www.edgecomputingworld.com
TOPIO NETWORKS FOOTPRINT – Tracking the 4th industrial revolution
2019
Edge Computing (Edgecomputingworld.com)
Community: 120KEdge Computing, AI for Edge, Big Data, Connected Cars, Autonomous Vehicles
2019
Distributed Apps (Edgecomputingworld.com)Community: 60KBlockchain, DApps
2015
Wearables(ReadWrite Labs)Community: 60K
2017
IoT(ReadWrite Labs)Community: 400KIoT, AI for IoT, Smart Home, IoT Connectivity*
Mapping the information & connections necessary to profit from emerging trends
• Identify relevant emerging trends
• Map timing, critical touchpoints and opportunities
• Create industry landscapes
• Develop and nurture impacted communities
• Enable companies and expert voices to contribute content
3 The Golden Age of Location-Enabled AI
Are you prepared to catch the next technology wave?
Open access to our industry research and analysis
Go to market identification from 100s use cases and markets based on market fit, timing and market sizing
Highly focused campaigns based on selected use cases and industries or selected accountsThought leadership opportunities
Our Speakers:
Gavin WhitechurchCo-Founder
Topio Networks & Edge Computing World
- Philippe Cases
- Philippe Cases
- Philippe Cases
- Philippe Cases/Gavin Whitechurch
- AI On Edge Devices:- Consumer- Enterprise
- AI via Connectivity- Telecom Edge- Data Center- Enterprise Core
WHERE COMPUTE ARISES
100
20%
100
25%
100
20%
100%
0%10%
15% 20%
5% Device Edge
70%55%
30% On-Prem Edge
Telco Edge
Cloud Edge
2015 2020
30%
2025
CONFIDENTIAL
Two types of compute:• Inference• Training
Type of activities at the edge:• Device isolated
insights• Data compression
and conversion
Edge AI Feature Requirements
• Edge resident analytics
• Ability to function autonomously
• Lower power consumption
• Latency
• Security, privacy
• Costs
• Scalability across a vast number of edge devices
• Issues associated with Edge:• Lack of device standardization• Infrastructure readiness (device maintenance and updates both hardware and software)• Potential loss of useful insights
Edge Business Benefits
• Faster response times (low latency)
• Reliability
• Reduced network bottlenecks
• Data filtering
• Costs (power, processing, communications)
Predictions about AI on Edge Devices
• Edge Analytics: 2 to 5 years to mainstream adoption
• Tripling number of Edge AI units between 2019 and 2024
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
••
•
•
Devices can go from an autonomous vehicle in the ten of thousand dollars to a proximity sensors at few dollars
Each devices require a different trade off:• Performance• Inference and training location• Size• Cost• Power• Software support • Price
Key KPIs for market entrants
• Clear value proposition (focus on a problem)
• Comprehensive software stack
• Developer community
• Strong support for partner ecosystem
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
••
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
••
•
•
• Trend 1: Going to the very edge
• Trend 2: Streaming Analytics
• Trend 3: Computational storage
• Trend 4: Healthy climate for innovation in semi conductor companies
• Trend 5: Healthy M&A activity (7 companies acquired)
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
••
•
•
Trend 1 - Going to very edge of the edge: IBM Effort
Trend 1 - Going to the very edge: TinyMLconsortium
• Focus on how best to implement machine learning in ultra-low power systems
• Backed by Google, Qualcomm, Arm, Nvidia, Facebook, Microsoft, Samsung
• Emerging start-ups:• Greenwave systems
• Babble Labs
• ETA compute
• GrAI Matter Labs
• Brainchip
Trend 2: Streaming analytics
• Purposes: Exploring data streams coming from the Edge
• Inference at the Edge
• Training in the cloud or in the edge
• Type of data: mostly sensor type data
Trend 3: Computational Storage
Computational resources into storage devices
Advantage:Lower latencyReducing Bottleneck
Telco Edge, Cloud, Hybrid
Raises US$ 40 millionUse of funds: Build the Edge AI net
Raises US$ 30 millionUse of funds: Accelerate Leadership in Human Security Market
Raises US$ 75 millionUse of funds:Innovating and evolving NLP and Artificial Emotional Computing technologies
Technology: Quantization of neural networks
For Xnor, quantization enables:- 32x memory saving - 58x faster operationsWhile being closer to 2.9% to a model that would operate with the full datasets
Xnor was focused on products that could run independently (on a FPGA Chip powered by a single solar cell…)
Technology: New architecture for Data Center AI Chips focused on deep learning
Habana optimizes on power and costs- 4x more efficient than traditional GPUs such as NVidia
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
••
•
•
Agent Target Systems
High-End ECU with AI Models
CA
N,
LIN
, M
OS
T, F
lexR
ay,
Eth
ern
et
Neural network for predictive maintenance
Source: IIC OTA SIG
Forthcoming Webinars
February 20 Data Platforms for the Smart Building Industry
Philippe Cases, CEO & Co-Founder, Topio Networks
Joseph Aamidor: Aamidor Consulting
Al Sargent: Product Marketing Influx
February 27 Blockchain – State of the Industry Q1 2020
Kyle Ellicott, Principal Blockchain and Dapps Analyst, Topio Networks
Philippe Cases, CEO & Co-Founder, Topio Networks
March 5 Privacy at the Edge
Jessica Groopman, Industry analyst and Founding Partner Kaleido Insights
Gavin Whitechurch, Principal Edge Analyst, Topio Networks