Machine learning(ML) & Artificial Intelligence(AI) are interesting fields creating a lot of buzz these days. If you manage a digital product or a service built around a digital product, you must have already started thinking about how to make use of these technologies to improve your product.
Andrew NG, former chief scientist at Baidu & co-founder Coursera proposed this rule of thumb to describe the current capabilities of ML/AI
“If a typical person can do a mental task with less than one second of thought, we can probably automate it using AI either now or in the near future.”
What this essentially means is, a lot of tasks currently done by humans could be automated or its efficiency improved manyfold by AI. If you are a into a service business built around a digital product, there will be many important tasks handled by your operations teams. You can analyse if they can be automated or improved.
- Validating listing content quality for a classified website
- Validating quality/appropriateness of user uploaded images
- Identifying fraudulent listings for a classified portal
Here, we might be tempted to ask:
“Can machines do things that humans cannot do in the domain of ML/AI?”
Here, we need to make a distinction between machine learning and other algorithms. Machine learning, unlike static algorithms, learn from and make predictions on data without the need of explicitly programmed rules.
Even if we apply the above distinction, the answer is yes, because of the sheer scale of certain problems. For example ‘anomaly detection’ which is an area involved with the identification of items, events or observations which do not conform to an expected pattern can be used for ‘fraud detection’ in online transactions. This task for a high scale online application is something humans cannot do. Another example is ‘network intrusion detection’.
If you are thinking of improving a digital product, there are a lot of areas impacted by ML. For a product manager, it is always better to know all the current applications of machine learning so that they can think laterally and come up with ideas suitable for their products. Most of the underlying technology problems have been solved and are available as APIs at a cost to application developers.
A list of product modules/features impacted by machine learning
|Search||Helping users find what they want. Currently there are ML approaches which improve relevance of search results and can learn from user behaviour|
|Recommendations||Giving relevant suggestions for users based on their current and past behaviour|
|Personalisation||Give a personalised product experience to a user based on her current and past behaviour (Example feed)|
|Handwritten user input||Character/handwriting recognition. Used in signature input and recognition; User interface for non standard languages by drawing etc.|
|Conversational user experience (chat based)||Letting the user interact with the product in natural language|
|Speech based user interface||Letting the user interact with the product using speech|
|Virtual Assistant/BOT||A digital assistant that can interact with user using natural language, understand and remember user context and execute a list of tasks|
|User Segmentation||Clustering users into multiple groups based on their behaviour and other attributes|
|Spam Detection||Filtering out unwanted content|
|Face/Object Detection in images||Identifying objects in images. Ex. face tagging in photoes|
|Fraud Detection||Identifying fraudulent transaction|
|Autocomplete/autocorrect||Completing/correcting user input|
|Association rule mining*||Identifying items that frequently go together. For example – Learn that, if butter and bread are purchased, customers also buy milk along with them. Use this learning to suggest milk if the same pattern occurs with another customer|
|Translation||Ability to translate from one language to another|
|Prediction/Forecasting||Predicting future trend based on past values|
|Failure Forecasting||Identifying machine/system failure based on monitored signal|
*Some of these might not fall under machine learning in the strictest definition
#This list is not exhaustive.
Once we understand the capabilities of machine learning algorithms, we can then try to find different applications in our business by thinking laterally and applying local context
As an example, Google Translate has a feature which allows user to draw letters in any language as the input for translation. The app would identify it and convert it to the relevant character.
Later Google released Autodraw which has the same concept as above but instead of characters it identifies pictures from drawings.
ML /AI algorithms are only going to get better from here and will be an integral part of all future software products. Product managers need to get ML ready now!
Sources and Other readings:
- Machine Learning for product managers by Neal Lathia
- Difference between data science and machine learning
- Computer vision blog
- Machine learning for product managers by Ken Norton
- “How Google is Remaking Itself as a Machine Learning First Company” by Steven Levy
- “A Visual Introduction to Machine Learning” by R2D3
- “The AI Revolution: The Road to Superintelligence” by Wait But Why
- “Machine Learning is Fun!” by Adam Geitgey
- “Machine Learning: An In-Depth Non-Technical Guide” by Alex Castrounis
- “What Every Manager Should Know About Machine Learning” by Mike Yeomans
- “AI, Deep Learning, and Machine Learning: A Primer” by Frank Chen
- “Introduction to Machine Learning” course by Andrew Ng (highly recommended)
- “Machine Learning: Google’s Vision” video from Google I/O 2016
- “Hello World – Machine Learning Recipes” video series from Google
- “Machine Learning Mastery” by Jason Brownlee
- TensorFlow Tutorials from Google
- Visualizing Neural Networks by Daniel Smilkov and Shan Carter
- The Master Algorithm by Pedro Domingos
- Introduction to Machine Learning by Nils Nilsson
- Deep Learning by Ian Goodfellow, Yoshua Bengio, and Aaron Courville
- Data Science for Business by Foster Provost and Tom Fawcett
- Neural Networks and Deep Learning by Michael Nielsen
Also published on Medium.