Tom Soderstrom, Chief Technology and Innovation Officer at NASA’s Jet Propulsion Laboratory, has demonstrated how internet-of-things (IoT) technology and cloud computing can form the backbone for monumental innovation. This combination has enabled private and public space exploration enterprises to dare greatly and, together, discover more of the solar system than ever before. Cloud computing, with its unlimited storage and compute resources, blends IoT, machine learning, intelligent assistance, and new interfaces with computers. It has the potential to allow humans to explore and colonize other areas of the solar system by enabling collaboration across millions of miles, and social networking on a planetary scale.
2. Can we answer the BIG questions?
How can we innovate at the edge?
Tom Soderstrom, IT Chief Technology and Innovation Officer, JPL
Tom.Soderstrom@jpl.nasa.gov
June 2018
3. Can we find Earth
2.0?
Why?To help us answer the BIG questions
Are we alone?
How do we protect Mother
Earth?
How do we divert an
asteroid?
Is/was there life
on Mars?
How did the Universe form
and where is it going?
T
4. Seven new missions launched in May 2018 (a JPL record)
Cold Atom Lab and Raincube
TEMPEST-D CubeSat
InSight lander
CubeRRT CubeSat GRACE-FO
Two Mars CubeSat Ones
5. Enjoy the benefits of surfing (user experience)
and
leverage the power and future of the wave (back end)
and
spend time doing it (priorities and focus)
How can we infuse emerging technologies into the enterprise?
8. What is the next Technology Tsunami?
8
Built-in
intelligence everywhere,
all the time
9. Our Vision of How We Will Work
Leverages IOT, Programming, Smart Data, Cloud, and Artificial Intelligence, which can evolve at different cadences
AI / Machine Learning
Intelligent
Assistance
(IA)
Deep LearningData Lakes
APIs
(Registry)
Sensors and Devices
(IoT)
Gesture
Touch
See
Type
Sense
Think
Hear
Speak
Click
10. Ubiquitous Computing Software Defined Everything
Accelerated Computing
Cyber Security challengesNew Habits
Applied AI Built in
intelligence
everywhere
Participate at: techwaves@jpl.nasa.gov or https://goto.jpl.nasa.gov/techwaves
These waves will impact everyone. These waves will impact primarily developers.
Surfing the next major technology waves - Summary
11. Ubiquitous Computing Software Defined Everything
Accelerated Computing
Cyber Security challengesNew Habits
Applied AI Built in
intelligence
everywhere
Work from anywhere, always connected,
gaming, sharing, open source, reduced
footprint, cord-cutting
At scale, authentication, encryption
by default, role-based training,
BlockChain
Serverless, edge computing, HPC,
GPUs, Neuromorphic, Quantum
Programming everything, APIs, Software Defined
Networks, containers, DevOps, Open Source, self-
healing, everything distributed
Mobile, smart devices, AR, IoT, NUI
Deep Learning, Machine
Learning, chatbots, NLP,
automation, data-driven, APIs,
analytics, combinations
Participate at: techwaves@jpl.nasa.gov or https://goto.jpl.nasa.gov/techwaves
These waves will impact everyone These waves will impact primarily developers
Surfing the next major technology waves - Summary
12. Ubiquitous Computing Software Defined Everything
Accelerated Computing
Cyber Security challengesNew Habits
Applied AI Built in
intelligence
everywhere
Work from anywhere, always connected,
gaming, sharing, open source, reduced
footprint, cord-cutting
Participate at: techwaves@jpl.nasa.gov or https://goto.jpl.nasa.gov/techwaves
These waves will impact everyone These waves will impact primarily developers
Surfing the next major technology waves - New Habits
13. 7/11/2018 13
New Habits – JPL Perspective
• Work from anywhere
• Always connected to each other, our data, our devices
• Sensors everywhere give us constant insight
• 3D print our future
• Experimentation makes us faster
• Open Source mentality helps us reuse and share
• Reconfigure our offices at ease makes us nimble
• Crowdsourcing helps us develop faster (e.g. JPL Pitch Day, Parking App Challenge, Digital
Assistants, ChatBots, Open Source Rover)
21. Ubiquitous Computing Software Defined Everything
Accelerated Computing
Cyber Security challengesNew Habits
Applied AI Built in
intelligence
everywhere
At scale, authentication, encryption by default,
role-based training, BlockChain
Participate at: techwaves@jpl.nasa.gov or https://goto.jpl.nasa.gov/techwaves
These waves will impact everyone These waves will impact primarily developers
Surfing the next major technology waves – Cyber Challenges
22. JPL/Caltech PROPRIETARY—Not for Public Release or Redistribution
Handling cybersecurity challenges at scale requires understanding threat trends
JPL Perspective
Industry Threat Trends CY17Q2 CY17Q3
Email Phishing
Vishing
Distributed Denial of Service attacks
Ransomware email escalations
End of Life software exploitation
Attacks on externally accessible websites
Poorly managed/configured systems
Weak security encryption deployments
Zero-Day / APT / Nation-State attacks
Dark Web Intel (TOR, Onion, Play Pen, etc.)
