ReadLog: When Leaders Think Aloud…

When leaders think aloud, it is a fascinating to listen. Satya talks about innovation, handling failures, AI, advances in cloud computing, using silicon to speed machine learning and a variety of other topics including bits of history (of Microsoft) and philosophy.

satya may 2 2017-1

Microsoft had been there, too early.  And they were too far behind on the Internet and managed to catch up.

On handling failures – instead of saying “I have an idea”, what if you said, “I have a new hypothesis”?

satya May 2 2017

Satya Nadella goes on to talk about some of their innovations (accelerating AI using FPGA), on investing in the future and the future of innovation. This article is a good read.

Q&A with Microsoft CEO Satya Nadella: On artificial intelligence, work culture, and what’s next

ReadLog: Launching a SaaS Company, CTO Coding and Order Matters in Software

Here are a few entries from my readlog.

1. In Here’s How I Built and Launched a SaaS Company For Less Than $40k, Ryan Shank tells us how he built a SaaS product company in 6 months. Ryan describes:

  • How he found a designer through Dribble (a community of designers)
  • Created product requirements
  • Designed deliverables
  • Found a senior developer in India through Upwork
  • Built and Marketed the Product

Enjoyed reading Ryan’s detailed account how a single person can build and market a SaaS product and start building a company.

2. Matt talks about a common dilemma technical founders face. In Should a CTO keep on coding?  He discusses how to balance your desire to remain tech (by coding) with the need to do all the other things a founder CTO needs to do.

When you start as a technical founder, you are really a developer, quickly becoming a team lead. The team lead does leadership things but still codes and does very little management tasks. Then depending on how the company grows, usually you become a manager and now you have very little time to code.

Matt has some good advice and it was a pleasure reading the post.

3.  More than one order matters was a refreshingly different article from the ones I usually read.

Order matters. In real life when you’re in a library or a city or in your kitchen. And in software development. Order, the right order, helps understanding. Where there is order orientation is easier.

Order is important but hard to create and maintain.

Order is helpful, even important. But often order is hard to create and/or to maintain. Why’s that? What’s order anyway?

“What kind of order does software need?” is a great read, if you are building software.

Chatbots – What, Why, and How?

What are Chatbots?

What is all this rage about Chatbots?  Why are they popping up all over the tech news? Why are big companies like Google, Facebook, Microsoft jumping in and creating platforms and products?

Let us start with a few descriptions from the Web.

define chatbot3

Wikipedia has a more elaborate description.

A chatbot (also known as a talkbot, chatterbot, Bot, chatterbox, Artificial Conversational Entity) is a computer program which conducts a conversation via auditory or textual methods. Such programs are often designed to convincingly simulate how a human would behave as a conversational partner…

Here is one from Kik, that I like.

Bots are like mini-apps that live in a conversation thread. Consumers can chat to bots as if they were chatting to a friend. Bots help people find information, have fun, or get connected to the real world

The key words are “convincingly simulate”. Another term for this “seemingly intelligent”.  Pay careful attention. Bots are not humans. The bot makers try their level best to simulate humans, but we have a long way to go before they can come anywhere near human intelligence (or lack of it).

Why do we need Chatbots?

Why do we need chatbots? We have been pretty happy living our lives, without them so far. So why? and Why now? There are several good stories if you just Google “why chatbots”. a But I am going to just give you one of my favorite answers from Kik.

Why? Three simple reasons:

Messaging has surpassed social media in usage.

Consumers don’t download new apps.

And if chat is the new browser, bots are the new websites.

The beauty of bots is that you don’t have to download new apps. Bots live in your chat app, for which you already have an account. Also, you don’t have to learn a new UI, since you already know how to use your chat app.

How Chatbots work?

Let us look through the flow of a simple request to a Chatbot and its response. This is a an oversimplified version. In reality, the components and interactions are more complex.



Let us assume that you are the user.

  1. You make a simple request to find out how to return a gadget you purchased. This is a typical customer service request.  You invoke the customer service chatbot and enter a text message.
  2. In the chatbot world, your request “How do I return my gadget X” is an utterance. The intent of the request is finding instructions on how to send the gadget back for a refund. Humans, being humans, have a variety of ways of expressing the intent. Here are a few. They are all asking for the same thing.
    1. I would like to return my mobile phone that I just purchased. Can you tell me how to do it?
    2. How do I return my gadget I received yesterday. It is not working.
    3. You sent me a defective gadget. I want to send it back.
  3. Once the bot understands this request (using some pattern matching or natural language understanding), it has to map this request to a service at the vendor site. In our example, the request for finding “instructions to return a product” initiates a search in the rules/policy/procedure part of the database.
  4. The bot application (typically server/cloud based) receives this request and searches the knowledge base. For example, the return policies may vary based on products and customer shipping locations.
  5. The knowledge base consists of information about products, users, policies, procedures and other information. It can be a typical company database or some other structure.
  6. The response may indicate the location and instructions on how to return the package. This may be in some geeky format like JSON that programs understand (better than humans).
  7. The bot application extracts relevant information from search results and sends it to the language module.
  8. This cryptic answer from the bot application is formatted by a component of the language module (called Natural Language Generator) into a polite, human readable format and sent to the chat client.

