CTO Guide To Machine Learning

At Google’s I/O conference held recently, CEO Sundar Pichai made one of the most important strategy announcements – the transition from mobile-first to AI-first and machine learning. Google’s goal is to equip the company’s line up of digital assistant products and services to understand user needs and process sensory inputs in unprecedented ways and scale.

Google’s machine learning and AI capabilities have innovated constantly and now impact everything from the cloud to email to search and the Android OS. But it has a bigger agenda – to enable individuals and small companies around the world to be a part of the AI ecosystem.

Google’s high level plan for machine learning and AI is to make it available, more accessible and useful and somewhat practical applications in the everyday life. It is now clear that other technology companies are in sync with this paradigm shift – less wow and more function.

All this talk of machine learning, AI and its impact on business, customer experience and such has revved up interest and curiosity in the subject. Because machine learning certainly has wide-reaching implications and benefits for everyone. But first things first. What exactly is machine learning? Machine learning is the science of data analysis that automates analytical model building. It uses algorithms that iteratively learn from data and more importantly machine learning allows computers to glean hidden insights without being explicitly told where to look. In simpler terms, we want machines that can learn from experience and derive insights and conclusions.

Machine learning is not new. Some of its earliest known use is the famous Turing Test in 1950 by Alan Turing to determine if a computer has real intelligence. In the first experiments, researchers interested in AI wanted to see if computers could learn from data. Machine learning has come a long way since then. It now uses neural networks – a computer system modeled on the human brain and nervous system. While the algorithm was first introduced in the 70s, it didn’t find recognition until 1986 – in a whitepaper authored by David Rumelhart, Geoffrey Hinton and Ronald Williams. It was this paper that threw light on how computers can learn and neural nets could be used to solve previously impenetrable problems.

The interesting thing about iterative learning is that the more machines and models are exposed to new data, the better they can adapt independently. They learn from previous calculations to deliver reliable and repeatable results and decisions.

Various machine learning algorithms have evolved and come in to play. But one of the most recent and interesting developments is the ability to apply complex mathematical calculations to big data – with increasing speed and accuracy. Some popular examples of machine learning at work include:

• Self-driving Google car
• Predictive online recommendations on eCommerce and entertainment sites like Amazon and Netflix
• Sentiment analysis of customer feedback on Twitter. This is a very interesting development that combines machine learning with linguistic rule creation.
• Fraud detection, Spam filtering, plagiarism checkers are some other well-known applications of machine learning in everyday life.

Who is using it?
Any industry working with large amounts of data benefits from machine learning. It is used by financial institutions to identify important insights in data, and prevent fraud. Government agencies like public safety and utilities are using machine learning in innovative ways to mine insights from vast pools of data, or to detect fraud and minimize identity theft. In the healthcare sector, machine learning is used in many different ways but the most interesting use is precision medicine – an emerging approach for disease treatment and prevention that factors individual variability in genes, environment, and lifestyle for each person.

The O&G sector is using it to find new energy sources, analyze minerals in the ground, and streamline oil distribution for efficiency and cost-benefits. In fact, every industry has begun exploring its potential, be it advertising technology (AdTheorent, Dstillery, Tapad), agriculture (BlueRiver, Tule, TerrAvion), retail (InVenture, Earnest Machine, Lenddo), legal (Everlaw, Ravel Law, Seal Software) and so on.

It is of course the default tool for data science across industries, allowing organizations to find new insights and patterns.

How to get started?
C-level executives are obviously interested in making the most of this opportunity. The starting point for implementing machine learning should be the strategy – so that machine learning is the means to an end. While the people charged with creating the strategic vision may be data scientists – who will define the problem and the desired outcome of the strategy – this will require CTO involvement and input at every stage. Organizations will also benefit from the role of chief data officer who will help identify gaps in the data, break down silos and democratize access to information. This also means greater interoperability of data among the various applications used by companies and more modernized systems that allow access to this data. Application modernization and integration are necessary for extracting the information.

Senior leadership and top management too has a role to influence and encourage behavioral change. C-level officers can approach applied machine learning in various phases – description, prediction and prescription. The description stage is where companies collect data in databases and most companies are well-versed with the OLAP – online analytical processing aspect of this phase. It is the prediction and prescription phase that require more effort and understanding from organizations. Advanced technology allows businesses not only to look at historical data but also to predict behavior or outcomes in the future – for example, warning credit risk officers with banks to determine which customers are likely to default etc. The accuracy of prediction depends on the quality of data and choice of correct algorithm.

Prescription is the most advanced stage of machine learning and is the opportunity of the future – so this aspect requires strong attention from the C-suite. The prescription stage of machine learning has steered a new age of man-machine collaboration. The C-suite here must be closely involved in the crafting and formulation of the objectives that these algorithms aim to optimize.

Summary
Machine learning is no longer the preserve of digital-first companies like Amazon, Google and Netflix. Algorithmic learning means every company will have to become a data-house and algorithms will be an indispensable part of tomorrow’s management vocabulary. To realize its full potential however CTOs will need to pair the best algorithms with the correct tools and processes.