Insider threat
Make it easy to be secure
• E.g. derived credentials test lab
• E.g. AI in the background (DDOS)
• E.g. BlockChain
27. Data Wall - Combining multiple senses for your use case
27
POC: M. Cox
28. Ubiquitous Computing Software Defined Everything
Accelerated Computing
Cyber Security challengesNew Habits
Applied AI Built in
intelligence
everywhere Serverless, edge computing, HPC,
GPUs, Neuromorphic, Quantum
Surfing the next major technology waves – Accelerated Computing
Participate at: techwaves@jpl.nasa.gov or https://goto.jpl.nasa.gov/techwaves
These waves will impact everyone These waves will impact primarily developers
Surfing the next major technology waves – Accelerated Computing
29. JPL
7/11/2018 29
Accelerated computing
Graphical Processing Units
explosion benefit
Artificial Intelligence, Virtual Reality,
and BlockChain
Serverless computing is the next
step for cloud computing
Edge computing scales by distributing intelligence
to the perimeters and is fast and resilient
Use others’ High Performance Computing
30. 7/11/2018 30
Accelerating the computing – JPL Perspective
• Cloud computing is the new normal
• Serverless computing experiment saw higher scalability and 100x less cost
• Edge computing prototype showed faster system performance and higher
resiliency
• High Performance Computing now uses multiple sites (JPL, Ames, Texas, cloud)
• Established Graphical Processing Units (GPU) test lab.
• Holding training classes in GPU programming
• Experimenting with Quantum Computing
• Getting ready for hugh increases coming in WiFi 802.11ax, 5G, and Bluetooth 5.
OSR uses Edge Computing
Alexa NASA Mars uses
Serverless Computing
NISAR is testing GPU computingJPL GPU test lab available now
AI/AR processing in near realtime
Is being used already
31. N A S A M A R S
N A S A ’ S F I R S T
A L E X A A P P
“Alexa, enable NASA Mars"
“How cold is Mars?”
”Can people live on Mars?”