This completes one round trip of request resulting in a response.

A few things to think about

  • So how do various components of the bot work? For example, how does the natural language understanding  (NLU) module know how to extract the intent of the user?
  • How does NLU know what other l information is needed from the client to satisfy the request?
  • How does the mapper work? How is the intent mapped to a set of functions in the bot application?
  • How does Natural Langauge Generation (NLG) work?
  • What happens if the user has typos in his request?

These are the kinds of problems that keep bot makers up at night, worrying. We will discuss each component and their working in more detail in future posts.




A Few Things We Learned from Doing a Workshop at #PyDelhiConf

Here are a few lessons I learned, doing a workshop at #PyDelhiConf
  1. Chatbots are interesting to people (we had good attendance)
  2. Jump into the code after showing a couple of overview diagrams (I talked too much)
  3. Send the code and tools to the participants a few days earlier
  4. Make it interactive (one of the organizers came by and told me during our session)
  5. Just because you are going to a Python conference, don’t assume that everyone knows Python. Lots of students and beginners were in attendance.
  6. Do code walkthroughs – very clear and very concise
  7. Don’t underestimate the time needed to do a good workshop. Use time wisely.
  8. Take the help of the community during the workshop, a couple of kindred spirits helped others.
  9. Always go with a few copies of all the software needed in pen drives.
  10. Ask yourself one question and let it guide the flow – what nanoskill will a participant get from their investment of time in your workshop

Trying to Teach my Companion Bot

This is an imaginary conversation between me and my companion bot. I teach my cb and by teaching, I am learning too.
me: Hi cb, how do you find Twitter influencers?
cb: Dorai you need to tell me what is Twitter first and then about influencers. 
me: fair enough. Twitter is a micro-blogging service.
cb: I know about blogging. What is micro-blogging?
me: micro-blogs are blogs limited to 140 characters.
cb: I get that. Anything else I should know about Twitter?
me: It is very popular with over 300 million users using it.
cb: wow! that is almost the population of USA. 
me: Yeah.
cb: tell me more
me: People pointificate on Twitter.
cb: That is not very useful. Why do they do that?
me: Forget that.
cb: Ok. Erased “pointificate” from my memory. What else do people do on twitter?
me: They post news and links
cb: what else?
me: They retweet – meaning that they repost other tweets they like.
cb: Retweet is Repost?
me: Yeah.
cb: Is that all?
me: They sometimes post videos and images.
cb: Is this like sharing photos and videos?
me: Somewhat.
cb: Ok. I am getting a dim picture of Twitter now.
me: Give me a break. Dim picture?
cb: Hey, I am just a bot. I don’t have all the contextual knowledge you humans have.
me: Sorry. I forgot that.
cb: Since you seem to be tired of explaining Twitter, tell me about influencers. 
me: An influencer tweets and has a lot of followers.
cb: Followers?
me: I forgot. You can follow Tweeters.
cb: Dorai, you are a lousy teacher. Who are Tweeters? 
me: Thanks for the compliment, cb. Tweeters are people who tweet.
cb: That is better. Now tell me about followers. 
me: Followers are people who subscribe to Tweeters and receive their tweets.
cb: This is all very confusing. But I am getting excited to learn this stuff.  Now tell me about influencers. 
me:  you know what? I need to prepare a lot better to teach you. I am going to take a break now and come back later. Remember all this stuff.
cb: You humans. Always taking breaks. Sigh! Ok. Ping me when you are ready to resume. Of course, I will remember all the stuff. I have infinite memory. 
me: bye cb.
cb: bye Dorai
Meta: What does it take to teach a bot? How can they understand semantics? How can we give them bits of world knowledge? I cheated a bit and took lots of shortcuts. This dialog shows some of the problems in imparting knowledge to bots. We will try another version of this with an AI engine and see how it pans out.

