Managing Complexity

Managing Complexity In Your Head

Finding it hard to understand a complex phenomenon?

How do you manage complexity in your head?

1. Visualization / Imagination

Imagination by Katrina Kaif

 

‘The true sign of intelligence is not knowledge but imagination’ – Imagination by Katrina Kaif

Language and words are discrete. Visualize, so you see ‘all’, ‘the whole’ at once.

  • Organization
  • Connected Structure – how parts are connected to form the whole; what happens when a part is changed.
  • Chunking

2. Abstraction

Create Abstractions to help you see whole in terms of abstract concepts, when it’s hard to see ‘whole’ at once.

  • Visualization is usually top-down; creating abstractions can be done both bottom-up and top-down.
  • Recursive definitions are one type of abstraction.
  • Naming.
  • Concept borrowed from Computer Science; applicable across disciplines.

3. Generalization

14724563_1788317928112078_2697529884478836366_n

 

‘Have a good day , and a great week , and just a wonderful life in general ! ‘ – Generalization by Katrina Kaif

Abstract classes and classes of behaviors into Generalizations.

Science is all about discovering generalizations.

4. Systems Thinking

Divide the whole into systems and the systems into sub-systems and their interactions – Systems Thinking.

5. Knowledge Ontology

Organize generalizations into Knowledge Ontology.

Move up and down between Multiple Levels Of Abstractions.

Concept borrowed from Artificial Intelligence (AI).

6. Point of View / Perspective

16641067_1843139239296613_5681191096692787784_n

Perspective by Katrina Kaif

Find new Points of view from which to look at the domain under consideration (examples include how American Physicist Richard Feynman found new way of looking at interactions between light and matter which helped him discard infinities and formulate QED).

 

16299489_1838457319764805_5693086054311300362_n

 

‘Seeing the world through different glasses’ – Point of view, zooming in and out by Katrina Kaif

  • Understand part (representative element / elements) to understand the whole – Part – Whole.
  • Lens tool – “zoom in” and “zoom out”.

 

 

 

 

14714958_1790952407848630_8243309853551691027_o

 

‘If plan A didn’t work stay cool, the Alphabet has 25 more letters’ – Google Alphabet by Katrina Kaif

  • Problem – Sub-problem
    • Examples
      • Recursion
      • Dynamic Programming
      • Mathematical Induction

 

  • Individual Element – Whole 
    • Examples
      • Matrix Multiplication: individual element = respective row * column
      • Proof of Inclusion Exclusion Principle.
      • Coloring Principle (Problem Solving Heuristic)
      • Telescoping Tool (Mathematical Problem Solving; Series Summation)
  • Local Behavior – Global Behavior
    • Invariance Principle (Problem Solving Heuristic)
      • Change in individual states – Invariant Global function.
    • Iteration in Computing
      • Thinking in terms of change in state in an iteration.

14712486_1786815188262352_863380145563592690_o

 

At the end of the day: ‘Hold on, I’ve gotta overthink about it!’ – Managing complexity by Katrina Kaif

Guidelines For Research In Neuroscience

Area Of Expertise: # Computational Neuroscience #NeuroEngineering  # Neuroscience

Guidelines For Research In Neuroscience:

Simulation is not enough (Specific cases),

Start building Models & Theories at successive levels of Abstractions (Generalizations).