…
• Separate IoT Network
• Serverless - safer and cheaper
• Natural user interfaces
• Using multiple senses
• Handles huge scale
32. Ubiquitous Computing Software Defined Everything
Accelerated Computing
Cyber Security challengesNew Habits
Applied AI Built in
intelligence
everywhere
Programming everything, APIs, Software Defined
Networks, containers, DevOps, Open Source, self-
healing, everything distributed
Surfing the next major technology waves – Software Defined
Participate at: techwaves@jpl.nasa.gov or https://goto.jpl.nasa.gov/techwaves
These waves will impact everyone These waves will impact primarily developers
Surfing the next major technology waves – Software Defined
33. 7/11/2018
Software Defined Everything
• Programming is a core competency
• Open Source is the new normal
• Automation is built-in
• The world is run by DevOps
• Application Programming Interfaces
(APIs) become critical
• Distributed enterprises rise using
Software Defined Networks
• Self-healing networks (e.g. NetFlix)
Credit: Techcrunch.com
34. 7/11/2018
• JPL code sharing mentality has grown 100%
• Software containers save time/money/risk
• Investigating in API management
• Rapid experimentation in joint software
and hardware development
• Focus on software training, experimentation
• Automation is expected and built-in
• Software defined… spacecraft, radios, …
• Building software-defined, auto-adjustable, self-
healing networks
JPL Perspective on Software Defined Everything
35. JPL/Caltech PROPRIETARY—Not for Public Release or Redistribution
Our 3 year transformation will improve network
resiliency, security, and performance
Future Network Infrastructure Capabilities
• High-capacity bandwidth (40GigE,
100GigE)
• Additional resiliency through Fabric
Architecture giving high speed and
network agility
• Software Defined Data Center and
Networking (SDDC, SDN) to automate
and orchestrate LAN & WAN resources
• Improved network visibility for
enhanced cyber analysis
• Increased adoption of off-prem cloud
and hosting facilities (e.g., DR)
Fabric Network Resiliency
100GigE Transformation
SDN – Intelligent Networks to
meet demand
2018
2019
2020
36. JPL/Caltech PROPRIETARY—Not for Public Release or Redistribution
Ubiquitous Computing Software Defined Everything
Accelerated Computing
Cyber Security challengesNew Habits
Applied AI Built in
intelligence
everywhere
Mobile, smart devices, AR, IoT, NUI
Participate at: techwaves@jpl.nasa.gov or https://goto.jpl.nasa.gov/techwaves
These waves will impact everyone These waves will impact primarily developers
Surfing the next major technology waves – Ubiquitous Computing
37. There are new IoT Use Cases every day…
White Cane 2.0
(Credit: MIT and Economist)
Connected Vehicles
Credit: The Daily Conversation/YouTube
Smart Cities and Smart Homes
Credit: Lux Reviews
Fleet Management
Credit: AT&T Enterprise/YouTube
Near-term areas:
Wearables
Voice
Healthcare
Transportation
Manufacturing
Security monitoring
Energy
…
38. 7/11/2018
Ubiquitous Computing – JPL Perspective
• JPL IOT Network allows for rapid
and safe experimentation
• JPL mobile apps and app store
• New mobile apps and capabilities
Bring Curiosity home Portal to anywhere
Science and engineering with AR glasses
(fewer experts with high fidelity)
AR on mobile phone
(democratizes AR)
39. Early promising IoT experiments
Data Wall - Combining
multiple senses
Controlling lights, A/V
equipment with voice
Understand and mitigate
hacking attempts
Alexa as Virtual
helpdesk and phone
As interface to
ChatBots
“Ask me anything”
E.g. JPL Info,
Roombot, AWS
Control robots
via voice
Alexa for public
outreach
NASA Mars
Acquisition
Intelligent Assistant
40. N A S A M A R S
N A S A ’ S F I R S T
A L E X A A P P
“Alexa, enable NASA Mars"
“How cold is Mars?”
”Can people live on Mars?”
…
43. 1. Answering questions… 2. Having a dialog
Rapid Iteration of Acquisition’s Digital Assistant
Iterated into …
POC: M. Cox
44. Ubiquitous Computing Software Defined Everything
Accelerated Computing
Cyber Security challengesNew Habits
Applied AI Built in
intelligence
everywhere
Deep Learning, Machine
Learning, chatbots, NLP,
automation, data-driven, APIs,
analytics, combinations
Participate at: techwaves@jpl.nasa.gov or https://goto.jpl.nasa.gov/techwaves
These waves will impact everyone These waves will impact primarily developers
Surfing the next major technology waves – Applied AI
46. Industries AI will transform:
- Security
- Health and medicine
- Manufacturing
- Education
- Business Intelligence
- Retail and eCommerce
- Marketing, Advertising
- Auto & transportation
- Customer service
- Agriculture
- Legal, Finance
- Government
- ...
