10 Things Everyone Needs to Know about Machine Learning in 2019

Artificial Intelligence (AI) and Machine Learning (ML) were two of the biggest buzzwords of 2018 and received a huge amount of hype following the mass of adoption of technologies such as Amazon Alexa and Google Home which have started our lives.

In simple terms, machine learning means learning from data. Every business has data in some or form and learning from that data can provide a vast amount of benefits. However, it is important to start small to think big.

Let’s take a toddler learning to say “mummy” and “daddy.” You will spend hours pointing at “daddy” and then “mummy”, asking the toddler to follow your instructions and repeat the words. Next time he sees “daddy” the toddler says “mummy” and you correct him. Eventually, he starts identifying daddy as daddy and mummy as

In the same way, we can teach machines to do these tasks, by feeding in a series of data and letting it learn for itself what to do and how to do it correctly. In its most basic form, this is machine learning.

Machine learning doesn’t need to be daunting and realistically, any business can benefit from its applications. Here are 10 things you need to know that will guide your machine learning strategy in 2019.

1.   Machine learning is all about “Quality In, Quality Out”

To start any machine learning project, you must have quality data going in. If we start trying to tell a toddler to say “daddy” but all the information we feed in relates to a gorilla, there is no way for the toddler to learn effectively. The same applies to any dataset, the results are only as good as what we feed into it. It is vital that data is formatted, and cleansed before even attempting to embark on a machine learning strategy.

2.   Machine learning doesn’t always require specialist professionals

2018 saw the rise of the “Citizen” Data Scientist. These are people who don’t necessarily have professional are able to use the platforms required and teach themselves code like R and Python (both data programming languages) to conduct enough analysis that allows for business decision making. This means hiring resource doesn’t have to be as costly as it used to be, at least to start with as unskilled resource have the tools and to carry out beginner type machine learning tasks.

3.   Keep it simple

We often think of machine learning as the big-ticket items, mainly because they are so highly like how Netflix can recommend shows or Spotify recommends music. It is important to remember that you need to start a lot smaller than this and get the basics right before trying to compete with the big boys. This might be as simple as just working out the age demographics of your customers and sending different age communications in a semi-manual process before embarking on. To engage citizens and help them more effectively, the Singapore government deployed a Facebook Messenger with to make information and government organizations more accessible.

4.   Most of your time is spent cleansing data

A common myth about machine learning is that a lot of time is taken to create scripts and algorithms with huge test and learn processes. Actually, one reason why companies are looking more towards the Citizen Data Scientist is because they that more time is needed to cleanse the data and ensure its quality before they can even start doing any sort of manipulation or transformation with the results. 

5.   Data can’t work miracles

As we’ve said, there is a lot of hype right now around artificial intelligence and machine fool yourself into thinking you can become the next Facebook overnight. The models used by the big technology companies have taken years and years as well as billions of rows of data to get right. Make sure that any strategy you have is appropriately phased (years rather than days or months) and realistically aligned with your budget and expected income.

6.   Be careful of “bias doom loops”

When machine learning systems fail, it is rarely due to problems with the actual more so where it has created a bias or “self-fulfilling prophecy” if you will. A great example of this in 2018 was with itself to appear racist by showing images of black actors/actresses to viewers of the same origin even though they were not the stars of the film. The data feeding into the algorithm had become biased causing the output to be incorrect. Again, this goes back to ensuring all data is clean and accuracy before applying machine learning scripts.

7.   Machine learning WILL NOT destroy humanity

A lot of people get their ideas and perception of machine assured, we are not quite ready for anything like SkyNet to happen just yet and I promise you we are not in The Matrix. Whilst machines can learn from the data they are given, we are not yet at a point where they are consciously aware which is part of the reason why cars are not more mainstream now. Would a car carrying a pregnant woman have a different context to how it operates than any other car? Maybe it will in time, but these are the sort of that mean machines aren’t ready for world domination. As Asia’s Number One Online Fashion Destination, Zalora continues to put their customers at the of it all by introducing a virtual assistant to help them navigate through the web of customer and commonly asked questions.

8.   Machine learning does not mean everyone loses their job

Another misconception about machine learning is that the core driver for its use is to replace people with computers or robots. There are absolutely some instances where this is the in 90% of situations, they are designed to help people improve their productivity and take away manual parts of their job so they can focus more on the things they enjoy doing. This could be automating data entry tasks or improving existing systems by data better. Check out Pam, Parkway's Assistance Manager, is a friendly virtual assistant built to help patients, families and friends answer all their health-related queries.

9.   Machine learning enhances segmentation, recommendations, lifetime value and customer experience

Arguably the most common application of machine learning across all industries is within marketing and customer experience. Machine learning can be used to work out how customers respond to different types of email or how they like a website to look. Maybe different customers expect different levels of customer service, some like a phone some would prefer to never hear from you after the point of purchase. Using your data, machine learning can ensure you have in place to speak to every individual customer in their preferred way, enhancing their lifetime value and ROI. Check out Singapore’s First Truly Digital Life Insurance Provider deployed a Messenger to complement its digital strategy.

10. To begin with, machine learning will seem overbearing and too hard

When just starting out with machine learning, you will need to do a whole lot of research. This could be working out the best applications for your business and all the options or even working out how to do statistics and basic algebra. Sometimes, it might feel like going back to doing all of that research is 100% necessary to avoid some of the common faux pas mentioned within this article and ensure you are equipped to succeed.

Machine learning comes with several benefits that everyone should be aware of in 2019 as the adoption of its techniques continues to grow. It is important that senior business potential and develop the appropriate strategies

Note: If you wish to explore how virtual agents can supercharge your productivity, check out local companies like now that you know that Machine Learning is not a friend who can help you to create an unfair advantage within your competition.

About the Author:

Andrew Chow is known to be pragmatic, visionary, competitive, intuitive and giving

While he is a successful social media and public relations strategist, entrepreneur and speaker based in Singapore, he is also the best-selling author of a highly popular series of books: Social Media 247, Public Relations 247 and Personal Branding 247

Andrew has spoken in over 15 countries within 5 years and addressed more than 20,000 people Digital Marketing, Personal Branding, Enneagram, Public Relations and Branding

Andrew’s career of 30 years; has seen him work with an array of clients including AXA Insurance, Abbot Medical Optics, Singtel and Sony Pictures, M1, Starhub, and Sennheiser