Building Distributed and Decentralized AI Systems: Security and … · 2019-04-15 · • Galvanize...
Transcript of Building Distributed and Decentralized AI Systems: Security and … · 2019-04-15 · • Galvanize...
Building Distributed and Decentralized AI Systems: Security and Privacy Challenges
Tim Llewellynn
9th April 2019Singapore
ARTIFICIAL INTELLIGENCE MARKETPLACE 2
h t tp s :/ / w w w .b o nse ye s .c o mCo p yrig h t Bo nse ye s 20 19 . All rig h t s re se rve d .
OVER 100 LEADING EUROPEAN DATA SCIENTISTS, DEVELOPERS, INDUSTRY EXPERTS PLATFORM FOR DISTRIBUTED AI SYSTEMS OF SYSTEMS DEVELOPMENT FOR EDGE COMPUTING
Countries : 9UK, Ireland, Serbia, Spain, Germany, Switzerland, Austria, Sweden, Greece
Partners : 14nViso SA, University of Castilla-La Mancha, Trinity College Dublin, FHNW, University of Edinburgh, Sciprom, ICCS, TechnischenUniversität München, University of Applied Sciences and Arts of Western Switzerland, Syno, ARM, ZF, RT-RK
Expertise : Compiler optimization, machine learning, computer vision, processor IP, embedded systems, application development, automotive manufacturing and components, and embedded software development.
ARTIFICIAL INTELLIGENCE MARKETPLACE 3
h t tp s :/ / w w w .b o nse ye s .c o mCo p yrig h t Bo nse ye s 20 19 . All rig h t s re se rve d .
DATA REDEFINING USER EXPERIENCES -> AI PREDICTIVE POWER INCREASESIMPROVES SIGNIFICANTLY WITH MORE DATA AND COMPUTE POWER DOMINATED BY A CLOUD CENTRIC MODEL
Indispensable to Business to Consumer Products• Asian tech leaders predict artificial intelligence and
cognitive computing will be the most disruptive technology impacting the global business-to-consumer (B2C) marketplace.
• Companies rapidly leveraging AI to enter into new markets and taking leadership positions.
FROM SENSOR TO USER EXPERIENCE
ARTIFICIAL INTELLIGENCE MARKETPLACE 4
h t tp s :/ / w w w .b o nse ye s .c o mCo p yrig h t Bo nse ye s 20 19 . All rig h t s re se rve d .
High Barriers to Entry
• Structural investment required• Systems improving overtime through
increasing generation of new data• Loyalty by continually improving experience
Chicken/Egg Problem for New Comers
• Not enough data to compete/attract users• Struggle to maintain critical mass• Latecomers to AI will face significant
barriers to entry that are almost impossible to surpass
CREATES STRONG DATA NETWORK EFFECTSTENDANCY TO CENTRALIZE DATA
ARTIFICIAL INTELLIGENCE MARKETPLACE 5
h t tp s :/ / w w w .b o nse ye s .c o mCo p yrig h t Bo nse ye s 20 19 . All rig h t s re se rve d .
AI can predict sleep patterns based on encrypted data traffic
ARTIFICIAL INTELLIGENCE MARKETPLACE 6
h t tp s :/ / w w w .b o nse ye s .c o mCo p yrig h t Bo nse ye s 20 19 . All rig h t s re se rve d .
REDUCE MOVEMENT OF DATADECENTRALIZED AND DISTRIBUTED AI
Cloud algorithms run at the data center on custom FPGA / GPUs or dedicated hardware and CPUs.
Edge algorithms runs on appliances between devices and cloud data centers running Intel x86 or GPUs.
Device algorithms run devices such as mobile phones, watches, and IoT devices often battery operated
Sensor algorithms run directly on sensor data at the sensor before communicating data to the network or main processing unit over a form of interconnection.
Centralized and Monolithic
High
LOW
Movem
ent of Data
ARTIFICIAL INTELLIGENCE MARKETPLACE 7
h t tp s :/ / w w w .b o nse ye s .c o mCo p yrig h t Bo nse ye s 20 19 . All rig h t s re se rve d .
• Galvanize EUROPEAN AI community around challenge based, user driven innovation.
• Create a sustainable research-development-industry virtuous cycle.
• Develop a suitable software infrastructure.
• Fill technology gaps.• Develop a Strategic Research and
Innovation Agenda for AI including ELSE (Ethical, Legal, Socio-Economic).
EUROPEAN AI ECOSYSTEMINDUSTRY DRIVEN INNOVATION
DATA
KNOWLEDGE
ALGORITHMS
TOOLS
AUTOMOTIVE
HEALTHCARE
FINANCE
MANUFACTURIN
RESEARCHERS INDUSTRY
MULTIPARTY COLLABORATION
AI SERVICE LAYER FOR EDGE AI
ARTIFICIAL INTELLIGENCE MARKETPLACE 8
h t tp s :/ / w w w .b o nse ye s .c o mCo p yrig h t Bo nse ye s 20 19 . All rig h t s re se rve d .
ENABLING MULTI-PARTY COLLABORATIONARTIFICIAL INTELLIGENCE VALUE CHAIN
The Bonseyes Marketplace provides the reuse of data, meta data, and models among separate legal entities, to significantly reduce design and development time in building systems of artificial intelligence. It support a number key actors in the AI value chain:
• Challenge Providers• Data Providers• Data Scientists• Application Developers• Infrastructure Providers• Solution Integrators
BENCHMARKKNOWLEDGE
Data Scientists App Developers
Infrastructure Providers
SERVICES & TOOLS
DATA MODEL APPLICATION
USE
CAS
ECH
ALLE
NGE
SDa
ta P
rovi
ders
DATA
-DRI
VEN
SOLU
TIO
NS
Inte
grat
ors
AI Marketplace Participants
80% Reduction in Time and Cost
ARTIFICIAL INTELLIGENCE MARKETPLACE 9
h t tp s :/ / w w w .b o nse ye s .c o mCo p yrig h t Bo nse ye s 20 19 . All rig h t s re se rve d .
INDUSTRIES THAT ARE SUPPORTEDWHY GET INVOLVED?
Boost Competitiveness• Top European intelligent product/service research• Reduce CAPEX, experts, time to adoption
Boost Technology Transfer• Mobilize talent, innovation hubs, pilots• Integrated assets, tools, data portals• European industry user-driven challenges
Boost Tech Convergence • Fertilize technology leadership• AI coalitions and interworking• Expand current market
AI SERVICE LAYER FOR EDGE AI
USECASESINFRASTRUCTURE
FinanceManufacturing
TALENT
AutomotiveHealthcare
DATA
USE CASES
Diagnostics Autonomous Driving
Predictive Maintenance Fraud
AI ENABLING TECHNOLOGIES
LEARNING VISION SPEECH NLP DIALOG
CPS ROBOTICS IOT BIG DATA HPC 5G
This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 732204 (Bonseyes). This work is supported by the Swiss State Secretariat for Education‚ Research and Innovation (SERI) under contract number 16.0159. The opinions expressed and arguments employed herein do not necessarily reflect the
official views of these funding bodies.
Questions?
www.bonseyes.com