20 April 2018

3 Myths about Machine Learning

Posted in Industry

A recent ServiceNow and Oxford University study found that the rate of enterprise investment in machine learning is due to double over the next three years. Business leaders clearly know that machine learning is an essential part of their organisations future success, however there are so many myths around machine learning that it often becomes a daunting task investing in it. Below are three of the most commonly accepted myths that surround machine learning and why they aren’t true.

1. Machine learning will displace your workforce

We've been writing a lot about this subject recently and it's because it's one of the contributors to peoples fear of machine learning. It's largely a media driven myth inspired by the idea that machines are being designed to supersede human intelligence rather than augment it. The reality is that people are integral to how machine learning understands and processes information and will make people more productive. Human in the loop is part of the 2.3 million new jobs Gartner predicts machine learning will help to create. Others could include IT roles (infrastructure, development and implementation), but it will enable more more customer facing and experience based roles which are reliant on uniquely human characteristics.

2. You need a data scientist to teach Machine Learning

Machine Learning may seem like a “dark-art” and something that can only be programmed and implemented by teams of data scientists. This is certainly how some of the larger vendors position their solutions but this is not the case for all applications. In many scenarios, machine learning relies on a vast volumes of learnable data to learn.Organisations may have vast quantities of data but it's not always in good shape and therefore it's not learnable data. It's therefore easier if machine learning can learn from people who handle data they touch every day, on a day to day basis. This data is current and the software is able to continuously improve with every transaction. Crucially, it is the people who are already involved in the process that are able to teach the machine learning by doing what they do every day, no special skills required.


3. Machine Learning can be applied to any task within an organisation

With new technology comes a wave of hype and misconceptions. It's easy to assume that machine learning can be shoe-horned in to solve an unsuitable problem because it's the latest trend. It's no wonder that initiatives fail which understandably leads to mistrust. With 89% of CIO’s surveyed by ServiceNow and Oxford University stating that they are planning on implementing machine learning, it's important that organisations first understand the problem they are trying to solve before they assume that machine learning is the answer.

Machine Learning can be applied to solve many different challenges within an organisation. Typically its used to deal with the plethora of unstructured and semi-structured data that is received on daily basis from customers, suppliers and employees through different channels. Some of the data that machine learning handles is machine generated semi structured but it can also handle human generated unstructured data. Before machine learning, the handling of this unstructured and semi-structured data was manually intensive and included all manner of work-streams such as enquiries, claims, complaints aswell as semi-structured data such as invoices.

If applied to the right challenge, the business benefits of machine learning are significant and rapid with improvements in productivity, efficiency, customer service.

Ultimately, it is important for organisations to understand the problem they are trying to solve and then do their research, refer to similar case studies and look beyond the myths to understand which technology solution is best for them – be that artificially intelligent or not.