March 22, 2021

Two Textbooks from Praxis Press

In two earlier posts, I had introduced the concept of DIGITALICs as a new discipline of study in management schools and had also explained it further in one of the famous Praxis Lockdown Lectures. Now we at Praxis Business School have taken the next step of actually introducing these concepts into the curriculum of the 2 year AICTE approved Post Graduate Diploma in Management (PGDM) Program.

A key challenge is the availability of textbooks that can introduce the correct technology to management students who do not come from computer science and other programming backgrounds. To overcome this Praxis Business School, under the Praxis Press Program has come out with these two textbooks.

Python for Business Managers - is a textbook for the Foundations of Data Science course that is compulsory for all first year management students. This is a small, 64 page book that not only teaches the student how to analyse data with Python but also sets him up for a deep dive into data science. In India, the book is available through Pothi and globally on the Amazon platform.

AI and Deep Learning for Business Managers  - is a textbook for the AI for Managers course that is offered as second year elective to the management students. This 100 page book not only explains the basics of neural network based AI and Deep Learning systems but, building upon the knowledge gained in the Python textbook, shows students how to actually build small AI systems from scratch. In India, this book is available through Pothi Global availability will be arranged through Amazon soon.

These two textbooks were written in the shadow of the infamous Covid19 lockdown and remain as a testimony to the trying times that the world had to go through.

March 14, 2021

Non Fungible Tokens in Journalism

One of the enduring tragedies of journalism today is its inability to monetize the significant intellectual effort that goes into creating great articles. With so much content being available 'free' readers have become increasingly reluctant to pay for articles that they read. Print magazines that have moved their contents to the digital format have tried protecting their IP behind paywalls and subscription mechanisms. Many users find this too cumbersome and opt either to read the few free articles or to simply go somewhere else. Which is a tragedy.

What we would want is a micropayment mechanism that allows individuals to pay for articles that they actually want to read without having to bother about annual subscriptions tied to credit cards and other relatively complicated payment mechanisms. In an earlier post that was also published in Swarajya in 2016, I had explored how browser based crypto-wallets could be used to make micro payments and I am delighted to note and report that MetaMask - the Ethereum wallet that can be added-on to both Chrome or Firefox - is now being widely used by the crypto-community. MetaMask is used both to establish identity and to make small payments. With a little bit of effort, websites can acquire the ability to make pages of the site available against small payments from MetaMask wallet.

But technology is moving even faster and now with non-fungible tokens we have a really new way to manage this process of protecting and monetising intellectual property.

image credit

What is a non-fungible token? Opensea is currently the best known marketplace for NFTs and their tutorials (1) The Beginners Guide ... and (b) The NFT Bible .... are good introductions to the topic. The fundamental difference between a traditional crypto-token ( like Bitcoin, Litecoin, or Ethereum) and an NFT is the concept of fungibility. A rupee coin ( or any fiat currency) can be exchanged for another without any loss of value. If you lend me a rupee and you take it back from me after a year, you will never ask me for that specific rupee coin that you had given to me. Any other rupee coin will do. That is fungibility. On the other hand, if you had leased a flat to me to live in and after a year you want me to give it back to you, you will demand that same flat. I cannot say that I am giving you back another flat in another part of the city of similar value. It never works that way because a flat - or anything physical - is a non-fungible asset.

What is true in the world of fiat currency (like rupee or dollar) and physical assets is also true in the world of crypto assets. Bitcoins or any other crypto-assets hold significant value ( as do fiat currency) but they are fungible. Cryptocurrency transactions are approved by the network participants ( also called 'miners') on the basis of total historical inflows being more than total historical outflows from a wallet resulting in a current wallet balance being more than the current outbound transaction. There is no need to track the specific coin or token or any part thereof.

