beyond the “i” in ai€¦ · © 2019 the mathworks, inc. 1 beyond the “i” in ai richard...
TRANSCRIPT
© 2019 The MathWorks, Inc. 1
Beyond the “I” in AI
Richard RovnerVP Marketing, MathWorks
© 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
© 2019 The MathWorks, Inc. 4
AI has tremendous potential to increase productivity
3x
AI 4x
2x
=
McKinsey Global Institute, September 2018
© 2019 The MathWorks, Inc. 5
But, AI is struggling
Why Most AI Projects FailOct, 2017
Most AI Projects Fail. Here’s How to Make Yours Successful.July, 2018
3 Common Reasons Artificial Intelligence Projects FailMay, 2018
© 2019 The MathWorks, Inc. 6
There are many ways Artificial Intelligence can fail
No data scientists
Not enough data
Too much data
Problem is a poor fit for AI
Low ROI
Beyond the skill of the team
Incomplete tools
Can’t interact with other systems
© 2019 The MathWorks, Inc. 7
AI is more than just the intelligence of the algorithm
Intelligence
Interaction
Insights
Implementation
Apply domain expertise
Operate within their environment
Span the entire design workflow
© 2019 The MathWorks, Inc. 8
Intelligence
Interaction
Insights
Implementation
Apply domain expertise
Operate within their environment
Span the entire design workflow
© 2019 The MathWorks, Inc. 9
Bring human insights into AI
Select data
Make tradeoffs
Evaluate results
AI
© 2019 The MathWorks, Inc. 10
Bring human insights into AI
• We are the domain experts
• Shortage of data scientists
• We need the right tools
© 2019 The MathWorks, Inc. 11
Improving New Zealand Dairy Processing• University of Auckland
• Auckland University of Technology
© 2019 The MathWorks, Inc. 12
Continuous Plant Process
Powdered Milk
Goal: Detect a bad product earlier
Days later
Raw Milk
© 2019 The MathWorks, Inc. 13
Data
Powdered Milk
Raw Milk
Goal: Detect a bad product earlier
AI modelPlant Process Predict Results
Near real-time
© 2019 The MathWorks, Inc. 14
Data
Powdered Milk
Raw Milk
They had lots of data
Plant Process
• Millions of data points
• 6 years
• 3 plants
© 2019 The MathWorks, Inc. 15
Data
Powdered Milk
Raw Milk
But…
AI modelPlant Process
© 2019 The MathWorks, Inc. 16
They had three key insights
1. Need to build a separate model for each plant
Plants behaved differentlyfrom each another
© 2019 The MathWorks, Inc. 17
They had three key insights
1. Need to build a separate model for each plant
2. Plant’s operating state changes each year
Each year was like a completely different plant
© 2019 The MathWorks, Inc. 18
True
Cla
ss
Predicted Class
Bulk density prediction results were inaccurate
• Many false positives
• Unused classes
© 2019 The MathWorks, Inc. 19
They had three key insights
1. Need to build a separate model for each plant
2. Plant’s operating state changes each year
3. Training data was biased
© 2019 The MathWorks, Inc. 20
True
Cla
ss
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. 21https://imgur.com/gallery/8B5Tx
「MATLABによる非常に生産的なデータの可視化と解析の素早さを目の当たりにした時の、業界関係者の驚きを隠せない表情を見るのは
非常に誇らしく思います。この結果が彼らの仮説を検証し、プロセス改
善の新しいアイディアを生み出す事に大いに貢献してくれました。」
- David Wilson, Industrial Information and Control Centre
© 2019 The MathWorks, Inc. 22
To succeed with AI …
Combine AI model buildingwith scientific and engineering insights.
Use tools that span boththe science and engineering and the data science.
© 2019 The MathWorks, Inc. 23
Intelligence
Interaction
Insights
Implementation
Apply domain expertise
Operate within their environment
Span the entire design workflow
© 2019 The MathWorks, Inc. 24
Intelligence
Interaction
Insights
Implementation
Apply domain expertise
Operate within their environment
Span the entire design workflow
© 2019 The MathWorks, Inc. 25
Implementation is about designing the solution
TestingData analysis
Reporting
Concept DevelopmentPrototypingDeployment
Requirements buildingModeling & simulationVerification & validation
Research Manufacturing Autonomous System
© 2019 The MathWorks, Inc. 26
“Deliver on the promise of self-driving cars today.”
© 2019 The MathWorks, Inc. 27
Voyage’s goal was to quickly get to market
1. Target retirement communities
© 2019 The MathWorks, Inc. 28
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. 29
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. 30
Voyage completed their AI system first
AIPerception
System
© 2019 The MathWorks, Inc. 31
Connect the AI to the rest of the system
AIPerception
System
Vehicle ControlSystem
Vehicle Dynamics
Environment
© 2019 The MathWorks, Inc. 32
Started with Simulink example
© 2019 The MathWorks, Inc. 33
Injected simulated vehicles to interact with while driving
© 2019 The MathWorks, Inc. 34
Deployed controller as ROS node and generated code
Robotics System ToolboxEmbedded Coder
Robot Operating System
© 2019 The MathWorks, Inc. 35
Train your AI faster with tight simulation loops
Field Data Algorithms
Usage
Better
SyntheticData
Simulated Usage
© 2019 The MathWorks, Inc. 36
Use simulation for data synthesis
Simulation-based workflow
Simulate Auto-label
Traditional deep learning workflow
Record Label
AI model
© 2019 The MathWorks, Inc. 37
“Simulink + ROS allowed us to deploy a Level 3 autonomous vehicle in less than 3 months.”− Alan Mond, Voyage
© 2019 The MathWorks, Inc. 38
To succeed with AI …
Use tool chains that span the entire design workflow.
© 2019 The MathWorks, Inc. 39
Intelligence
Interaction
Insights
Implementation
Apply domain expertise
Operate within their environment
Span the entire design workflow
© 2019 The MathWorks, Inc. 40
Intelligence
Interaction
Insights
Implementation
Apply domain expertise
Operate within their environment
Span the entire design workflow
© 2019 The MathWorks, Inc. 41
Interaction within complex environments
© 2019 The MathWorks, Inc. 42
What was the larger system the vehicle had to operate in?
© 2019 The MathWorks, Inc. 43
“Proactive patient care”
© 2019 The MathWorks, Inc. 44
Statistics and Machine Learning ToolboxSignal Processing ToolboxMATLAB CoderEmbedded Coder
© 2019 The MathWorks, Inc. 45
EarlySense’s AI can predict critical events before they happen
ContinuousMonitoring
EarlyDetection
EarlyIntervention
BetterOutcomes
AI AI
© 2019 The MathWorks, Inc. 46
Dashboards at nurses’ stations and on hallway monitors
© 2019 The MathWorks, Inc. 47
Alerts on hand-held devices carried by staff
© 2019 The MathWorks, Inc. 48
Address problems before they become emergencies
© 2019 The MathWorks, Inc. 49
To succeed with AI …
Design how our systems will integrate and interact within their environment.
© 2019 The MathWorks, Inc. 50
AI is a transformative technology But AI projects can and do fail
Success requires more than just intelligence
© 2019 The MathWorks, Inc. 51
© 2019 The MathWorks, Inc. 52
Intelligence
Interaction
Insights
Implementation
Apply domain expertise
Operate within their environment
Span the entire design workflow
© 2019 The MathWorks, Inc. 53
Intelligence
Interaction
Insights
Implementation
Apply domain expertise
Operate within their environment
Span the entire design workflow
How will you apply AI to your projects?
© 2019 The MathWorks, Inc. 54
Go Beyond the “I” in AI