  • Trying to understand the brain by creating exact models of the brain with enormous computing power (costing Billions of dollars) is not going to work – as is intended. 
    • We are never going to understand how the brain works completely with data only from the lowest level (molecules, channels, neurons) with so much complexity involved.
    • There is so much genetic variation from mouse to mouse, primate to primate, that you can’t draw general conclusions from data of genetic expression of a single mouse. What happens in the brain data scanning initiatives is that data is gathered from a single organism. And what is required is something similar to functional genomics – sort of functional neuroscience – trying to understand the relation between behavior and what happens in the brain – not just cataloging what data from a particular brain looks like.
    • What we need is new models, new theories, new abstractions – that can explain all these data.
    • We don’t need to simulate large parts of the brain on computers. Our goal should be simulation of small parts and theoretical models that can explain data from those small parts of the brain. 
  • We need to start building models, theories, abstractions. And then on top our first attempts at building models and theories, we will start building more accurate models, models that connect data from different levels of the brain. 
    • Our first attempts at building models, abstractions might concentrate on data from only one level. Next, our newer models would connect different levels of structural abstractions and their corresponding different levels of functional abstractions found in the brain.
    • Different structural levels of abstractions found in the brain:
      • Molecules, Receptors, Neurotransmitters. 
      • Neuron, Channels, Synapse, Glial Cells.
      • Collection of neurons
      • Brain regions (e.g., Primary Visual Cortex)
      • Brain – Behavior;  
  • Neuroscientists individually work on a tiny part at “only one level” (among all these levels, from molecules and neurons to whole brain) of the brain. We need scientists who can connect different levels of structural abstractions. 
  • The new breed of Neuroscientists, with the aim of building models, abstractions, theories of the brain, would try to learn how scientists with different backgrounds are studying Neuroscience.     
    • What diseases are Neurologists seeing in patients? How do the Neurologists explain them in terms of lesions, etc. in a particular brain region
    • Examples:
      • Speech – Broca’s area [2].
      • Synesthesia [3] – Cross-connections among nearby brain regions.
    • What diseases are Psychiatrists seeing in patients? How do they explain them in terms of excess or reduction in neurotransmitters? 
      • Examples: 
        • Schizophrenia – Excess of Dopamine [4]. 
    • Data from neurons, channels, molecules.
    • Data from specific brain regions (e.g., MRI, fMRI data).
    • Data from optogenetics – switching neurons on and off with light. 
    • Systems Neuroscience
    • Computer Models of brain. Connectomics.
    • What are we learning from our research in Artificial Intelligence about the requirements of intelligence? 
    • Cognitive Neuroscience – latest research on higher mental functions and brain. 
  • Psychologists have built models. Researchers interested in both Artificial Intelligence and Neuroscience (e.g., Marvin Minsky [1]) have built models. Why not start by trying to explain those models with our understanding of the brain?


Large Neuroscience Projects



References

Managing Complexity

Managing Complexity In Your Head

Finding it hard to understand a complex phenomenon?

How do you manage complexity in your head?

1. Visualization / Imagination

Language and words are discrete. Visualize, so you see ‘all’, ‘the whole’ at once.

  • Organization
  • Connected Structure – how parts are connected to form the whole; what happens when a part is changed.
  • Chunking

2. Abstraction

Create Abstractions to help you see whole in terms of abstract concepts, when it’s hard to see ‘whole’ at once.

  • Visualization is usually top-down; creating abstractions can be done both bottom-up and top-down.
  • Recursive definitions are one type of abstraction.
  • Naming.

3. Generalization

14724563_1788317928112078_2697529884478836366_n

‘Have a good day , and a great week , and just a wonderful life in general ! ‘ – Generalization by Katrina Kaif

Abstract classes and classes of behaviors into Generalizations.

Science is all about discovering generalizations.

4. Systems Thinking

Divide the whole into systems and the systems into sub-systems and their interactions – Systems Thinking.

5. Knowledge Ontology

Organize generalizations into Knowledge Ontology.

Move up and down between Multiple Levels Of Abstractions.

6. Point of View / Perspective

16641067_1843139239296613_5681191096692787784_n

Perspective by Katrina Kaif

Find new Points of view from which to look at the domain under consideration (examples include how Feynman found new way of looking at interactions between light and matter which helped him discard infinities and formulate QED).

 

  • Understand part (representative element / elements) to understand the whole – Part – Whole.
  • Lens tool – “zoom in” to “zoom out”.
  • Problem – Sub-problem
    • Examples
      • Recursion
      • Dynamic Programming
      • Mathematical Induction
  • Individual Element – Whole 
    • Examples
      • Matrix Multiplication: individual element = respective row * column
      • Proof of Inclusion Exclusion Principle.
      • Coloring Principle (Problem Solving Heuristic)
      • Telescoping Tool (Mathematical Problem Solving; Series Summation)
  • Local Behavior – Global Behavior
    • Invariance Principle (Problem Solving Heuristic)
      • Change in individual states – Invariant Global function.
    • Iteration in Computing
      • Thinking in terms of change in state in an iteration.

