© 2019 The MathWorks, Inc. 1
Beyond the “I” in AI
Chris Hayhurst
© 2019 The MathWorks, Inc. 2
Watt Steam Engine
© 2019 The MathWorks, Inc. 3
Artificial intelligence is a transformative technology
based on McKinsey’s latest AI forecast – September 2018
Seoul National University Hospital
© 2019 The MathWorks, Inc. 4
AI has tremendous potential to increase productivity
3x
AI 4x
2x
=
© 2019 The MathWorks, Inc. 5
Yet AI is struggling
Why Most AI Projects Fail
Oct, 2017
Most AI Projects Fail. Here’s
How to Make Yours Successful.
July, 2018
3 Common Reasons Artificial
Intelligence Projects Fail
May, 2018
© 2019 The MathWorks, Inc. 6
TayTweets AI project taken down within 24 hours
March 24, 2016
…
© 2019 The MathWorks, Inc. 7
There are many ways Artificial Intelligence can fail
No data
scientists
Not enough data
Too much data
Problem is a
poor fit for AI
Poor ROI
Beyond the skill
of the team
Problem is
unsolvable
Incomplete
tools
Can’t integrate with
other systems
© 2019 The MathWorks, Inc. 8
AI is more than just the intelligence of the algorithm
Intelligence
Integration
Insights
Implementation
Apply domain
expertise
Connect with
other systems
Span the entire
design workflow
© 2019 The MathWorks, Inc. 9
Intelligence
Integration
Insights
Implementation
Apply domain
expertise
Connect with
other systems
Span the entire
design workflow
© 2019 The MathWorks, Inc. 10
Bring human insights into AI
Select data
Make tradeoffs
Evaluate results
AI
© 2019 The MathWorks, Inc. 11
Improving New Zealand Dairy Processing
• University of Auckland
• Auckland University of Technology
© 2019 The MathWorks, Inc. 12
Continuous
Plant
ProcessPowdered MilkRaw Milk
Days later
© 2019 The MathWorks, Inc. 13
Powdered Milk
Raw Milk
AI modelPlant Variables Predict Results
Near real-time
© 2019 The MathWorks, Inc. 14
Powdered Milk
Raw Milk
They had lots of data
Plant Variables
• Millions of data points
• 3 plants
• 6 years
© 2019 The MathWorks, Inc. 15
Powdered Milk
Raw Milk
But…
AI modelPlant Variables
© 2019 The MathWorks, Inc. 16
Key Insights Made
1. Predictions were wrong
© 2019 The MathWorks, Inc. 17
Key Insights Made
1. Predictions were wrong
2. Need to build a separate
model for each plant
Plants behaved differently
from each another
© 2019 The MathWorks, Inc. 18
Key Insights Made
1. Predictions were wrong
2. Need to build a separate
model for each plant
3. Plant’s operating state
changes each year
Each year was like a
completely different plant
© 2019 The MathWorks, Inc. 19
Tru
e C
lass
Predicted Class
Bulk density prediction results were inaccurate
• Many false positives
• Unused classes
© 2019 The MathWorks, Inc. 20
Key Insights Made
1. Predictions were wrong
2. Need to build a separate
model for each plant
3. Plant’s operating state
changes each year
4. Training data was biased
© 2019 The MathWorks, Inc. 21
Tru
e C
lass
Predicted Class
100%
90%
83%
93%
93%
93%
97%
10%
10%
3%3%
3%
3%
3%
3% 3%
3%
3%
Resampling data resulted in higher predictive accuracy
• Resampled data
• Reduced the number of bins
© 2019 The MathWorks, Inc. 22https://imgur.com/gallery/8B5Tx
“It’s great to sit down with our industry
partners and watch their jaws drop when
they see how productive we are with
MATLAB and how quickly we can
analyze and plot data.”
- David Wilson, Industrial Information
and Control Centre
© 2019 The MathWorks, Inc. 23
To be successful with AI, you must …
Combine AI model building
with scientific and engineering insights
© 2019 The MathWorks, Inc. 24
Intelligence
Integration
Insights
Implementation
Apply domain
expertise
Connect with
other systems
Span the entire
design workflow
© 2019 The MathWorks, Inc. 25
Intelligence
Integration
Insights
Implementation
Apply domain
expertise
Connect with
other systems
Span the entire
design workflow
© 2019 The MathWorks, Inc. 26
Implementation is about designing the solution
Testing
Data analysis
Reporting
Developing concept
Prototyping
Deployment
Requirements building
Modeling and simulation
Verification and validation
© 2019 The MathWorks, Inc. 27
“Deliver on the promise of self-driving cars today.”
