The machine learning revolution
Wednesday, 15th June 2016
No matter what business you’re in, you’re in the business of information. Data sits at the heart of all organisations. Machine learning has not only improved the quality of traditional business modelling, but it has allowed us to analyse new types of information more quickly and easily than ever before. Business has been quick to use machine learning to improve the quality, speed and accuracy of existing business models. The second in our series on machine learning, or deep learning, writer Charlotte Roy explores the human element of machine learning and what it means for our future.
What is deep learning
Deep learning, or machine learning is the iterative process of adjusting a model to account for variability between predicted and actual outcomes. It’s what we refer to in human terms as “learning from life experience”.
Over a lifetime, we accumulate a large body of experience, which we describe as “wisdom”. We can identify patterns, and cope with variability outside normal parameters, by experiencing the anomalies that occur over a long period of time, and learning from those statistical outliers.
One of the most important ways we use wisdom is to develop insight for predict the future. Whether this be at a basic, individual level, such as knowing when we will next need to eat, or a sophisticated level, such as understanding our environment for weather forecasting, the greater our wisdom, the more accurate our predictions are about future outcomes.
Machine learning in business
In a business sense, information is essentially “data under management”. It is the insights obtained from statistical analysis of quantitative data. Machine learning is the process of improving the quality of our analytical models. It plays a vital role in improving the accuracy of the resulting business information we use to make decisions.
No matter what business you’re in, you’re in the business of information. Data sits at the heart of all organisations. Machine learning has not only improved the quality of traditional business modelling, but it has allowed us to analyse new types of information more quickly and easily than ever before. Business has been quick to use machine learning to improve the quality, speed and accuracy of existing business models.
Machine learning has enabled us to use computers to analyse sounds and images quickly, easily, accurately and importantly, inexpensively. It has uncovered new types of data for analysis that were previously too sophisticated for computers to handle. This radical change has opened up new horizons for discovery and analysis across all parts of human existence.
The ability of computers to process images and sounds is truly revolutionary. Image processing occurs across a very wide range of industries, from healthcare (pathology) to infrastructure (licence plate reading on toll roads). When computers undertake these activities, not only do they lower costs and increase the speed of processing, but they can also identify broad trends within extremely large data sets that would not be recognised by individuals analysing smaller sets of data. New avenues of exploration, such as listening to the sound the body’s joints make to diagnose medical problems, become possible when computers are used to analyse large sets of auditory data.
Machine learning is still in its infancy. It would be fair to say that it has unlocked new frontiers for analysing, understanding and predicting our environment. Although it is already used widely in many businesses, the opportunities it affords us for discovery remain largely untapped.
Machine learning will change business at all levels
Although it is impossible to accurately predict the future, the following assumptions can provide us with some insight into what that future could look like. If we assume technology will continue to support better data capture, it is likely we will have an increasing volume of data covering a wider range of variables to work with. If we also assume that computing power increases, whether that be at the individual or networked level, we will be able to analyse larger sets of data using more sophisticated models, and produce the resulting information at a faster rate.
The more often a model is used, the more results we will have to compare against actual real-world outcomes. Additionally by using more variable, our models can become more accurate. We can therefore conclude that the iterative process of machine learning will lead to an increased sensitivity in our models.
When taken to an extreme, an infinite amount of data, analysed instantaneously by completely accurate models, would provide information and signals for business in real time. What machine learning can do, however, is take this one step further, by using historical data to predict future outcomes based on current variables. When we relate this conclusion to the original premise that information is simply data under management, we can see that machine learning will have profound affects on business.
In future, it is likely that machine learning will replace human experience in decision-making. Not only are machines able to use more points of data than any human ever realistically could over the span of a lifetime, but they are devoid of the inerrant biases every individual possesses. So no matter how diligent we are, machines will always have the upper hand in analysing data.
As the accuracy of predictive modelling improves, business will rely more heavily on the insights it provides for decision-making. Most businesses already do to some extent for production and inventory management. However, it is likely machine learning will become more pervasive, and influence decision making higher up the organisational structure. It is will become a more important strategic tool in the arsenal of weapons businesses use to compete in the global marketplace. Machine learning already differentiates businesses; providing industry leaders with a powerful competitive advantage. This trend will only increase as data capture and computing power grow.