• 140 startups acquired since 2011, 30 of
which were in 2016. Valued at more than
$2.5B. Most in Healthcare. (CB Insight)
• 20% lift in revenues for companies using
AI by 2020 (Infosys)
• 300% YoY AI investment rise in 2017 (Infosys)
• 40 ZB of data in 2020 (from 4ZB in 2014)
AI by the numbers:
Why AI? Why me? Why now?
• 76% of sales growth for large companies with
Machine Learning (MIT Sloan)
The right conditions finally exists:
successes, compute power, data, open source, skills
47. Decide which AI you want to start with - and experiment
(1) Machine Learning – E.g., Deep Minds, Self driving cars
(2) Natural Language Processing (NLP) – ChatBots, Online help
(3) Computer Vision – Medical diagnosis
(4) Smart Robot - Roomba, Hube
(5) Intelligent Digital Assistant - Alexa, Siri, Google, Cortana
(6) Speech to Text Translation – Google Translate, Azure Media Server
(7) Context Aware Computing – IBM Watson, Google Calendar, Waze
(8) Gesture Control – MS Kinect, Intel Realsense
(9) Speech Recognition – Helpdesk, online banking
(10) Automatic Content Recognition – Shazaam, YouTube
(11) Recommendation Engines – Amazon, Netflix, Chatbots
Categories suggested by Access AIThink and use “IA” to learn and gain adoption
48. Decide which AI to start with and experiment
(1) Machine Learning – E.g., Deep Minds, Self driving cars, NASA Curiosity
(2) Natural Language Processing (NLP) – ChatBots, Online help, NASA Mars
(3) Computer Vision – Medical diagnosis, NASA Curiosity
(4) Smart Robot - Roomba, Hube, JPL RovE, Open Source Rover, Div. Rob. Swarm
(5) Intelligent Digital Assistant - Siri, Google, Cortana, Lex, RoomBot, Search
(6) Speech to Text Translation – Google Translate, Azure Media Server, JPLTube
(7) Context Aware Computing – IBM Watson, Google Calendar, Waze, Acquisition
(8) Gesture Control – MS Kinect, Intel Realsense, JPL OnSight
(9) Speech Recognition – Helpdesk, online banking, JPL Alexa skills
(10) Automatic Content Recognition – Shazaam, YouTube, JPLTube, JPL Search
(11) Recommendation Engines – Amazon, Netflix, Chatbots, Research Assistant
Categories suggested by Access AI
(JPL examples in red)
Think and use “IA” to learn and gain adoption
49. Deep Learning (subset of Machine Learning)
What it is: A technique to implement
Machine Learning that uses neural networks to emulate the
human brain (neurons) and the way humans learn. Needs a
lot of training data. Most active area of research with many
successes including reinforcement learning.
Opportunity: Chose/run open source in the cloud. E.g.
AWS (MxNet, Rekognition, GPUs, Comprehend,
SageMaker) Google (Tensorflow, Vision, TPUs,
Dialogflow, AutoML)
Microsoft (Cognitive Toolkit)
Other (Torch, Caffe, Keras, DeepLearning4J, Theano, … )
Credit:
Wikipedia
50. 7/11/2018 50
ML and Stocks
Applied AI opportunities: New algorithms, massive data sets, and Intelligent Assistants
1 2 3
53. Other Intelligent Digital Assistants aka “Chatbots”
• Acquisition IDA
• RoomBot
• HRBot
• CloudBot
• CyberBot
• Conference Room IDAs
• DataWall
• Room AV control bot
• Conferences IDA
• Proposals IDA
• …
Lessons Learned
• Keep it focused
• Use Natural Interfaces
• Deploy quickly and iterate
• Access to the data matters most
• Access to displays matters too
• X Functional Teams work best
• Keep it light, fun, and fast
54. Is face recognition ready and helpful in our enterprises?
7/11/2018 54
POC: R. Ma
56. jpl.nasa.gov
Can we use ML to detect anomalies in Soil Moisture Active Passive (SMAP) telemetry?
