Skynet? Really? How close are we to self aware, self replicating machines? In this fun session learn some of what computers can do and what they can’t. You think you know. You may be surprised.
The emerging focus on Cognitive computing, general AI, Computer Vision, Internet of Things, etc. signpost the way to new opportunities and new challenges for computers and humans alike. We decided to see how far we could get in building our own version of an all powerful controlling entity.
In this session we’ll cover how we did it, what we learned and answer those important questions like: “Can we build a Skynet yet?”, “Can my computer be my best friend?”, ”Will I ever able to program without a keyboard?”, ”Can a computer read my mind?” and the all important “will drones be able to deliver beer at the right temperature?”
Building Real-Time Data Pipelines: Stream & Batch Processing workshop Slide
North americai iotskynet-v2
1. THE RISE OF THE MACHINE - IS
SKYNET CLOSER THAN EVER?
2. JVM Developer
DevOps practitioner
Developer Advocate
Robot Builder
IoT speculator
AI explorer @spoole167
Work at IBM’s UK Research
and Development Laboratory
Means
I get to play
with cool stuff
Steve Poole : IBM
3. How close are we to
building Skynet?
And can anyone do it?
4. “Skynet is a fictional neural net-based conscious group mind
and artificial general intelligence (see also superintelligence)
system that features centrally in the Terminator franchise
and serves as the franchise's main antagonist.”
https://en.wikipedia.org/wiki/Skynet_(Terminator)
7. Meet ‘Handle’
Another Robot from Boston Dynamics
http://www.bostondynamics.com/
Handle is 6.5 feet tall, can jump 4 feet and travels at speeds of up to 9 mph.
and can travel up to 15 miles between charges
11. The test of the world's largest micro-
drone swarm in California in October
2016 included 103 Perdix micro-drones
measuring around six inches launched
from three F/A-18 Super Hornet fighter
jets,
13. Bot Nets
500,000+ hijacked internet-connected things
like cameras, lightbulbs, and thermostats
launched the largest DDoS attack ever
against a top security blogger
IOT
An integrated end-to-end solution that enables
your apps to communicate with, control, analyze,
and update connected electronics
14. 50 Billion devices connected to
the internet by 2020
Wikipedia: “Skynet gained self-awareness after it had spread into millions of
computer servers all across the world”
15. AI’s that can beat humans at
Chess, Go and even Jeopardy
Jan 2016
Feb1996
Feb 2011
31. My toolkit
• Java
• OpenCV
• Cuda4J
• Apache Spark
• DeepLearning4j
• Neuroph
• Processing
• Gazebo
• Raspberry PI
• Robot HAT
• Robot chassis with wheels
• Ti SensorTag
• Ultra sonic range finder
• Webcam
• Loudspeaker
• Microphone
• Batteries x many, many
62. Poorly trained
Loop 5000 times
pick random angle
get expected answer
get answer from neural net
apply positive & negative feedback
Loop throw angles 0..360
get expected answer
get answer from neural net
apply positive & negative feedback
Properly trained
67. Terminator4J RobotDelivery4J
• Training is stupidly hard
• I hadn’t appreciated how hard.
• You need something like Gazebo
• But it still takes significant time to train and AI
• Plus visualising the AI decision process is difficult if not impossible
http://gazebosim.org/
68. Computer Vision : face recognition
• How do you turn a picture of a face into a ‘key’ ?
• Especially when the face is at a different distance or relative angle?
69. “face landmark estimation"
Convert the image into HOG format
Identify key points around the face.
Train an AI to find those points in an image
Use the located points to morph the face
image into a standard form
Invented by Vahid Kazemi and Josephine
Sullivan.
http://www.csc.kth.se/~vahidk/papers/KazemiCVPR14.pdf
http://sharky93.github.io/docs/gallery/auto_examples/plot_hog.html
70. Once you can identify
the key parts of a face
you can morph the image
in other interesting ways
face landmark estimation
71. ‘Encode” the face
• With a standardised face image
• Identify the ‘important’ parts of the face so you can do pattern matching
• The challenge
• We don’t know what the ‘important’ parts are
• The answer
• Use a Neural Network to work it out.
• http://www.cv-foundation.org/openaccess/content_cvpr_2015/app/1A_089.pdf
75. Results
“love they do is”
“Open the pod bay doors hal.”
Speaker 0: Open the pod bay
doors hal.
Speaker 1: Sorry. Great I can.
“Open the pod bay doors hal.
I’m sorry Dave. I can’t do that”
Sphinx4 Watson
78. Chat Bots
User: "Alexa, open History Buff"
Alexa: "History Buff. What day do you want events for?"
User: "August thirtieth."
Alexa: "For August thirtieth, in 2003, [...] . Wanna go deeper in history
User: "Yes."
Alexa: "In 1995, Bosnian war [...] . Wanna go deeper in history?"
User: "No."
Alexa: "Good bye!"
https://github.com/amzn/alexa-skills-kit-java
79. Chat Bots: a little bit of a cheat
GetFirstEventIntent get events for {day}
GetFirstEventIntent give me events for {day}
GetFirstEventIntent what happened on {day}
GetFirstEventIntent what happened
GetFirstEventIntent {day}
GetNextEventIntent yes
GetNextEventIntent yup
GetNextEventIntent sure
GetNextEventIntent yes please
AMAZON.StopIntent no
AMAZON.StopIntent nope
AMAZON.StopIntent no thanks
AMAZON.StopIntent no thank you
81. Artificial Intelligence
• Relies on ‘good’ data representations
• Relies on appropriate internal data structures
• Relies on the ‘right’ sort of training
• Embeds knowledge inside its data structures
• Very quickly becomes opaque to humans
• Needs significant processing power
This is
‘art’
not
science
83. How long do we have left?
In the movie (set in 1984) the first Terminator came from 2024
84. Checkpoint
• Building you own cognitive robot or service is hard
• It takes time.
• It takes data.
• It takes patience.
• What AI techniques are giving us is the ability to mine data effectively and be able to teach systems to ‘understand’
that data
• It’s clear that companies are creating AI’s that will be significant assets.
• It’s also clear that differentiator will be the quality and quantity of data used to train the AI
• No sign of sentience yet.
• However…
85. • We design the AI
• We train the AI
• Our desires / agendas / biases can easily get encoded into the AI
• How do we ‘trust’ AI’s?
The Skynet idea still has an achilles heel …
Don’t worry about Skynet yet.
Worry about how we learn to understand and visualise the AIs we’ve created.