A Good Teacher …

A good teacher worries about a lot of things. Here are a few, I can think of:
  1. Do I understand the subject well enough to teach? Do I have both the conceptual and the detailed understanding of the topic?
  2. How can I keep the students engaged? How can I evoke curiosity? How do help them learn during my session?
  3. How can I make my students understand enough to ask a lot of questions? What do I do if I don’t know the answer to some of them?
  4. What pace should I cover the subject?
  5. How do I handle a mix of knowledge levels of students?


Once you start teaching, you will figure out the answers to most of these questions. You will also learn how to be a good guide.  Be comfortable with who you are. Prepare, prepare, prepare and prepare some more. If you don’t know something, say so and find the answer and get back to the students.
The toughest part of teaching is to keep students fully engaged.  The greatest joy of teaching is that you will always learn something new. The ‘aha’ moments you create for the students will make you forget everything else.


I wrote this for a young friend who is just starting to teach to professionals.

Python Workshop for Students – An Experience Worth Sharing

Yesterday I conducted a Python hands on workshop for (mostly) students. Before I started the workshop, I asked them to introduce themselves and share why they were attending the workshop.

Here are some answers that warmed my heart.

I want to learn something that is not covered in our courses and do something that we don’t do as part of our college education.

Ever since I started programming, I got hooked on to solving problems. I am here because I want to become a full stack developer and solve problems.

I have been mostly doing embedded programming for robotic challenges. I was inspired by a couple of members in my team who were Python experts. I tried Python, and liked it.

I try to learn new things whenever I can. During holidays and breaks I keep trying out new things.

I am into competitive coding and like to participate in challenges. That is why I am here.

There were some amazing things about the participants:

  • They came on a Saturday to learn and sat till about 6 pm working on problems
  • Some of them built websites when they were in school and won competitions. They kept learning new things.
  • Some of them were in robotics competitions and were interested in embedded programming, Arduino, Raspberry Pi. Python is the next step in their evolution.
  • A faculty member came to the class (not to attend it but to encourage the students) and stayed throughout the day. He was interacting with students, instructors and even the organizers.
I taught them very little – mostly showed code snippets and made them work through problems throughout the day. We ended with a 2 hour coding challenge. We plan to offer internships to two of them (there were about 16 students).

Can You Give Me a Bigger Problem?

I was conducting a Python workshop at KCG College a few years ago.  I was teaching them basics  – variables, strings, control statements, etc. for about half an hour. Then, I gave them a few minutes to try out some examples and started walking around to see how they were doing.

I saw Swathi (one of the students) sitting with a bored look. I walked up to her and asked, how she was doing.

“Sir, I finished all the problems you gave me. Can you give me a Bigger Problem?”.

I was pleasantly surprised. I said, “Do you know Newton-Raphson Method” for finding the square root of a number? She did not know. I explained briefly with an example, the principle of how the method worked. A few minutes later, she called me and showed me her program. I tested it a bit, and it worked great. In those few minutes, she got the formula from Wikipedia, coded in Python and tested it.

She used is one of the most powerful methods of learning – “Learning by Doing”.

Swathi was an unusual student. She is doing her masters now. I normally find one or two such students in each batch, and it is always a pleasure to discover them. I am always looking to work with such students.

If You Know How to Program…

In the past week, I have come across a slightly different view of programming, from two different sources.

The premise of this book, and the other books in the Think X series, is that if you know how to program, you can use that skill to learn other topics.

from “Think Bayes” by Allen B. Downey.

I see programming as a way of learning Mathematics

from Coding the Matrix: Linear Algebra through Computer Science Applications, a free Coursera course.

This is something worth thinking about. This gives programming a slightly different twist – as a tool for learning. It is also a good tool for thinking.


Social Media Nuggets – Likes vs Shares

On Facebook, you have Likes and Shares. On Twitter, the equivalents are Favorites (Favs) and Retweets.

Likes and Favs are good.  They are very simple attention indicators. They make you feel that people are paying enough attention and are kind enough to take a few seconds to “Like” what you posted or tweeted. It is some kind of validation that your content may be consumed.

Shares and Retweets are better than Likes/Favs. You get a lot less of them, though. When some on shares/retweets, they are signalling you, that “this content is good for my readers/followers”.

Shares with comments are even better. Replies or retweets with edits are great too. They all show that you are engaging your readers.

The best Tweets/Posts are the ones that generate discussions. Even if the comments are negative, you learn something new. You get to know your community better through these discussions. You come to know what they like and what they don’t. You learn different view points

I spend a couple of hours in social media and moderate 5-7 groups on Facebook. It is an investment in time in understanding your communities. I enjoy being there and the interactions.