Simplifying the Complex: What is Machine Learning? - Business Insight - Canon Singapore

    Simplifying the Complex: What is Machine Learning?

    According to Cisco, the internet traffic surpassed exabyte threshold in 2001, and the zettabyte threshold in 2017. Now, imagine the amount of data that is generated from transactions and communication processes through the years. The numbers of internet traffic might be huge but it doesn’t even begin to compare with the total data that companies are storing worldwide.

    With the huge amount of data that needs to be properly and securely stored, that’s where Machine Learning comes in.


    What is Machine Learning?

    Data science has revolutionised the way we perceive and make sense of stored data. It aims to analyze any existing datasets and derive meaning from it. While Machine learning, as a subset of Data Science, allows prediction based on the analysis of patterns and trends from datasets.

    Broadly speaking, there are two types of machine learning techniques: supervised or unsupervised learning. Supervised learning allows any machine learning model to be able to predict future outputs based on the known inputs and outputs, while unsupervised learning is a machine learning model that relies on the grouping and interpreting of data based on input data only.

    All in all, the ability to “learn” from an immense amount of data and make data-driven recommendations is what makes Machine Learning powerful as a tool for solving problems.


    What machine learning can or cannot do…

    Machine learning might be powerful but it does have its limitations. Broadly, we’ll share our take on Machine Learning for a better understanding of what it can do.


    #1: Machine learning can organise data BUT not create something new (yet).


    In Machine Learning, various algorithms can be applied. These are then used to sort through and organise the large amount of data. Through data mapping and prediction, it is able to automatically identify what are some of the useful data to keep, from data that is seldom or never used. All of which are possible with the adaptive nature of algorithms, allowing it to “learn” from the data through computational methods.

    These technological advances are useful and cost-effective in data management, as data can be stored in the fast or slow storage automatically based on the frequency of its usage. This saves a lot of time for data architects and storage managers from addressing storage optimisation manually, where human errors are hard to avoid.

    With the advancement in technology, the creation of new data such as AI-generated images might become a reality progressively. While that continues to develop, what Machine Learning gives will ultimately depend on the data a Machine Learning model is fed with. Any poor results are usually indicative of either a lack of data or the lack of good data, as machine learning algorithms require large amounts of data before any useful results can be generated.


    #2: Machine learning can protect data BUT not beyond its task.


    Data privacy has always been a concern, and Machine Learning has been a solution for that. Through machine learning algorithms in solutions such as ESET, computers are able to learn by experience. By analysing the data it is fed, malicious patterns can be identified. This allows the detection of anomalies through constant monitoring of any network, in real time.

    Other than threat detection, machine learning algorithms can also keep our data safe when we are browsing online, identifying any potential dangers that might be a current or emergent threat. This extends to the protection of data on cloud storage through cybersecurity solutions such as OpenText, by detecting suspicious cloud activity, logins, or location-based anomalies.

    However, as Machine Learning relies on structured data, it cannot run beyond what the machine learning algorithm, framework and parameters were pre-defined to cover. In such cases, new algorithms will need to be in place to carry out what the previous algorithms were not tasked to cover.

    #3: Machine learning can analyse data BUT it doesn’t get smart instantly.

    The automation of data analysis and the making of data-informed predictions are essentially the core functions of machine learning algorithms. It is able to analyse data, for example, footages that are recorded or collected in real time. With such algorithms in place, these video analytics solutions are able to sort and handle different types of actions and trigger the appropriate response.

    Another application of such data analysis can be seen in Canon’s Smart Workspace Solution Facial Recognition technology. Such solutions can be applied widely within office settings, based on the company data it is fed with, to improve efficiency, productivity and office space management.

    Once any structured data is analysed, the algorithms are able to uncover efficiencies, predict and provide useful insights through solutions such as UiPath to improve any digital transformation journey that businesses may be embarking on.

    However, with the large input of data, time is required for any algorithm to fully learn and be able to produce good, quality predictions or results. The more it is fed with good data, the faster it will be able to learn through the patterns and trends it detects.

    Regardless of its application, as we journey through Industry 4.0, machine learning is definitely here to stay. The wide and extensive application of machine learning algorithms will help many businesses to improve their work efficiencies and systems in the decades to come, paving the way for greater innovations in Industry 4.0 and beyond.

    Follow us on Facebook and LinkedIn for more business insights and updates on digital solutions!