupware
Marketplaces Auctions Wikis
Weblogs
RSS
E-mail
Conferencing USENET
Social Networks
Community Portals
IM
Artificial Intelligence
Personal Assistants
Intelligent Agents
Semantic Webs
Knowledge Bases
Knowledge Management
Ontologies
Taxonomies
Decentralized Ledger
Privacy-Preserving Computation
Knowledge Networks
Degree of Social Connectivity
Degree
of
Information
Connectivity
Source: Novas Pivack
Enterprise Minds
Group Minds
Lifelogs
“The Relationship Web”
Decentralized Communities
WEB 3.0(Semantic Web)
Connects Knowledge
WEB 2.0(Social Web)
Connects People
WEB1.0(The Web)
Connects Informatiion
WEB X.0(Meta Web)
Connects Intelligence
Semantic Weblogs
Smart Marketplaces
Multi Agent System
PlatON
The Global Brain
3. Market Capilazation of the Web
1997 2020 2037
Share
of
Global
Market
Cap
0.0%
7.5%
15.0%
22.5%
30.0%
Web 1.0
Green Shoots of E-commerce
Desktop Browser Access
Dedecated Infrastructure
$13 Trillion
$2 Trillion
~$0
$20 Trillion
$30 Trillion
Web 2.0
Social Networks
Mobile always on
Cloud-driver Computation
Web 3.0
AI-driver Networks
Decentralized Data Architecture
Privacy-preserving Computation
17% CAGP
Information Technology Internet Artificial Intelligence Data Souce: ARK INVEST
4. 180
160
140
120
100
80
60
40
20
0
Size And Security of the Global Data
2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025
46%
25%
29%
51%
24%
25%
Requires security unprotected Requires security protected Does not require security Data Source: IDC’s Data Age 2025 Study
Zettabytes
33%
32%
35%
13%
42%
35%
5. Problems and Solution
Challenges with the current web
l Centralization: The capturing of behavioral data is locked into silos that are
dominated by large players.
l Privacy and security: With the increasing amount of data being captured large
data centers act as honeypots for organized crime.
l Scalability: With larger data sets from billions of connected devices, today’s
client server model is not likely to scale for the next generation web.
Problems of artificial intelligence
l Centralization and De-democratisation:A handful of tech giants have
monopoly control over data, algorithms, and top AI talents, AI has not been
democratized.
l Data abuse and privacy leakage:The centralization of AI opens the door
to all sorts of abuse. Due to privacy leakage and government regulation, there is
not much, and higher quality data for AI model training.
l Expensive training cost: Total AI training costs continue to climb, state-of-
the-art AI training model costs are likely to increase 100-fold, from roughly $1
million today to more than $100 million by 2025.
AI
Blockchain Cryptography
6. Let us
compute
https://www.platon.network
What is PlatON
Build an infrastructure for autonomous AI agents and their
collaboration, facilitate the emergence of advanced AI, and explore
the path to AGI.
Make the power of AI available to anyone and make AI technology
work better for the public.
AI
A Decentralized Privacy-Preserving AI Network
Web 3.0 to WebX.0
An evolutionary path to AGI
Democratizing Artificial Intelligence
7. The Privacy-Preserving AI Network
A decentralized and collaborative AI network on
the top layer in which AI agents collaborate at scale,
accomplish complex AI services through group minds.
A privacy-preserving computation overlay network on
top of the underlying blockchain network, which supports
AI-level complex computation, creates a decentralized AI
marketplace by incentivizing more data, algorithms, and
computing power to join through cryptoeconomics.
The underlying trust layer that implements consensus
protocols and smart contracts.
PlatON's Layer1 base protocol is deeply customized for
privacy-preserving AI.
8. The Stack Of Privacy-Preserving AI Network
PlatON will
l Focus on technologies related to
privacy-preserving AI training,
serving and interoperability.
l Collaborate extensively with the
community to create an ecosystem
of AI computing and services.
9. Competitive Landscape
Layer Technical Features Application
Layer1 Layer2 Layer3 TPS TTF Privacy-Preserving
Computation
Smart
Contract
AI Model
PlatON ✔ ✔ ✔ 10k 3s Cryptography EVM
WASM
✔
Ethereum ✔ ✔ 10 6m ✗ EVM
eWSM
Cosmos ✔ 1k 6s ✗ ✗
EOS ✔ 4k 163s ✗ WASM
Solana ✔ 50k 1.5s ✗ Rust
Oasis ✔ 1k 6s TEE EVM
eWSM
Enigma ✔ TEE
Arpa ✔ MPC
Phala ✔ TEE
10. Competitive Advantages
Decentralization
Any user and node can connect to the network permissionless.
Any data, algorithms and computing power can be securely
shared, connected and traded. Anyone can develop and use
artificial intelligence applications.
Low training costs
With blockchain and privacy-preserving computing
technologies, anyone can share data and algorithms in a
secure and frictionless marketplace, truly reducing marginal
costs and drastically reducing training costs.
Privacy-Preserving
Modern cryptography-based privacy-preserving computing
techniques provide a new computing paradigm that makes data
and models available but not visible, allowing privacy to be fully
protected and data rights to be safeguarded.