© 2019 The MathWorks, Inc. 28
Voyage’s goal was to quickly get to market
1. Target retirement communities
© 2019 The MathWorks, Inc. 29
Voyage’s goal was to quickly get to market
1. Target retirement communities
2. Use off-the-shelf components
wherever possible
© 2019 The MathWorks, Inc. 30
Voyage’s goal was to quickly get to market
1. Target retirement communities
2. Use off-the-shelf components
wherever possible
3. Bring in the right software tools
across the entire workflow
© 2019 The MathWorks, Inc. 31
Voyage completed their AI system first
AI
Perception
System
© 2019 The MathWorks, Inc. 32
But they needed to connect the AI to the rest of the system
AI
Perception
System
Vehicle
Control
System
Vehicle
Dynamics
Environment
© 2019 The MathWorks, Inc. 33
Started with Simulink example that they could build upon
© 2019 The MathWorks, Inc. 34
Started with Simulink example that they could build upon
Find out more:
자율시스템을위한센서융합및추적
센서신호처리및무선기술
서기환
© 2019 The MathWorks, Inc. 35
Deployed controller as ROS node and generated code
Robotics System Toolbox
Embedded Coder
© 2019 The MathWorks, Inc. 36
Train your AI faster with tight simulation loops
Field Data Algorithms
Usage
Better
Synthetic
Data
Simulated
Usage
© 2019 The MathWorks, Inc. 37
One example of leveraging simulation for data synthesis
Simulation workflow
Simulate Auto-label
Traditional workflow
Record Label
Transfer
Learning
AI model
Preliminary
AI model
© 2019 The MathWorks, Inc. 38
“Simulink + ROS allowed us to
deploy a Level 3 autonomous
vehicle in less than 3 months.”
− Alan Mond, Voyage
© 2019 The MathWorks, Inc. 39
To be successful with AI, you must …
Use tool chains that span
the entire design workflow
© 2019 The MathWorks, Inc. 40
Intelligence
Integration
Insights
Implementation
Apply domain
expertise
Connect with
other systems
Span the entire
design workflow
© 2019 The MathWorks, Inc. 41
Intelligence
Integration
Insights
Implementation
Apply domain
expertise
Connect with
other systems
Span the entire
design workflow
© 2019 The MathWorks, Inc. 42
Integration with complex systems
© 2019 The MathWorks, Inc. 43
What was the larger system the vehicle had to operate in?
© 2019 The MathWorks, Inc. 44
“Proactive patient care”
© 2019 The MathWorks, Inc. 45
Statistics and Machine Learning Toolbox
Signal Processing Toolbox
MATLAB Coder
Embedded Coder
© 2019 The MathWorks, Inc. 46
EarlySense’s AI can predict critical events before they happen
Continuous
Monitoring
Early
Detection
Early
Intervention
Better
Outcomes
AI AI
© 2019 The MathWorks, Inc. 47
Integrate into nurses’ stations
and hallway monitors
© 2019 The MathWorks, Inc. 48
Integrate with hand-held
devices carried by staff
© 2019 The MathWorks, Inc. 49
Address problems before they
become emergencies
© 2019 The MathWorks, Inc. 50
To be successful with AI, you must …
Design how our systems will
integrate and interact within their environment
Find out more:
딥러닝과강화학습
인공지능과딥러닝
김종남
© 2019 The MathWorks, Inc. 51
Summary
▪ AI is a transformative technology
▪ But AI projects can and do fail
▪ Success requires more than just intelligence
© 2019 The MathWorks, Inc. 52
© 2019 The MathWorks, Inc. 53
Intelligence
Integration
Insights
Implementation
Apply domain
expertise
Connect with
other systems
Span the entire
design workflow
© 2019 The MathWorks, Inc. 54
How will you apply AI to your projects?
We have the right tools → MATLAB and Simulink
–Discover and apply insights to fully understand your system
–Implement your complete system across the entire workflow
–Design the systems which will integrate into a larger world