5 Lessons We All Need To Learn From Machine Learning

Published on:

Machine Learning is one of the tech breakthroughs that are set to power the future. There are five valuable lessons that we can learn from the science of Machine Learning.

While not in most people’s top 10 (or even top 100) new technologies lists, Machine Learning is one of the tech breakthroughs that are set to power the future. IFI Claims Patent Services ranks the technology third among the 8 fastest growing technologies of 2017.

Basically, Machine Learning is a field within the broad science of Artificial Intelligence (AI) which almost everyone has heard about, I’m sure. However, the two do not have much to do with each other outside of the shared concept of “machines doing things” as discussed below.

Is Machine Learning The Same As Artificial Intelligence?

On its part, AI involves programming machines to think and perform tasks just like regular humans would. Some AI applications include; language translation, data mining and automated trading. On the other hand, Machine Learning is based on the idea of feeding computers with data and giving them the ability to ‘learn’ how to perform tasks and progressively improve performance based on data alone, as opposed to manual programming. It is through Machine Learning that innovations like speech recognition and Self-Driving cars are now a reality.

That being said, there are 5 valuable lessons that we can all learn from the science of Machine Learning as follows:

#1 Data Quality Over Quantity

With Big Data now a fad in the techie scene, most developers are mistakenly led to believe that using large datasets in their Machine Learning models makes them look cool and sophisticated. Well, that may sometimes work, depending on the specs of the project but in most cases, small datasets work just as good if not better. Matter of fact, Big Data is expensive to collect, store, analyze and manage not to mention other overhead costs involved and the fact that the current version of Hadoop is often accused of crashing or slowing down when dealing with large datasets.

Back to the point, if you’re working on an algorithm, do some pretests to see whether smaller datasets can train your model satisfactorily. If they can, use them in the project instead of trying to show off. The same should also apply when piloting scaled experiments as above the apparent advantages; small data saves your precious time. It’s very disappointing to spend a day or two training your algorithm using big data only to discover a bug on the last day.

If your project specifically needs Big Data, you can take some representative samples instead of whole sets. Considering the defects and anomalies that are present in every sort of data, sampling does give more accurate outcomes not only in ML but also in other fields.

#2 Understanding Over Memorization

The foundation of Machine Learning, as the name suggests, is feeding computers with enough amounts of relevant data filled with examples and solutions so they can learn to solve similar but not necessarily identical problems in future. Unlike conventional programming where computers are set to do specified tasks during specified conditions, ML allows machines to react appropriately to all types of situations similar to the teacher data or not. By crawling through datasets and analyzing given variables, machine algorithms pick up knowledge on the tasks they are expected to perform and learn how to approach different situations.

The takeaway from this is to always focus on understanding concepts on a deeper level instead of merely memorizing the problems and solutions. Whether in academics, work or life in general, knowing how to solve different issues just off your experiences and life lessons is an essential survival skill.

#3 Data Is King

Successful applications of Machine Learning, particularly in the field of marketing and advertising, show that statistical analysis is the best way to predict and prepare for, future occurrences. For instance, YouTube bots collect and analyze user data regarding their movements and search words while using the website and recommend similar videos under the ‘Recommended for You’ tag, which also helps in targeted advertising. Google also uses our browsing and personal data to determine which ads to place in our paths.

As opposed to human emotions and using probabilities, data analysis often gives much better results, provided the data is accurate and up-to-date. This explains why Machine Learning programs can predict our actions and needs better than our fellow human beings, particularly sales teams.

#4 Strategy Is Critical To Success

Succeeding in Machine Learning experiments requires you to have a coherent, well thought out plan with everything else coming second. Before you begin work, you must be aware of, among other things, the algorithms to use relative to the type of data available to you. Secondly, determine what tests will you will carry out online and which ones are better done offline depending on your resources and model specifications. Not having a concrete strategy from which to base your work on makes you vulnerable to costly mistakes such as using too much data with algorithms of low capacity. This applies even in the life where you must sit and devise plans on how to deal with various issues in your life, paying attention to the pros and cons of every choice instead of making rash decisions.

#5 The Human Brain Is A Masterpiece!

If you’ve ever doubted or been critical of your brain’s power to analyze situations and react to various circumstances, you might be tempted to change your mind (no pun intended) when you study Machine Learning. For one, Machine Learning attempts to copy the functions of the highly efficient neural networks of the human brain, sometimes succeeding but mostly unsuccessfully.

Just like our reactions and abilities to perform tasks are determined by the ‘data’ or knowledge we’ve acquired in our lifetimes, ML algorithms are also built to do the same which explains why they can drive vehicles just as well as human drivers. And that’s where the similarities end.

We are still in the early stages of the Machine Learning age, but it’s an open fact, based off reported failures of ML applications, such as fatal accidents involving Self-Driving vehicles, that AI or ML can never replace or duplicate human intelligence. The neurons in our brains are self-organizing and very unlikely to slow down while ML algorithms require a human hand and electrical power to function and are highly susceptible to making mistakes or crashing. This means that no matter the advancements in this field, humans can never be replaced. The lesson here: don’t worry about bots and algorithms taking all the jobs!


It’s not often that we take lessons from technologies as they are a product of human knowledge in the first place. Nonetheless, there’s a lot to learn from Machine Learning and how computers absorb data and apply what they learn to solve problems and perform tasks, just like us.



Sharing is caring!