Deep Learning Applications for Smart cities

Background and Approach

This blog is based on my talk in London at the Re.work Connected City Summit on Deep Learning Applications for Smart cities. The talk is based on a forthcoming paper created with the help of my students at UPM/citysciences on the same theme. Please email me at  ajit.jaokar at futuretext.com  or follow me  @ajitjaokar  for more details.

Here are some notes on our approach:

  • When we speak of Machines – the media dramatizes the issue.  Yet,  city officials and planners plan for ten to twenty years in the future. They will have to consider many of these issues in a pragmatic way.
  • Deep Learning / Artificial Intelligence will impact many aspects of Smart cities. We decided to approach the subject in a pragmatic manner and to explore the impact of Deep Learning/AI technology on the lives of future citizens.

How could self-learning machines affect humanity in cities?

Initially, we started off with the usual Smart City approach i.e. domains such as Security – Transport – Health – Governance – Environment etc

Then, we were inspired by a statement “Man becomes the sex organs of the machine world – the bee of the plant world – enabling machines to evolve ever new forms” – Marshall McLuhan

It indicates that disruptive innovations like Deep Learning and AI cannot be viewed in silos. Instead, we decided to reframe the problem in a more disruptive way by asking the questions;

    What can Machines learn from Observations?

    What can Machines learn from Data?

    What impact does it have on new services, culture, citizens ?

    What are the threats?

    How will the lives of future citizens be impacted through self learning machines?

 

The shortest introduction to Deep learning:

Here is a brief introdcution to Deep Learning.  I have spoken of the Evolution of Deep Learning models and An introduction to Deep Learning and it’s role for future cities

Deep Learning can be seen more as a specific form of Machine Learning that leads to creating Self Learning Machines.  The whole objective of Deep Learning is to solve ‘intuitive’ problems i.e. problems characterized by High dimensionality and no rules.  With Deep learning, Computers can learn from experience but also can understand the world in terms of a hierarchy of concepts – where each concept is defined in terms of simpler concepts. The hierarchy of concepts is built ‘bottom up’ without predefined rules . This is similar to the way a child learns ‘what a dog is’ i.e. by understanding the sub-components of a concept ex  the behavior(barking), shape of the head, the tail, the fur etc and then putting these concepts in one bigger idea i.e. the Dog itself.

More specifically, a form of Deep Learning called Reinforcement Learning is making a huge impact in areas such as AlphaGo. Reinforcement Learning (RL) is based on a system of rewards. RL is a form of unsupervised learning – An RL agent learns by receiving a reward or reinforcement from its environment, without any form of supervision other than its own decision making policy.

In machine learning, the environment is typically formulated as a Markov decision process (MDP) as many reinforcement learning algorithms for this context utilize dynamic programming techniques. The main difference between the classical techniques and reinforcement learning algorithms is that the latter do not need knowledge about the MDP and they target large MDPs where exact methods become infeasible. Reinforcement learning differs from standard supervised learning in that correct input/output pairs are never presented, nor sub-optimal actions explicitly corrected. Further, there is a focus on on-line performance, which involves finding a balance between exploration (of uncharted territory) and exploitation (of current knowledge). (adapted from wikipedia)

Analysis

Here are the trends we note from the themes noted above. Link sources from Home of AI info and the web

What are machines learning from Data and Observations?

  • New computer program first to recognize sketches more accurately than a human
  • Deep Learning Algorithm ‘Paints’ in the Style of Any Artist it Copies
  • New big data system developed at MIT is more intuitive than humans
  • Artificial intelligence breakthrough as intuition algorithm beats humans in data test
  • MIT Develops Device That Can See People Through Walls
  • Lie-detecting algorithm spots fibbing faces better than humans
  • Machines That Can See Depression on a Person’s Face
  • An algorithm aims to be able to replace human intuition
  • ‘Psychic Robot’ System Guesses Intentions From Your Movements
  • MIT’s intelligent drone can avoid crashes and fly at 30 MPH
  • Facebook working on AI that can tell what’s in photos
  • Computer Algorithms Could Aid Schizophrenia Diagnose
  • Machines That Can See Depression on a Person’s Face
  • Robot Radiologists Will Soon Analyze Your X-Rays
  • Predicting change in the Alzheimer’s brain
  • A new computer program that can diagnose cancer in just two days!
  • Machine learning to help predict online gambling addiction
  • Predicting people’s daily activities with deep learning
  • MIT Scientists Create An AI System That Can Determine How Memorable Your Face Is
  • This Algorithm Is Better At Predicting Human Behaviour Than Humans Are
  • New Artificial Intelligence: Russia Endows Robots With Collective Mind
  • Scientist Develop New Machine Which Can Calculate Pattern Recognition with Near Human speed
  • Machine Vision Algorithm Learns to Recognize Hidden Facial Expressions
  • Artificial Intelligence: Scientists Developed a Handwriting Algorithm
  • Computer With Built-In Algorithm Beats Man In A Turing Test
  • Machine learning to differentiate between positive and negative emotions using pupil diameter

 

Self learning for Robots(from observation)

