Of all the technologies that form part of the Fourth Industrial Revolution (4IR), machine learning is probably one of the most misunderstood. It is often confused with Artificial Intelligence (AI) but there is a distinct difference.
AI is the ability of machines to perform tasks and functions that previously required human intelligence whereas machine learning is a function of AI. Machine learning is the ability of machines to develop and learn from experience without being explicitly programmed, constantly improving their functionality by processing algorithms from each process undertaken.
The main variance of machine learning compared to AI is that just like statistical models, it aims to understand the structure of the data, according to mathematical theories and assumptions. While many machine learning algorithms have been around for a long time, the ability to automatically apply complex mathematical calculations to big data over and over, faster and faster is a recent development.
Machine learning has huge implications for business, industry and governments worldwide. Being able to continuously learn means it is possible to compose and automatically produce learning models that can analyse complex data and deliver fast, accurate results on a large scale. This allows the potential for greater use of resources, better outputs and higher productivity, it can also mean avoiding unknown risks.
Let’s take a look at a few applications of machine learning already in use.
Virtual Personal Assistants
Virtual assistants, as demonstrated through an app or function on a smartphone or a smart speaker like Google Home or Amazon Echo, are becoming increasingly popular. They are designed to understand your language, your preferences and habits, meaning that as they constantly gather data, they are able to become smarter and potentially, more helpful.
While still very much in the novelty stage, predominantly being used to play music, receive a weather forecast or ask questions, the future potential of virtual assistant technology is significant. They have the potential to become the command centre of a connected home, controlling internet-connected lights, thermostats and basic home appliances. This is the content of the Internet of Things (IoT) – networking previously inanimate objects to work from a central control system. As a result, we are seeing a stream of significant partnerships between organisations that facilitate this networking potential and drive integration.
For example, Hisense will put the Amazon assistant into its TVs, while Kohler is devising a bathroom mirror with built-in microphones so people can use Alexa to dim the lights and fill a bathtub. HP, Asus and Acer have confirmed they will integrate Alexa into their computers, while Panasonic, Garmin and other electronics makers will do the same for devices that go into cars.
One of the biggest threats to the success of 4IR is safeguarding the security of accumulated data. A breach can bring industries and even governments to their knees and therefore the risk cyber-threats pose to economies and organisations alike are immense.
Machine learning has the potential to significantly improve the identification of threats and protection of data. As Jack Gold, president and principal analyst at J. Gold Associates says; “Most of the major companies in security have moved from a purely “signature-based” system to detect malware, to a machine learning system that tries to interpret actions and events and learns from a variety of sources what is safe and what is not.”
Whilst still in its infancy, using machine learning to bolster cyber-security is the future.
Machine learning can help farmers with a multitude of decisions regarding crops, weather, timing, soil, threat identification and myriad other crucial factors. The technology allows them to make more accurate decisions in a faster space of time. It has the ability to lower costs, reduce risk and increase yields.
For example, plant breeders are constantly searching for a specific trait to increase efficiency water and nutrient use, adapt to climate change and resist disease. Machine learning can help the producer to identify a beneficial trait in a sequence of genes by using deep learning algorithms to sift through a raft of data. In doing so, the millions of combinations can be easily narrowed down to the few winning combinations.
Machine learning can also help eliminate waste with early identification of disease through pattern recognition.
McKinsey estimates that big data and machine learning in pharma and medicine could generate a value of up to $100 billion annually, through better decision-making, optimised innovation, improved efficiency of research/clinical trials, and new tool creation for physicians, consumers, insurers, and regulators.
Machine learning is also able to recognise disease or risk factors and determine disease progression, often with greater accuracy than a seasoned doctor with many years’ experience. Statistical analysis of test results and big data collected from patients globally, enable pattern detection to aid diagnosis and treatment plans based on the most statistically successful therapies.
Personalised medicine, is thought to be the next big leap in medicine as we now understand that each person reacts slightly differently to therapies depending on their genetic make-up and health. Supervised learning, a subset of machine learning allows physicians to select from limited diagnoses based on symptoms and genetic information and allows suggestion of a treatment plan also based on data analysis compared to genetic data.
The Future of Machine Learning
The future of machine learning is set to include improved unsupervised algorithms that are able to identify trends without any guidance or influence in order to find more accurate outcomes. It is also likely to include deeper personalization as the accuracy of automated recommendations online use more sophisticated, more precise targeting predictions through smarter data. Machine learning is also likely to progress intelligent features including emotion detection through speech, vision and facial recognition which will also help to personalizing computing experiences.
Machine learning has a bright and exciting future with the potential to do a great deal for industry, societies and the environment. Although our knowledge and application of what we can do with this technology is still in its infancy, the potential is huge.