Emergent Gravity


I was watching on internet a public lecture by Erik Verlinde on Entropic Gravity which he recently gave at Perimeter Institute in Canada. Apparently physicists are struggling with some great cosmological mysteries. He has a new theory on gravity often taught as a fundamental force in physics textbooks. According to him, gravity is an emergent phenomenon which arises from entangled quantum bits. It is emergent because it is not a fundamental force, but arises out of more basic and microscopic fabric of the universe. Being a computer scientist, I am intrigued by the central role of entropy, that is, information content, in this new perspective. Only with time, further work and observations, we will know whether the new ideas will prevail. All part of scientific progress.

“The most beautiful thing we can experience is the mysterious. It is the source of all true art and science. ” Famous quotation by Albert Einstein. We are restless beings, propelled onward by desire to understand, create, express and explore.

M’illumino, d’immenso; by Giuseppe Ungaretti. I flood myself with the light of the immense. Beautiful exclamation by the poet.

What we understand about the universe is literally based on the light of the immense! But visible light makes up only 0.005% of the universe. Next 5% is matter, and the remaining 95% is complete mystery.

Even though it may be all very mysterious, it is great tribute to human mind to reach this stage in which we can formulate these mysteries and propose new theories. The light of the immense will continue to make us march onward!



Kabir on Transience


Had-had tape so auliya

Behad tape so peer

Had-anhad dono tape

Wako naam Fakir.


One who transcends limits is protector

One who transcends limitless is spiritual guide

One who transcends both limits and limitless

Is called Fakir.


Kabir, the ultimate Fakir and a renowned mystic poet from 15th century India, saw clearly how temporary everything is.  As children growing up in India, we had Kabir’s Dohe (couplets) in our textbooks. Here is translation of one of his poems Mat Kar Maya ko Ahankaar, Mat kar Kaya ko Abhimaan. Liberty is taken not to translate verbatim but to capture the meaning. It was fascinating to read 500 year old Hindi.

Kabir first tells about a powerful king in past and how small is human life span in backdrop of flow of time.

There was a strong emperor

All powerful and ruler of lands

Elephants in his majestic court

All in the end

As ephemeral as dew drops

Then Kabir reminds us that we all face grief over losses due to impermanence of things.

Parents and family

And all attachments

Why feel grief over loss?

Everyone eventually passes away

As ephemeral as dew drops

Finally, Kabir tells us that message of humility is what one should learn.

Don’t be so arrogant of wealth

Don’t be so proud of looks

This body is more frail than clay

Just a gust of wind, even a small one,

Can turn you into dust

Falling Autumn Leaves Cad Frunze Toamna Ruginie Pictures

Sahir Ludhianvi and Mystery of Time


When I was growing up in India, like everyone around me, I was fortunate to experience some great music from Bollywood. Wonderful voices of Rafi Sahib, Asha Ji, Kishore Da, Lata Ji, and many other great singers, combined with poetry of many talented writers, provided a deep aesthetic experience for our growing minds. Radio used to be always on in my childhood home.

In particular, Sahir Ludhianvi’s poetry stood out. I remember two songs “sansaar ki har shay ka” and “aage bhi jaane na tu” which made me wonder as there seemed to be a jewel hidden in the songs.

Later I learned how physicists are grappling with the nature of time. Whether it is Arthur Eddington’s Arrow of Time, or The Wheeler-DeWitt equation of a timeless universe, or emergent space-time in quantum theory, time remains a fundamental mystery. Even the notion of the present moment and how we perceive it, is amazing as shown by Einstein in relativity of time and in relativity of simultaneity. Thanks to my daughter who keeps me updated with all this wonderful science. 🙂

In “sansaar ki har shay ka”, here is the translation of one of the stanzas:

where is this path from, where is it leading to,

nobody knows the secret of this mystery.

on the Eyelid of this moment, rests the cosmos,

till the closing of the Eye, all this is a beautiful game

and, here is from “aage bhi jaane na tu”:

you don’t know what is ahead, you don’t know what is behind you,

whatever is, is this moment alone

Sahir is expressing his awe and sense of mystery, and how our subjective experience of the present moment creates the magic of universe and brings it to life to us.

A poet’s expression and a theoretical physicist’s equation. Two sides of a great effort to understand.

I wish all the bright minds, and students in sciences, and new generations, very best to unravel this deep mystery.