  • Giving robots a more nimble grasp
  • Why it is hard to teach robots to choose wisely
  • Machine learning plays vital role in the evolution of Man
  • Designing Robots That Learn as Effortlessly as Babies
  • How Robots Can Quickly Teach Each Other to Grasp New Objects
  • Why IBM just bought billions of medical images for Watson to look at
  • Read my lips: truly empathic robots will be a long time coming

 

Learning Culture, Humanity, emotions and ethics

  • Smart Programs Read Shakespeare
  • Artificial intelligence learns how to put together interactive stories just as good as a human
  • How do you teach a machine to be moral?
  • ‘Psychic Robot’ System Guesses Intentions From Your Movements
  • Lie detection software learns from real court cases
  • Why Helping Humanity Should Be Core to Learning
  • Could Artificial Morals and Emotions Make Robots Safer?
  • AI: In search of the sarcasm algorithm
  • Microsoft Teaches Computers To Be Funny
  • Microsoft’s Project Oxford Can Now Detect Emotions from Photos
  • Robots are learning to disobey humans: Watch as machine says ‘no’ to voice commands
  • Robots could be converted to religion someday: Scientists
  • Intimacy & Falling In Love With A Robot Could Happen In 50 Years Because Of Artificial …
  • Health
  • If We Want Humane AI, It Has to Understand All Humans
  • Humai Is Working On A Way To Bring Your Loved Ones Back From The Dead
  • Mum Robot Goes Darwinian on Her Kids

How does that (self) learning affect services and our lives in future cities

  • Artificial intelligence comes to toys
  • Beyond the Pill: Data Is the New Drug – Google Life Sciences Rebrands As Verily, Uses Big Data To Figure Out Why We Get Sick
  • Nvidia Aims To Power Flying Vehicles with Jetson TX1 Board
  • Motorcycle-riding robot may take on world champion racer
  • Meet Mercedes-Benz’s Vision Tokyo, a self-driving car for the megacity
  • How artificial intelligence could lead to self-healing airplanes
  • Trains with brains: how Artificial Intelligence is transforming the railway industry
  • A self-driving sailboat to patrol the oceans and monitor the environment
  • Malaysia testing ‘artificial intelligence’ for prisons
  • Real-Time Seizure Detection Possible with Learning Algorithm
  • Facebook Is Helping People With Blindness “See” the Photos on Their Walls
  • Mitsubishi Electric uses machine-learning tech to detect distracted drivers
  • Tinder matches made easy with new intelligent algorithm
  • Deep Learning Algorithm Successfully Identifies Potential Intracranial Haemorrhaging
  • An artificial intelligence based third Umpire
  • When children talk to toys, some are talking back
  • Predicting change in the Alzheimer’s brain
  • Robotic Automation Meets Agriculture
  • Food delivered by drones, driverless cabs and cyber PAs to organise your party: A revolution in …
  • AI will soon be forecasting the weather
  • How Artificial Intelligence Can Fight Air Pollution in China
  • Starfish-killing robot to protect Great Barrier Reef
  • Self-Driving Car Tech Allows Vehicle To ‘See’ Environment In Real Time
  • US Company On Plan To Bring People Back From Dead Using Artificial Intelligence
  • A trillion tiny robots in the cloud: The future of AI in an algorithm world
  • Teforia Is A Tea Brewing Robot That Uses Algorithms To Pour The Perfect Cup
  • Japanese artificial intelligence passes university exams (but still can’t quite get into the country’s …
  • Facebook AI built to help visually impaired people
  • Problem of Climate Change and Global Conflicts Can Be Solved Using Human and Computer …

 

Risks to humanity and cities

  • ‘Only movies build bad robots‘ – famous last words?
  • Why human-in-the-loop computing is the future of machine learning
  • As Robots Steal Millennials’ Jobs, Young Workers Focus On Skills, Not Careers
  • Millions of jobs at risk from artificial intelligence
  • Davos report projects 5 million jobs will be lost to new technologies by 2020
  • Can Humanity Rein In The Rise Of The Machines?
  • Christian leader warns of ‘Frankenstein monsters’ due transhumanism
  • The rise of the killer robots — and why we need to stop them
  • Producer of Russia’s Armata T-14 plans to create army of AI robots
  • Inside the Pentagon’s Effort to Build a Killer Robot
  • How Technology Could Prevent Another Paris-Like Attack
  • Kaspersky deepens security offering through machine learning
  • Robots will declare war on humans within 25 years, claims artificial intelligence expert
  • Law firm bosses envision Watson-type computers replacing young lawyers
  • Hitachi Hires First ‘Artificial Intelligence’ Boss To Manage Workers

Conclusion and Evolution

We reframed the problem of Deep Learning and Smart cities by asking the Question:

How could self-learning machines affect humanity in cities?

    What can Machines learn from Observations?

    What can Machines learn from Data?

    What impact does it have on new services, culture, citizens

    What are the threats?

Please contact me at ajit.jaokar at futuretext.com to know more updates – especially if you are a city official. We are also planning to explore the implementation of these ideas by working with companies like Nvidia.

I would also like to thank the students who helped me with this project.