This was a top level presentation on some of the 30+ subcategories of Artificial Intelligence at the Hackaday LA June Meetup - Wheels, Wings, and Walkers. Sponsored by SupplyFrame Design Labs in Pasadena CA
15. Underwater Toxic Metal Detector Robot
Boats use copper based coatings to prevent barnacle growth on their hulls.
Over the course of many years the copper leaches out of the paints and into the marine
environments.
Marina Del Rey has one of the highest concentrations of copper which is toxic to most marine life.
21. Machine Learning (Find Patterns)
- Neural Networks
- Deep Learning
- Reinforcement Learning
- Gradient Boost Machines
- Support Vector Machines
- Conformal Prediction
Reasoning
Robotics
Computer Vision
Internet of Things
- Everything connected
Natural Language Processing
- Chatbots
Philosophy
Knowledge Engineering
Rules Engines
Logic Programming
Multi Agent Systems
Turing Tests
…… and many more
30+ Subcategories of Artificial Intelligence
22. Machine Learning (Find Patterns)
- Neural Networks
- Deep Learning
- Reinforcement Learning
- Gradient Boost Machines
- Support Vector Machines
- Conformal Prediction
Reasoning
Robotics
Computer Vision
Internet of Things
- Everything connected
Natural Language Processing
- Chatbots
Philosophy
Knowledge Engineering
Rules Engines
Logic Programming
Multi Agent Systems
Turing Tests
…… and many more
30+ Subcategories of Artificial Intelligence
23. Developed in the 1940’s
Think of them as “Giant Rocket Engines”
- Andrew Ng, Chief Scientist, Baidu
Neural Networks
24. Limited cup of rocket fuel, ie data
Rocket engines sputter because of lack of fuel.
Artificial Intelligence gets a bad wrap a few times referred to as AI Winters
25. 2006 - BIG DATA
Explosion of data from Twitter, Facebook, Youtube, etc is making these “rocket engines”
come alive.
Amazing advances in cancer detection, speech recognition with background noise, etc
26. Big Data
2.5-quintillion bytes of data are being created every day
90% of the data in the world today has been created in the
last two years alone
27. Machine Learning (Find Patterns)
- Neural Networks
- Deep Learning
- Reinforcement Learning
- Gradient Boost Machines
- Support Vector Machines
- Conformal Prediction
Reasoning
Robotics
Computer Vision
Internet of Things
- Everything connected
Natural Language Processing
- Chatbots
Philosophy
Knowledge Engineering
Rules Engines
Logic Programming
Multi Agent Systems
Turing Tests
…… and many more
30+ Subcategories of Artificial Intelligence
28. Deep Learning
Used in Deutsch VW RRR Campaign which recognizes human vocalized car sounds.
29. Deep Learning
- Developed by Geoff Hinton of Univ of Toronto & Google
In 1959 David Hubel & Torsten Wiesel discovered “simple cells” and “complex cells” in cat visual system
Software that emulates the Cat’s visual cortex system.
The first layer of a cats eye recognizes edges of objects.
The next layer recognizes what’s attached to that edge, is it a nose or an eyeball, etc
The next layer recognizes whether the eyeball is attached to another cat or human…..
30.
31. VW RRR
Audio Recognition Image
Recognition problem
- Collected 1000s of audio samples of people making 3 types of car sounds, ie
acceleration, deceleration & screeching!
- Process converted audio wave files into frequency spectograms.
- Train using supervised learning methodology, ie tell the Deep Learning engine that this
image represents a human making a screeching car sound, etc.
32. Machine Learning (Find Patterns)
- Neural Networks
- Deep Learning
- Reinforcement Learning
- Gradient Boost Machines
- Support Vector Machines
- Conformal Prediction
Reasoning
Robotics
Computer Vision
Internet of Things
- Everything connected
Natural Language Processing
- Chatbots
Philosophy
Knowledge Engineering
Rules Engines
Logic Programming
Multi Agent Systems
Turing Tests
…… and many more
30+ Subcategories of Artificial Intelligence
36. C. DEDUCTIVE : conclusion
guaranteed
Reasoning
So what is Sherlock Holmes doing?
B. INDUCTIVE : conclusion merely likely
A. ABDUCTIVE : taking your best shot
37. C. DEDUCTIVE : conclusion
guaranteed
Reasoning
So what is Sherlock Holmes doing?