• Good test candidate, good relationships
• ~4,000 telemetry channels
• Power, CPU, RAM, Thermal, Radiation, counters,
switches
• 4B values
• Challenges
• Semi-supervised
• Complexity, diversity
• Scale
Copyright 2017 California Institute of Technology. U.S. Government sponsorship
acknowledged
POC: K. Hundman
57. Use Deep Learning to detect spacecraft anomalies automatically
Most Recent Data
(current day)
Model Predictions
(for current day)
Prediction
Errors =
ANOMALIES
Minus
Equals
58. jpl.nasa.gov
AI for SMAP Telemetry Anomaly Detection
Eclipse day – Aug 21st, 2017
Copyright 2017 California Institute of Technology. U.S. Government sponsorship
acknowledged
59. JPL AI and IOT related experiments
Data Wall - Combining
multiple senses
Cyber Security
Chatbots through
typing, speaking,
texting
Acquisition
Intelligent Assistant
Use of IoT and
Serverless architecture
Detecting Spacecraft
Anomalies
People CounterFace Recognition MEMEX
60. Additional AI, IoT and Natural User Interfaces experiments
Detecting interesting
rocks on Mars
Understand and mitigate
hacking attempts
Alexa as Virtual
helpdesk and phone
Control robots
through voice
and AI
Methane detection
Derived Credentials
(front end UE)
DDOS Remediation
(back end UE)
Relevance tweets
61. Help evolve JPL’s exploration of these technology trends Summary
Ubiquitous Computing Software Defined Everything
Accelerated Computing
Cyber Security challengesNew Habits
Applied AI Built in
intelligence
everywhere
Work from anywhere, always connected,
gaming, sharing, open source, reduced
footprint, cord-cutting
At scale, authentication, encryption by default,
role-based training, BlockChain
Serverless, edge computing, HPC,
GPUs, Neuromorphic, Quantum
Programming everything, APIs, Software Defined
Networks, containers, DevOps, Open Source, self-
healing, everything distributed
Mobile, smart devices, AR, IoT, NUI
Deep Learning, Machine
Learning, chatbots, NLP,
automation, data-driven, APIs,
analytics, combinations
Participate at: techwaves@jpl.nasa.gov or https://goto.jpl.nasa.gov/techwaves
62. An innovation approach – The Process
1. Question Farm - with end users to find low-hanging use case.
2. Experiment - with users and developers using 1 or 2-pizza teams.
3. Take the easy path - Make it easy to understand, build, and use.
4. Measure - what works… abandon what doesn’t.
5. Focus on the data - it is your currency!
6. Double down - on what had an impact… and iterate.
7. Find or develop the skills – Look at prescriptive analytics group for skills
8. Partner with Cyber Security – Speeds up progress by orders of magnitude
9. Move forward – we can take small steps more quickly and easily.
63. An innovation approach – The Technology
1. Open Source – available, inexpensive, growing
2. Cloud computing – for maximum leverage and speed
3. Crowdsourcing - partner internally and externally (e.g. Kaggle)
4. Internet of Things – for Interacting Naturally and collecting data
5. Analytics – extend the analytics efforts into IA
6. APIs – how you access the data
7. AI frameworks and libraries – makes it easy to get started
8. IA – evolve to AI when users trust the IA
9. Combinations – especially of IOT + APIs + AI, all in the cloud
64. Always remember our key priorities
Photo credit Wikimedia.org
Photo credit Wikipedia.org
”With great power
comes great responsibility”
– Voltaire
”If we stop moving, we die”
– Great White Shark
66. Seven new mission launched in May 2018 (a JPL record)
Cold Atom Lab and Raincube
TEMPEST-D CubeSat
InSight lander
CubeRRT CubeSat GRACE-FO
Two Mars CubeSat Ones
67. W E W I L L
C O L L E C T
1 0 0 X M O R E
D A T A O C O - 2
2009
S M A P
2015
S W O T &
N I S A R
2021
1 0 0
T B / D A Y
Understand the Earth’s water in detail with SWOT and NISAR