A picture paints a thousand words …

A picture paints a thousand words. And words are painted across our brains.  In a paper titled “Natural speech reveals the semantic maps that tile human cerebral cortex” by Huth et al, and published in Nature in April 2016, authors have mapped words to different regions of brain using functional MRI data while subjects listened to hours of narrative stories. Interestingly, despite having our individual maps, our minds are organized in similar and consistent manner, and words cluster as per semantic domains. Here is the amazing video:

Nature video on how brain maps words to different regions

Here is a screenshot of part of brain and words mapping to that region:

Screen Shot 2016-05-01 at 11.56.09 AM.png

(Figure Credit: Nature video on brain dictionary, April 2016)

One cannot help observing connection with current research in Natural Language Processing (NLP) in the field of Artificial Intelligence. Machine learning models such as deep recurrent neural networks can work with words. But since computer models work with numbers, words have to be first converted into numeric representation in the form of vectors. Here dimensionality of these vectors can be large. In a way, the words are being converted into spatial points in a high-dimensional spaces and then semantics becomes spatial concept. Words which are semantically similar, map to close by regions, and their relative displacements capture semantic concepts. See the following reference for technical details of word2vec (word to vector) approach:

Vector representation of words

Screen Shot 2016-05-01 at 10.16.50 AM.png

(Figure Credit: Tensorflow tutorial on word2vec)

It seems we are making progress in unraveling how mind works.

At the same time, a lot has yet to be discovered and understood.

How does a new born baby develop this semantic map within a matter of few years, which seems to be consistent across individuals?

And, where are our thoughts in all this?

In a semantic world, where words become colorful entities in space, perhaps our thoughts are nothing but mysterious dances in this surreal landscape.

And, where are our dreams in all this?

One can only wonder, as a wise philosopher did long time ago:

“Once upon a time, I dreamt I was a butterfly, fluttering hither and thither, to all intents and purposes a butterfly. I was conscious only of my happiness as a butterfly, unaware that I was myself. Soon I awaked, and there I was, veritably myself again. Now I do not know whether I was then a man dreaming I was a butterfly, or whether I am now a butterfly, dreaming I am a man.” – Zhuangzi, 4th century BC

dali boat.jpg

Data Science for fighting Human Trafficking


Recently one of my dear friends has made me aware of human trafficking problem in India. Human trafficking is a global scourge with its roots in poverty, lack of opportunities, and related difficult socioeconomic conditions.

I am a data scientist and it makes me wonder if Big Data can be used to understand and to solve such social problems. How can we employ tools of modern data science for social good?

At New College of Florida, the honors liberal arts college of Florida, my colleagues and students and I were recently discussing about role of data science for socially relevant projects. This post is inspired by this discussion and my friend’s work in India.

It will have to be interdisciplinary work in which people from social sciences, computer science, economics, policy making, law and enforcement, and non-governmental organizations will have to come together to pool in their efforts.

First step will be data collection. Due to many stakeholders, the data can be collected at multiple points and in different forms. Data could be collected from hotlines, from monitoring of real-time microeconomic factors, and from census and surveys about detailed picture of local society, education, economy and employment.

More than 80% of data science is hard work involved in collecting data and cleaning it. The data will be in different formats, both structured and unstructured. Getting the data is truly a big task in itself. Having lots of real-time good quality clean data can go a long way in making the exercise successful.

The data can be then summarized and visualized. It will portray a complex picture of not only actual incidents and traffic patterns as collected by law and enforcement but of underlying economic, demographic, geographical and social factors.

One can then proceed to statistical analysis of the data to answer many questions:

  • What are the patterns in human trafficking?
  • How can human trafficking be detected?
  • What factors correlate with high incidence of human trafficking?
  • What are potential causes?
  • What law and enforcement techniques are most effective?
  • What is the profile of perpetrators? How do they operate?
  • What is demographic profile of affected areas?
  • What is the profile of victims?
  • How can people in affected areas be educated to combat the problem?
  • What are the sources and destinations of human trafficking at regional, national and international levels?
  • Which programs for victim rehabilitation are most effective?

Here role of statistics and data science is to provide technological tools. These tools come in form of distributed data collection, database software, visualization software, geographical mapping software and statistical techniques.

Backed by data science and deep understanding of ground realities, one will be able to conquer this social evil which plagues humanity. My hope is that such effort will lead to application of data science for social good in many other projects.