B. INDUCTIVE : conclusion merely likely
A. ABDUCTIVE : taking your best shot
38. Machine Learning (Find Patterns)
- Neural Networks
- Deep Learning
- Reinforcement Learning
- Gradient Boost Machines
- Support Vector Machines
- Conformal Prediction
Reasoning
Robotics
Computer Vision
Internet of Things
- Everything connected
Natural Language Processing
- Chatbots
Philosophy
Knowledge Engineering
Rules Engines
Logic Programming
Multi Agent Systems
Turing Tests
…… and many more
30+ Subcategories of Artificial Intelligence
39. Czech playwright Karel Čapek introduced the word robot in 1920 in his play
Rossum’s Universal Robots
50. Flippy - a burger-grilling robot from Miso Robotics
@ Caliburger in Pasadena
51. HANDEDNESS
Would your robot be left handed or right handed?
Why?
If it doesn’t matter why did evolution make us one
or the other?
The idea of handedness in humans is tied to the
use of one hemisphere of the brain over another,
known as "lateralisation."
55. Machine Learning (Find Patterns)
- Neural Networks
- Deep Learning
- Reinforcement Learning
- Gradient Boost Machines
- Support Vector Machines
- Conformal Prediction
Reasoning
Robotics
Computer Vision
Internet of Things
- Everything connected
Natural Language Processing
- Chatbots
Philosophy
Knowledge Engineering
Rules Engines
Logic Programming
Multi Agent Systems
Turing Tests
…… and many more
30+ Subcategories of Artificial Intelligence
56. Modelling Human Visual System in Software is Difficult
The bar in the middle is only one color.
But when placed on a gradient background your brain makes it appear to have a gradient itself
61. WATSON COGNITIVE DRESS
iOT Technology Roadmap - 28 Mar 2016
Computer
WIRELESS OPTIONS
xBee / Zigbee
WiFi
Bluetooth
Cellular/GSM
Radio RF
MicrocontrollerIBM Watson Components
OPTIONS
Mbed series
Odroid XU4
RaspberryPi 3
Arduino Mega
BeagleBoneBlack
Intel Edison
Particle Photon
Particle Electron
WIRED
Ethernet
CONTROL CENTER
DRESS
OPTIONS
Servos
Motors
Muscle Wire
Switches
Sensors
SmartFilm
Power
OPTIONS
Voltage
Amperage
LiPo
NOTES:
1. Design considerations - # of components to determine weight, heat, operation time & power constraints
2. Is the communication to the dress in one direction? Or will the dress send sensor data back to Watson for further cognitive training?
3. With each dress design iteration, a full matrix of an optimized hardware solution will be generated, detailing wireless protocol,
microcontroller, battery size, etc.
65. Some of today’s players in the fashion wearable space
Julia Koerner
https://www.juliakoerner.com/
Anouk Wipprecht
http://www.anoukwipprecht.nl/
CuteCircuit
http://cutecircuit.com/
Elektrocouture
https://elektrocouture.com/
66. Machine Learning (Find Patterns)
- Neural Networks
- Deep Learning
- Reinforcement Learning
- Gradient Boost Machines
- Support Vector Machines
- Conformal Prediction
Reasoning
Robotics
Computer Vision
Internet of Things
- Everything connected
Natural Language Processing
- Chatbots
Philosophy
Knowledge Engineering
Rules Engines
Logic Programming
Multi Agent Systems
Turing Tests
…… and many more
30+ Subcategories of Artificial Intelligence
67. Natural Language
Natural Language Processing - NLP
- Voice to Text, Text to Voice, Translation
Natural Language Understanding - NLU
- Understand the context of whats being said
Natural Language Dialog - NLD
- AI generates natural sentences.
69. CHATBOTS
LOEBNER PRIZE
- Annual competition
Awards prizes considered by the judges to be the most human-like.
- Controversy
Promotes deceit vs true conversation
- Mitsuku by Steve Worswick http://www.mitsuku.com/
2013/2016 winner built on free pandorabots.com chat
platform
70. Why Natural Language Understanding (NLU) is hard!
What is “it” in each case?
“The trophy doesn’t fit in your suitcase because it is too large”
“The trophy doesn’t fit in your suitcase because it is too small.”
- Yann LeCunn, Facebook Artificial Intelligence Research (FAIR)
77. POSSIBLE MAJOR DISRUPTOR IN DEEP LEARNING SPACE
Bayesian Program Synthesis (BPS)
Deals in Probabilities
Company: Gamalon
Funding: $12 mil from Darpa & Felicis Ventures
Trains on few pieces of data (one shot) with same accuracy as Deep Learning
Can train on one iPad vs many servers required for Deep Learning.
100 times more efficient than Google’s TensorFlow
Cleaning enterprise unstructured data is their current business model.
78. Robots that can TASTE & SMELL.
Computer learns to recognize sounds by watching video
Quantum based, Machine Learning.
qBit is a 1 & 0 …….at the same time! ← SPOOKY!
