Slides from Talk @ Intel IoT DevFest IV
With both Facebook and Google's recent shift in direction towards a "Future is Private" world, learn how you too can train and deploy your AI models in a privacy-preserving way, with Decentralized AI and a combination of AI and Blockchain. These techniques will become even more rampant as we move into a world where users will own their own data and companies will start using “ethically sourced data” and move towards a path for Ethical AI for the IoT space.
In this session, you will learn:
- Use cases for Decentralized AI, with combined benefits of AI + Blockchain for IoT applications
- Federated learning & related privacy-preserving AI model training techniques for IoT applications
- How to build Ethical AI solutions for IoT using these techniques
Designing IA for AI - Information Architecture Conference 2024
Private and Ethical AI Roadmap for IoT
1. F U T U R E I S P R I V AT E
R O A D T O W A R D S E T H I C A L A I F O R I O T
J U N E , 2 0 1 9
G E E TA C H A U H A N , C TO
2. RESILIENT ANTI-FRAGILE PLATFORMS
• Data Center as a Computer
• Microservices
• Elastic Scaling
• Self healing
• Chaos Monkey
• Canary Deployments
• A/BTesting
• Continuous Learning
• Continuous Monitoring
• Software Defined Networking
• AI Ops, DevSecOps
• Decentralized AI driven Cloud
• Human in loop AI Crowd-sourcing Platform
• HPC on demand SaaS for Chip Design
• AI/Deep Learning, Recommendation Engines
• Data Platforms for Telecom
• Private Cloud Platform for Telecom
• IoT Analytics
• Mobile Device Management
• Social Media Monitoring Platform
• Web Monitoring Platform
• Speech Analytics Platform
• Developer Platforms w/ Real-time Debugger
17+ Platforms, Countless apps, 1Billion customers…
3. C E N T R A L I Z E D A I I S L I K E T H E
C L O S E D S O U R C E O F T H E 1 9 9 0 S
4. CHALLENGES?
Privacy Problem Can entities train model without disclosing data?
Influence problem
Can 3rd parties contribute towards behavior of AI model in a way that is quantifiably
influential?
Economic Problem Can 3rd parties be correctly incentivized to contribute to knowledge & quality of AI
models?
Transparency Problem Can the activity of behaviour of AI model be transparently available to all parties
without a trusting middleman?
Latency Problem Centralized AI is inappropriate for use-cases where AI needs to interact in real time
with the real world
6. FEDERATED AI
• Subset of devices selected,
each downloads the model
• Train model with local data
• Model updates – gradients –
sent back to server
• Server aggregates
• Cancer treatment centers
training models
7. WHAT IS BLOCKCHAIN?
• An Immutable record of digital events
shared peer to peer between
different parties
• Distributed Ledger, AuditTrail
• Open +Trust + Secure
• Smart Contracts → dApps, DAO
• Non-FungibleTokens, SecurityTokens
• Fully Democratize Internet
Information Age → Internet ofValue
Source: Economist.com
8. W H A T I S H O M O M O R P H I C E N C R Y P T I O N ?
10. DATA EXCHANGE
• Blockchain for Data Provenance
• User Owned Data
• Time expiry for data
• Ethically Sourced Data
– Transparency
– Fairness
– Privacy
11. ETHICAL AI
• Personal Data Rights & Individual
Access Controls
• Well-being metrics
• Awareness of Misuse
• Respect for Privacy
• Governance of AI Autonomy
• Accountability,Transparency
• IEEE Ethical Design, Europe
Trustworthy AI
• California “Data Dividend”
proposal https://cnn.it/2N9KEay
13. PRIVACY CHALLENGE
• FederatedVirtual Assistant
• Fine-grained privacy control
• User controls where data resides
• Stanford Almond project
• Privacy-minded alternative to Alexa, Siri
• $3 Mil NSF grant
14. AI
MARKETPLACE
• Data Competition each
week
• Encrypted data released
• Crowdsource Data Science
Model
• Data Scientists retain IP –
encrypted models
• Participants paid in Bitcoin
based on accuracy of their
guesses, payouts to top 60
• Originality paid extra
15. LATENCY
CHALLENGE
• Slow Inference problem
• Real-time scenarios
• AI needs to interact in
real-time with real-world
• Need compute on / close
to edge devices
Democratizing cloud computing
for cloud resource providers
and application developers
16. LATENCY CHALLENGE FOR IOT
• Slow Inference problem
• Real-time scenarios
• AI needs to interact in
real-time with real-world
• Need compute on / close
to edge devices for your
apps / microservices