Vision Of Collaborative Research Platform

With the proliferation of global collaboration platform the World Wide Web, we need to direct research in the right direction – one that centers on solving real problems collaboratively, not just publication of random research papers (for getting promoted!).

A collaborative research platform could define substantial problems, each of them, if solved, can bring about revolutions, and break the substantial problems into manageable pieces, i.e., sub-problems and the sub-problems into even more manageable pieces, i.e., sub-sub-problems. Individual researchers or research groups could pick and work on manageable problems that match their abilities and interests and publish and share results on the platform. Researchers could define new problems, define new sub-problems, suggest improvements and changes on the platform.

A research paper could be co-authored by hundreds or even thousands of researchers scattered throughout the world.

For example, a high level problem could be “Codify Biology to the point that you can control biological processes and organisms”. The problem should have measurable, quantifiable goals. This high level problem could be divided into more manageable pieces and then into even more manageable pieces until they are solvable by individual researchers or research groups.

This would help the entire scientific community move forward towards practical goals much more rapidly.

What Google Could Focus On Next

Research Problems

  • Cracking Artificial Intelligence with great internal teams & external collaboration – new products, increasingly better existing products – essentially changes the world.
  • It was search, rules, first order logic, probability and other representations. Now, it’s all about learning from massive datasets. Revolutions are coming! Ultimate search engine!
  • Intelligent Personal Agent
  • Solving big problems (or as Google calls them “moon shots“) with (problem focused collaborative research) great internal teams & external collaboration / massive collaboration / open innovation
  • Solve for X, Google X
  • Application of Big Data: understanding complex systems better: finding answers to age old questions: Sociology, Behavioral Science, Political Science, Economics & Business
  • Exploration of new computing architectures – continuation of Moore’s law: exponential increase in processing power – computational power for data processing, intelligence.
  • Parallel computing architectures (e.g., GPUs); Molecular computing; Optical Computing; Quantum computing; Cognitive computing / Neuromorphic computing.
  • Brain-computer Interfacing
  • Google in our brain!
  • Integrating data from all the sources => Knowledge Mining, New Applications

CEO

People focused Social Problems

  • Education & Learning platform
    • Google Play for Education, ChromeBook. Singularity University. Technology and Education are great equalizers.
  • Platforms that empower people – the enormous potential in each and everyone of us is materialized
  • Wikinomics platform 
    • Global collaboration
    • Android, Play, Chrome
  • Products that take on social problems and scale using profit to reach millions
  • Funding / Growing / Working with entrepreneurs to solve problems people face.
  • Solving local, social and global problems (clean water, cheap energy etc.) utilizing massive collaboration and exponential technologies
    • Google.org. Google Ideas. For-profit.

Healthcare & Biomedicine

  • Computational Biology / Bioinformatics, Systems Biology, Biology Engineering
  • Healthcare Informatics
  • Lifespan extension; Curing Diseases; Preventive Medicine; Personalized Medicine

Physical Digital Integration

  • Robotics, Automation
  • Car 2.0
  • Electric / hybrid Car? App Market Integration; Automation; Advanced Collision avoidance system
  • Ultimate goal: Self Driving Car, Intelligent Transportation System
  • Computational Materials Design
  • Internet of things; Smart cities; Smart Home; Sensor Web; Big data

 

Google bought 8 robotics companies in past 6 months.

Google started with Search; currently focusing more on platforms and services. Lately, excited about the prospects about Physical Digital Integration?

(Developers Users, Apps).
Next: Developers developing tools, components Users developing apps?

Platforms:

  • Android
  • Google Play (Streaming services?)
  • Chrome
  • Google App Engine
  • Google+
  • Google TV
  • (New) Google glass, Motorola, ChromeOS, ChromeBook

 

Services:

  • Maps
  • GMail
  • Youtube
  • Google Books

 


 

Followups

Fundamental Shift In Education: A Requirement Of Our Time

Fundamental shift in education is a requirement of our time.Searchable facts are available online. When you have the Internet at your fingertips aided by Google you can learn anything ever known to mankind.