Answer Set Programming
Vatican weighing in on the religious aspects of A.I.
On the horizon!
80. How to get started in Machine / Deep Learning!
Deep Learning in the Browser
http://cs.stanford.edu/people/karpathy/convnetjs/
Free Coursera Online Class
https://www.coursera.org/learn/machine-learning
Kaggle Competition
https://www.kaggle.com/
81. Open Source Deep Learning Software
Google
Tensorflow
https://www.tensorflow.org/
Facebook
Deep-learning modules for Torch
https://github.com/facebook/fbcunn
82. Cloud based AI Tools
- all have free trials.
IBM WATSON
https://www.ibm.com/watson/
MICROSOFT
https://www.microsoft.com/en-us/ai
AMAZON
https://aws.amazon.com/amazon-ai/
GOOGLE
https://cloud.google.com/products/machine-
learning/
84. Organizations
ASSOCIATION FOR THE ADVANCEMENT OF ARTIFICIAL INTELLIGENCE
http://www.aaai.org/
ALLEN INSTITUTE OF ARTIFICIAL INTELLIGENCE
http://allenai.org/
MACHINE INTELLIGENCE RESEARCH INSTITUTE
https://intelligence.org/
FUTURE OF LIFE INSTITUTE
https://futureoflife.org
85. Conferences
NIPS 2017
https://nips.cc
O’REILLY ARTIFICIAL INTELLIGENCE CONFERENCE
https://conferences.oreilly.com/artificial-intelligence/ai-ny
IEEE INTERNATIONAL CONFERENCE ON ROBOTICS AND AUTOMATION( ICRA )
http://www.icra2017.org
LIST OF MANY MORE CONFERENCES
http://www.kdnuggets.com/meetings/
The central problems (or goals) of AI research include reasoning, knowledge, planning, learning, natural language processing (communication), perception and the ability to move and manipulate objects.
The field was founded on the claim that human intelligence "can be so precisely described that a machine can be made to simulate it."
The central problems (or goals) of AI research include reasoning, knowledge, planning, learning, natural language processing (communication), perception and the ability to move and manipulate objects.
The field was founded on the claim that human intelligence "can be so precisely described that a machine can be made to simulate it."
The central problems (or goals) of AI research include reasoning, knowledge, planning, learning, natural language processing (communication), perception and the ability to move and manipulate objects.
The field was founded on the claim that human intelligence "can be so precisely described that a machine can be made to simulate it."
The central problems (or goals) of AI research include reasoning, knowledge, planning, learning, natural language processing (communication), perception and the ability to move and manipulate objects.
The field was founded on the claim that human intelligence "can be so precisely described that a machine can be made to simulate it."
The central problems (or goals) of AI research include reasoning, knowledge, planning, learning, natural language processing (communication), perception and the ability to move and manipulate objects.
The field was founded on the claim that human intelligence "can be so precisely described that a machine can be made to simulate it."
The central problems (or goals) of AI research include reasoning, knowledge, planning, learning, natural language processing (communication), perception and the ability to move and manipulate objects.
The field was founded on the claim that human intelligence "can be so precisely described that a machine can be made to simulate it."
Inductive reasoning begins with observations that are specific and limited in scope, and proceeds to a generalized conclusion that is likely, but not certain, in light of accumulated evidence.
Inductive reasoning begins with observations that are specific and limited in scope, and proceeds to a generalized conclusion that is likely, but not certain, in light of accumulated evidence.
The central problems (or goals) of AI research include reasoning, knowledge, planning, learning, natural language processing (communication), perception and the ability to move and manipulate objects.
The field was founded on the claim that human intelligence "can be so precisely described that a machine can be made to simulate it."
The central problems (or goals) of AI research include reasoning, knowledge, planning, learning, natural language processing (communication), perception and the ability to move and manipulate objects.
The field was founded on the claim that human intelligence "can be so precisely described that a machine can be made to simulate it."
https://en.wikipedia.org/wiki/Mirror_test
The central problems (or goals) of AI research include reasoning, knowledge, planning, learning, natural language processing (communication), perception and the ability to move and manipulate objects.
The field was founded on the claim that human intelligence "can be so precisely described that a machine can be made to simulate it."
the network of physical objects—devices, vehicles, buildings and other items
—embedded with electronics, software, sensors, and network connectivity that enables these objects to collect and exchange data.
https://chargetech.com/ 110volt chargeable power
https://chargetech.com/ 110volt chargeable power
The central problems (or goals) of AI research include reasoning, knowledge, planning, learning, natural language processing (communication), perception and the ability to move and manipulate objects.
The field was founded on the claim that human intelligence "can be so precisely described that a machine can be made to simulate it."