Non-fungible assets or tokens are different because one has to keep track of the ownership of the entire token. Non-fungible tokens exploded into the crypto landscape through a trivial game called Cryptokitties that allowed people own and trade images of kittens all of which were different from each other. Almost all non-fungible tokens are based on smart-contracts built with the Solidity language and deployed on the Ethereum network. For more information on how these things are done, please see my recent post on how to build smartcontracts. Stripped of all jargon, what this means is that the contract defines a digital artifact -- an image, an audio or video file -- and an owner, whose identity changes when there is a transfer of ownership. All transfer of ownership is recorded in an immutable Ethereum blockchain and so there is never any doubt about the ownership of an asset.

But there are two problems :

  • What if I make a copy of the digital artifact? Frankly there is nothing that can stop you. You can go to the Louvre and take a picture of the Mona Lisa but that does not mean that own the Mona Lisa. Ownership does not change. You have a copy that is obviously a copy. In the case of digital assets, it is impossible to distinguish the original from the copy so does it really make much sense to go to such extent to protect a digital asset? 
  • What if the asset is something physical like a car or a house? Does it make a little more sense? It surely does! If the asset is physical, the ownership is clearly established but whether you can enjoy the asset would depend on other factors like being able to evict squatters or illegal occupants with the help of the judiciary or the police.

These are important issues but resolving them needs a little more understanding of the process of creating a non-fungible token.

During the process of creating an NFT, we usually provide only two things : 

  • The asset, which is a digital file that is stored either on the blockchain ( which is very very expensive) or otherwise on some server and only the location is stored in the blockchain. 
  • The wallet that owns the asset as defined by its address or the public key. Anyone who has access to the private key of the wallet is technically -- de facto or de jure --  the owner of the asset.

Obviously there is no fun if the asset can only be seen by the wallet owner? What is the point if I own a million dollar painting that no one can see. So all owners will make their assets visible. But then people can copy it? Which makes ownership pointless! 

So where is the catch? Why are non-fungible assets suddenly so valuable? Not because of stupid Cryptokitties!

The salvation lies in one specific feature of non-fungible tokens called Unlockable Content. This is a piece of information that can be viewed only with the private key of the wallet that owns the asset and can be used to store truly private information. This is where one can keep stuff like access keys or passwords with which one can actually use the asset, rather than just see it.

  • If the asset is password protected PDF file, then the PDF can be publicly visible and downloadable but the password will be an Unlockable Content that is only available within the wallet.
  • If the asset is house or a flat, that is visible from the road, then access key of an electronic lock to the house can be kept as an Unlockable Content within the wallet. Without this one cannot enter the house.

Armed with the concept of Unlockable Content, one can design a simple mechanism to protect and distribute journalistic content using a general purpose eBook reader that has an embedded crypto wallet to handle non-fungible assets. This is very similar to the MetaMask add-on that can handle fungible assets like Ether. We will refer to this wallet enabled eBook reader as an eReader/Wallet.

On the Ethereum blockchain there are two standards : ERC-20 for fungible tokens and ERC-721 for non-fungible tokens and both are in the public domain. So any website that publishes content can set up the mechanism to interact with any eReader/Wallet. What we need is a way for the website to request fungible tokens from the eReader/wallet and when this is approved by the human user, the website will receive the ERC-20 fungible token and send the ERC-721 non-fungible asset to the  eReader/Wallet. Obviously, the owner of the eReader/Wallet must have loaded up the wallet with ERC-20 fungible tokens like Ether or equivalent prior to any transaction. What is interesting is that the same eReader/Wallet can interact with any website that is built as per ERC20/721 standards. So there is no need to subscribe to any specific website separately.

The non-fungible asset (say a PDF file) may or may not move to the users machine. It may still reside on a distant server but the unlockable content (say, the password) will be on the eReader/wallet and using this the user can open the asset and view it within the eReader/Wallet.