Deep Learning brainstorming at a Lucknow (India) school

Screen Shot 2016-01-23 at 6.01.51 AM

I recently got opportunity to give talks at two very different places in India. One is the Indian Institute of Technology (IIT), Kanpur, one of elite technical institutes in India with stellar alumni. I spoke on Computer Vision, Machine Learning and Deep Learning, highlighting how my own career has been intertwined with progress in these fields. As expected, audience was highly learned and technically strong. I have now signed up with IIT Kanpur’s Machine Learning Special Interest Group and am happy to see the group is very active and versed with the latest developments.

This post is however about my experience at a very different place. I was invited to talk at St. Anjani’s Public School in suburb of Lucknow, Uttar Pradesh, India. Students at St. Anjani’s School come from modest backgrounds in this North Indian belt where people are still struggling hard to make ends meet.

When I arrived on the campus, I was greeted by the school manager and the school principal and they took me to one of their class rooms where I spoke on Machine Learning, Deep Learning and Data Science in an interactive format assisted by a power-point presentation. It was a great experience for me as the audience consisted of 10th and 11th grade students.

I was highly impressed by the questions asked by students. How do Internet Search Engines work? What are the conditions in which Artificial Intelligence cannot be used? What steps can students take to learn more about Data Science, Deep Learning and AI? How can one succeed in entrepreneurship in the IT sector?

Students came up with ideas for applications where data science and machine learning can be used which were at par with those being considered and funded in Silicon Valley. Here were some of children’s suggestions: smart home, ensuring safety of children using robotic babysitters, applications of deep learning in health care, smart governance, remote medicine, and educational apps.

I came back with the following observations:

  • There are smart children everywhere, including the poorest areas of the world. And, all these children harbor in their hearts desire to learn about the cutting edge in sciences and technology.
  • Success depends on opportunities. Not everyone gets opportunities and resources to succeed.

Next day in Indian newspapers, I read articles about the latest trends in technology. As the technology juggernaut of data science, deep learning and AI marches forward in Silicon Valleys of the world, I paused to ask if deserving children around the world will all have opportunities and resources to participate in this effort or if they will get held back because of unfortunate circumstances not in their control. I am very much part of Silicon Valleys of the world, and it is my hope that some of the aspiring students I met will make informed decisions in future when they have to choose their college majors and careers. That also made me reflect on corporate social responsibility programs and how I can be a part of such an initiative to ensure scholarships for these deserving children who will design our future world.

Ultra Advanced Autonomous Artificial Intelligence


Computer Vision, Speech Recognition and Natural Language Processing are experiencing convergent and synergistic progress these days. Breakthroughs are being reported in mass media. Large neural networks with many layers and deep learning are active areas of research at present.

I believe this is just the beginning and coming decades and centuries will witness great progress. And that brings me to the topic of this post.

What if we can eventually build ultra advanced autonomous machines? Should we do it?

Recently Stephen Hawking warned us against the dangers of Artificial Intelligence. To not have a human in the loop could lead to unpredictable scenarios. As any computer programmer knows even a simple program has to face all kinds of situations when deployed in the real world, so if it is a highly complex and completely autonomous system, then it will be very hard to foresee all possibilities in advance.

To be on the safe side, we may choose not to build fully autonomous systems and always have human in the loop. At least if they are to be used on the Earth. This may make them less efficient but we may choose to live with this decision. But what if we need to move out into space. Let’s say we want to terraform Mars. Then it would be perfectly reasonable to first send robots there. And they will have to be autonomous with no humans around.

Oh, on a lighter note, in Hollywood, it is fashionable to suggest that such machines can even become conscious and have feelings. Unless we first understand what states of matter can cause consciousness to emerge, it really is rather pleasant and entertaining fiction at this time, which I do enjoy when I go to see such movies with my children. In this post, I am talking about highly complex state transition of such machines from rather geeky software engineering point of view. Of course, one day it will be very exciting to learn about scientific breakthroughs in understanding human consciousness which undoubtedly will have far reaching implications.

Since I am in favor of technological progress and I see great practical benefits of smart software – in medicine, space exploration, scientific research, energy, education, commerce and many other fields, I think it is really for us to become responsible in use of technology as a society and have appropriate rules in place.

Technology just gives us tools. How we use them, it is for all of us to decide together. 🙂


Previous Older Entries