Low development threshold
Visualize AI model development and debugging, automated machine
learning (AutoML), MLOps simplifies the whole process of managing
AI models from model development, training to deployment, reducing
the development threshold of AI models and improving development
efficiency.
High-performance
High-performance asynchronous BFT consensus is achieved
through optimization methods such as pipeline verification,
parallel verification, and aggregated signatures, and its safety,
liveness, and responsiveness are proven using formal
verification methods.
Regulatable and auditable
All data, variables and processes used in the AI training decision
making process have tamper-evident records that can be tracked and
audited. The use of privacy-preserving technologies allows the use of
data to satisfy regulatory regulations such as the right to be forgotten,
the right to portability, conditional authorization, and minimal
collection.
1
5
6
4
2
3
12. A Few Lines to Privacy
l Compatible with Tensorflow, switch to
privacy mode in 3 steps:
I. Import Rosetta package
II. Set MPC algorithm
III. Get private input data
l No knowledge of cryptography required for developers
l Adapting common statistical analysis algorithms, machine
learning algorithms and deep learning algorithms to
privacy-preserving algorithms
Privacy-Preserving AI Framework (Rosetta)
Hareware Acceleration
CPU GPU FPGA ASIC
TensorFlow Pytorch Spark Flink
Privacy-Preserving AI Lidlary (RosettaMLLib)
MPC ZKP HE
Federated
Learning
TEE
Mechine Leaning Deep Leaning Statistic
Network
13. Privacy-Preserving Computation on PlatON
l layer1 has built-in privacy algorithms
(including homomorphic encryption and
zero-knowledge proof) that can be
integrated into smart contracts.
l Layer 2 provides secure multi-party
computing protocols for privacy-preserving
training of AI models, and the trained
models can be deployed to Layer 3.
Decetralized Resource Management
Blockchain
Resource Publish Resource Discovery
Computing
Market
Data
Market
Algorithm
Market
Economic
lncentive
Supervision
Audit
Resource Scheduling Resource Location
RELOAD Overlay Network
DID(Decentralized ldentity)
Zero-Knowlage Proof Library
EVM Visual Machine WASM Visual Machine
Data
Provider
Data
Provider
Data Area
Computing
Node
Computation Area
MPC
Protocol
Proxy
Re-Encyption
AI Network
Secret
Share
Both on-chain and off-chain
data privacy are supported:
Computing
Node
Computing
Node
Data Node
Data Node
Data Node
14. AutoML
Privacy-Preserving AI Service Platform
Available out of the box
1. Drag-and-drop model orchestration and visual debugging
2. One-stop machine learning, end-to-end training process
automation
3. MLOps pipeline automation, continuous training and
deployment
4. Data visualization cockpit
5. One-click Docker deployment, support cloud and local
deployment
Security Compliance
1. Complete mathematical and cryptographic proofs
2. Strict authentication and authorization
management
3. Compatible with both international open source
and Chinese commercial secret systems
4. Blockchain depository audit
5. Secure visualization cockpit
Flexible Extensions
1. Plug-in support for extended algorithms, data
sources, authentication and authorization
2. support for security multi-party computation, federal
learning, TEE and other privacy-preserving
computation algorithm
3. Support SQL, Python and other high-level
languages
Feature Enginnering
Model Selection
Hyper-parameter Optimization
Visualization
Modeling
Data
Sources Pre-Processing
Multi-party
Deployment
Joint Training
Joint Prediction
Model Monitor
Data and
Algorithm
Marketplace
Discovery
Trading
Evaluation
Joint
model
Continuous Training and Deployment
Privacy-preserving
Computation Network
15. Application Ecosystem
Marketing/Advertising
User Portrait
Customer Match
Intelligent City
Intelligent
Manufacturing
Intelligent
Transportation
Intelligent
Building
DeFi
Model Oracle
Privacy-Preserving Dex
DeFi Credit
Biologic
Drug Screening
Clinical Trials
Target Discovery
Finance
Anti-fraud
Risk Control
PlatON
DeFi
16. Research Team and Results
World-leading research team
We have established a research funding for privacy-preserving computation and have a large pool of
top cryptographic talent, including professors and PhDs from major universities in China and the U.S.
Research results
Our research team has been conducting and publishing exploratory and in-depth research in the
fields of cryptography, Internet of Things, human intelligence, as well as economics and governance.
l Papers
l LEAF: A Faster Secure Search Algorithm via Localization, Extraction, and Reconstruction, In ACM CCS, 2020
l Compact Zero-Knowledge Proofs for Threshold ECDSA with Trustless Setup, In PKC, 2021
l Mystique: Efficient Conversions for Zero-Knowledge Proofs with Applications to Machine Learning, In USENIX
Security, 2021
l Doubly Efficient Interactive Proofs for General Arithmetic Circuits with Linear Prover Time, In eprint 2020/1247
l zkCNN: Zero Knowledge Proofs for Convolutional Neural Network Predictions and Accuracy, In eprint 2021/673
l Open-source privacy-preserving AI framework Rosetta
Research grants and project collaborations
We regularly fund academic research and conferences on cryptography and privacy-preserving
computation, and we collaborate deeply with various projects and research groups.
l Conferences: Crypto, Eurocrypt, Asiacrypt, CCS, AsiaCCS, IDASH, etc.
l Project in Ethereum: MPC implementation of Proof of Custody in Eth 2.0