There are two very interesting aspects to this mechanism

  • The user who has received the PDF file cannot send it to someone else. They will not be able to read the file without the access key stored as the Unlockable Content in the owners wallet.
  • The user, if they so want, can transfer the asset to someone else through a standard ERC-721 wallet transaction but then they lose access! This is no different from buying a paper magazine at news stand, reading it and then passing it on a friend.
What if the owner of a journalistic asset makes the Unlockable Content public? Theoretically that is possible. After all, if the owner of a house decides to hang the key on the gate or the password of a PDF file is used as an extension of the name of the PDF file there is little that technology can do. However, it is entirely possible to make this quite difficult if the eReader/wallet is so designed that it can ONLY use the unlockable content to open the asset within itself but will NOT allow it to be exported or published. While reverse engineering cannot be stopped and 'jail-broken' eReader/wallets cannot be ruled out, a process like this will ensure that a vast majority of assets will be protected successfully.

Publishing journalistic content as a non-fungible asset on the Ethereum blockchain and creating an eReader/Wallet could be a very good way to solve the problem that is currently facing journalists. The Brave browser, with its built in support for cryptowallets could be a step in this direction.

March 02, 2021

Deep Learning for Business Managers

Why this book?

 After years of being talked about and featured in science fiction, artificial intelligence is finally showing up in the world around us. But before we delve into artificial intelligence, do we understand what normal, natural human intelligence is?

For sure, there is no simple, single definition of intelligence, but it can be broken into a number of tasks, or categories of tasks. At one end of the spectrum we have raw, computational power that allows us to solve complex but logical problems, while at the other end we have abstract, philosophical concepts like being conscious and  self-awareness and have motivation and morality. In between are four kinds of tasks that are seen as intelligent - the ability to predict future outcomes, to react successfully to unexpected circumstances, to resolve between ambiguous situations and to create something new and original.

Complex logical problems can be resolved with traditional computer programming while the philosophical issues are best kept aside for academicians and spiritualists. The other four namely predictions, uncertainty, ambiguity and creativity - is what practical artificial intelligence is all about.  From this perspective certain tools and techniques have been found to be very useful. Deep learning -- the combination of a tool called artificial neural networks and a technique called machine learning -- is one approach that drives many of the most exciting innovations in the business world.

Managers in mainstream companies are cut-off from this surge of technology that is flowing past them but it will have a significant impact on their careers. Many tasks that were being carried out by humans are now done by machines. This was already evident in factories run by robots and now back-offices are adopting robotic process automation tools to eliminate white-collars. There is no escape from the relentless march of this technology.

Sun Tzu’s Art of War tells us to know the enemy. Even though AI is not an enemy in that sense, it is important that managers today understand the nuts and bolts of this critical technology. This book will help business managers enter this field, find their way around and prepare them to leverage the potential opportunities that lie within this magical world.

It is naive to believe that AI can be understood through text and slides and managers can leave the coding part to professional programmers. Zukerberg and Musk built global businesses but they did not outsource the coding to India. Computer coding is as much a part of the corporate landscape today as Mathematics and English is and every manager should have the ability to at least read, follow and understand computer code.

This book uses Python to demonstrate how intelligence arises, or manifests itself, from pure data through the medium of neural networks that are loosely modelled on human, biological, brains. If the reader invests a little effort in simply following along and understanding what the Python codes are doing they will be rewarded with five interesting applications of artificial intelligence :

  • Predict the incidence of diabetes and cancer on the basis of pathological data
  • Classify images based on what they contain.
  • Predict what a person is likely to type next on a keyboard
  • Get a Taxi to learn how to navigate around a city and Moon Lander to learn how to land safely on the moon
  • Create original art-work

There are many books that talk about and describe deep learning but to really appreciate the magic that they bring to the table,  it is important to actually build simple systems that demonstrate intelligent behaviour. No coding is required but readers with even a faint memory of computer coding should have little difficulty in navigating through this book.  

This book was written for second year MBA students in Praxis Business School where Python is taught as a compulsory subject in the first year. The syllabus and pedagogy at Praxis has a strong bias towards high technology and the author is grateful to Dr Subhasis Dasgupta, a member of the faculty who teaches these subjects in the Data Science program, for his help with some of the examples used in this book.