Skip to main content

A Business Principal's guide to Machine Learning - The origins

As a Business Principal, I have been coming across the term Machine learning for a long time. Several business executives have wondered what this animal is and how does it help them.

 The series of posts is to list out what Machine learning means to a business principal and some basic industry standard approaches that has been tried out in the past few years.





What does a Business Principal do

Lets suppose the business executives are working through a particular business challenge. Some challenges are tactical and some are strategic. They ideate internally or with external consultants or both.

Today, this ideation is severely limited by the scope of the challenge that is discussed as well as with the knowledge that the stakeholders bring to the table.

The Business Principal is primarily a facilitator that helps this ideation achieve the outcomes that solve the business challenge. The business principal structures the conversation, layers the business challenge to make it more transparent, simulates the various business scenarios for more clarity and provides the next step of action items that would be required from that ideation exercise.

Machine Learning for Business Principals

In this context, Machine learning is now the buzzword that keeps getting throw at Business principals - more as solution in these ideation exercises. It becomes critical for a Business Principal to know about Machine learning now more than ever.

The series of posts are my attempt at abstracting my experience in the last 4+ years on interacting with several business teams, data scientists to articulate my view point.

I have drawn a lot of the points and structure I am making from the following Medium post: https://medium.com/@yaelg/product-manager-pm-step-by-step-tutorial-building-machine-learning-products-ffa7817aa8ab

The Origins of Machine Learning

It is important to know a little about what Machine learning has evolved from and then get into the details.


What is Machine Learning


Machine Learning Enables us to draw insights from the data
at scale, at a level of granularity that ranges from a single user interaction to worldwide trends and their impact on the planet


What is AI?

AI - Artificial Intelligence - is about helping computer reason. When presented with a problem, computers should be able to reason away to a solution.
It is also about:
  • Knowledge Representation: Understand and interact with the real world
  • Planning: Understand how to navigate and plan around the world that we live in - Find out where it is safe to go.
  • Natural Language processing: Speak and understand language, Understand context. Teach computers as much of subtlety in communication and language as we can.
  • Perception: See, hear and feel things in the world. Perceive things through sight, sound ,smell & touch.
  • Generalised Intelligence: Discrete parts to combine, Would get to fully autonomous, thinking, interacting robot


AI Winters

AI Winter 1: Successful language translation. However, Could not understand context
AI Winter 2: Successful interpretation of language and context. However, could not understand if it deviated from the pre-programmed context.
AI Winter 3: Expert systems. However, learning curve for each expert system did not help the next expert system

Deep Learning

Earlier, Codified rules from experts that help the computers interpret data. Later, Algorithms that interpret data and help the computer learn.
We started with Interpreting handwriting of zip codes for postal department. Next was speech to text algorithms.
Neural Networks - Computer algorithm and data structures to mimic the brain.
Instead of codifying computers, the algorithms help the computer learn for themselves from the data.
Machine Learning

Whole set of techniques where we are trying to program computers to exhibit Human intelligence. E.g. Machine learnings, Search optimisation, Expert systems.
Machine learning is one of those sets of techniques. It is a peer to search techniques, constraint satisfaction, logical reasoning, probabilistic reasoning and control theory.
Machine learning has a set of techniques within itself - Deep learning, Associative rule learning, Decision trees, Random forest, Support vector machines, bayesian networks, Reinforcement learning, Genetic algorithms.
Of these, Deep learning has been the most productive and accurate of all.
More on the next set of posts.....

Labels:

Comments

Post a Comment

Popular posts from this blog

The right business model for your product

How would you decide what is the best business model for your product? Deciding on the right business model is very important for the product. It does not necessarily mean what is the price of your product. There could be no price for your product - though that is a myth. For example, we think google or Facebook is free. However, we are never the customers for Google or Facebook. It is the advertisers that are its customers. What we need to figure out when we decide on a business model are as follows: Who is my target audience? Who is my customer What is the price. How is the support model going to work Who are my partners While we do not absolutely need to price a product, we need to understand how we would make money. For example Open ERP does not price its product. However, its product is not the ERP software but really the hosting, service and training that it offers. While pricing is an important part, what is more important is in realising who is you...
Post a Comment
Read more

Finding your target audience

The biggest task of the product manager is to find your target audience. Typically it is a very tight rope walk. You define it too narrowly and you would struggle to generalize the product later. You define it too broad and you would struggle to satisfy a lot of people and the product might not take off at all. There is this awesome post by Dave Mcclure on why Niche target segment works. Blog link here . What he says is mostly right and that is what we did as well. We did it mostly by accident. We had to define our niche because we could only talk to those kind of retailers and get data from them to start developing the data. It did have its upside. We were very clear on what we wanted to develop and we are slowly talking to retailers and adding more scenarios. As we are doing this, we are also talking to a slightly different set of retailers. This helps us incrementally build our product while also generalizing the existing features. In our case, we did not fin...
Post a Comment
Read more

Constructing the Product Roadmap

When I started out in the team, I was tasked out with coming out with a plan so that we can demonstrate the product with real data. The idea was to demonstrate it to our prospect with their own data and win them over. However, when I started working, I found out that what we lacked was not a plan but a vision. I quickly put together a one day workshop with our team and our product owner. When we started talking, we talked about several things and figured out that we needed a plan of various proportions. We needed a plan that could help us one the following: Provide a focus for the long term with near term focus on just one thing Help us talk to prospects on what we are going to do Help us decide on when we would need to focus shift on which function - Engineering , sales etc We also did a quick “Product in a box” exercise so that all of us in a team can practice and imbibe the elevator pitch. We needed this since we had to do this pitch at several levels - Inter...
Post a Comment
Read more