Computational devices / apps {e.g., WolframAlpha} are available. We don’t need to do multiplications or divisions or even algebraic calculations anymore. Computer based Numerical Computing and Symbolic Algebra Systems (SageMathematicaMatlab) and Statistical Computing Systems (R) are available. All we need to learn is to turn real world problems into computer programs that can solve our problems.

 

Exponential growth of human knowledge means that much of what you know today would become obsolete ten years from now. So lifelong education is the norm.  

Breadth of knowledge will become increasingly important. You need to know what you need to know and once you know what you need to know, you can find it online.

Boundaries between different fields of study are opening up. We organized different fields of study (Physics, Chemistry, Biology etc.) more than hundred years ago. We organized everything related to life under Biology. The study of atoms and molecules and reactions was named Chemistry. Physics concerned itself with forces and matter of the universe and their interactions with the ultimate goal of reducing all forms of interaction into few fundamental laws (one law – “The Theory Of Everything” or so the Physicists say – if you are a bit ambitious!). But now in the light of new knowledge and better tools, new organization of fields of study is required.

As we gained knowledge about the chemical processes of life, we organized our knowledge under the name Biochemistry. Fields that used to be completely separate are merging in the light of new knowledge and better tools to tackle common problems. For example, Biology and Computer Science, once separated, have combined to form Computational Biology / Bioinformatics as a result of explosion both in genomic and more generally biological data and computational power. New fields of study like Bioengineering, Biophysics, Cognitive Neuroscience, Computational Materials Science have popped up. As we look forward into the future, we can see Computer Science and Physics combining to form Physical Digital Integration and creating a world that is much different and better than ours. 

The Education of future Engineers would be based on Science and Mathematics: learning a few concepts that can explain all the different tools, technologies, possibilities and limits.

With the proliferation of global collaboration platform The World Wide Web, we need to direct research in the right direction – one that centers on solving real problems collaboratively, not just publication of random research papers (for getting promoted!).

 
In short, a complete reorganization of Education as it is today. Are we ready?


Followups

Research Problems I Want To Work On

  1. How do you create Advanced Artificial Intelligence that is better than human experts and understands and can reason about everything on the Web and in the real world?
  2. How do you Codify Biology at different levels of abstraction (DNA, Proteins, Metabolic and Signalling Pathways, Cells, Tissues, Organs, Body) so that you can predict and control? (Codification of Biology, Engineering Biology, Systems Biology, Computational Biology, Curing Diseases & Disabilities, Increasing Lifespans.) 
  3. How do you invent better tools, technologies (Imaging, Optogenetics) and Models for Understanding and Engineering the BrainHow do you Cure Neurological and Psychiatric Disorders?
  4. How do you Understand Complex Systems consisting of lots of interacting agents? (Application of Big Data; Inventing better Models, Mathematics, Algorithms to understand complex systems better; finding answers to age old questions in Sociology, Behavioral Science, Political Science, Economics & Business)
  5. How do you Design Materials and Nanostructures with required properties using Computers?
  6. How do you make Software Developers say 100 times more productive?
  7. How do you design an Effective Education and Learning and Research Platform?
  8. How do you create Platforms that empower people – so that the enormous potential in each and everyone of us is materialized?
  9. How do you take Human – Computer (networked) collaboration, intelligence to the next level?
  10. How do you integrate the Information World and the Physical World? (A world where the Physical world is completely aware of everything utilizing information from the Information World; the world of information is embedded in the Physical World.)
  11. How do you design Next Generation Manufacturing Technologies? (Fab-lab, 3D Printing, Automation)
  12. Exploration of new computing architectures – continuation of Moore’s law: exponential increase in processing power; computational power for data processing, intelligence; (Parallel computing architectures (e.g., GPUs); Molecular computing; Quantum computing; Cognitive computing / Neuromorphic computing)
  13. How do you solve Local, Social and Global problems (clean water, cheap energy etc.) utilizing massive collaboration and exponential technologies?
  14. How do you predict and prevent Natural Disasters?
  15. Joining the pieces of puzzle together to get a complete picture of the Ultimate Reality