What is machine learning and how is it used?

Machine learning is a term that's used frequently in the tech world but for some, it's still a source of confusion. Of course, we're all familiar with the concept of AI, and that's what machine learning seems to be describing, however they aren't interchangeable. Machine learning isn't so much about your smart fridge becoming self aware and asking difficult existential questions – It's more easily thought of as the algorithm that allows your preferred media streaming platform to learn what you like and make smart recommendations.

Systems built around machine learning have created massive changes in the digital world and this technology is helping facilitate positive change our analog lives as well. Let's take a look at what machine learning is, how it works (we'll keep it simple, don't worry) and how it's already making waves.

What is machine learning?

Machine learning describes algorithms developed to process and apply complex calculations to large sets of data, detect patterns and "learn" without being explicitly programmed. While such algorithms have existed for a while, the reason they are becoming increasingly prevalent in big data and analytics is because we've developed the tools to let these algorithms do a lot more, a lot faster. Big data guru and best-selling author Bernard Marr describes it in very simple terms, which we'll paraphrase here:

Think of your data set as a huge number of images. A considerable number of these images are of cats, and each image is labeled either "a cat" or "not a cat". The algorithm analyses each image and determines that all photos labeled "a cat" have similar characteristics. The algorithm has essentially learned what a cat looks like. From here, you could feed the algorithm images and ask it to identify which ones are of cats.

Machine learning: If an algorithm can learn what a cat looks like, what else can it do?This one is definitely a cat.

The key difference between AI and machine learning

The concept of artificial intelligence is over 3,000 years old.

In order to differentiate these two terms, we first need to look briefly at what artificial intelligence is. This concept is over 3,000 years old – artificial beings have been present in myths and legends dating back to the Indian philosophy of Charvaka circa 1500 BC, though the concept has existed in its current form since the birth of computer science in the early 20th century. At its most basic level, artificial intelligence describes a computer capable of mimicking the human decision making process.

Modern AI development generally falls into two areas. The first is applied AI, in which a system is designed to undertake a particular task. For example, this might be AI designed to react to real-time environmental data in Google's self-driving car. The second area is generalised AI, where a system is designed to be theoretically capable of handling any task.

An example of this would be AI research firm Deep Mind's "Alpha Go" algorithm, which beat the best living human Go player (for some context, Go is a 2,500 year old abstract strategy board game that's far more complex than chess – there are famously more possible board configurations than there are atoms in the visible universe). So, while applied AI reacts, generalised AI preemptively strategises.

Machine learning is a subset of general AI, and it was born from researchers experimenting to see whether computers could learn from patterns in data. The definition of AI might encompass what we know as machine learning, however machine learning only describes a fraction of what defines AI.

What's the difference between AI and machine learning?Machine learning doesn't refer to a machine capable of mimicking the human thought process, but such a machine would (will?) be built on similar tech.

Where is machine learning applied?

We've mentioned self-driving cars and algorithms that recommend video or audio media based on your history, so we'll skip over those for now.

  • Healthcare

Machine learning in healthcare is one of the most meaningful and profoundly game changing applications to date. Analysis of medical data is already being used to detect the early stages of breast cancer, and to reduce the incidence of avoidable hospitalisation for patients with diabetes. Additionally, it's allowed us to identify tuberculosis in chest x-rays and to gain a better understanding of risk factors for disease in large populations.

  • Security

From identifying malware variants to predicting security breaches, machine learning is helping us to stay protected in an age where cybercrime is one of the greatest threats to businesses. Not limited to digital security, algorithms using image recognition can help identify red flags that human personnel might miss in screening processes at airports, stadiums and other public places.

  • Marketing

With a dramatic increase in the revenues for collecting customer data, machine learning has helped us to refine marketing strategies to be more personalised. We're able to make smart recommendations of products, custom tailored offers, specialised newsletters and accurately targeted ads, all of which shepherd customers towards a sale. It's difficult to engage customers without knowing what they want, but with machine learning we don't have to.

  • Financial services

Banks, investment platforms and other financial institutions can use machine learning technology to gather valuable insights from data which can identify investment opportunities and inform investors when to trade. The speed with which these insights can be generated is responsible for these companies making smart decisions that regular old humans didn't have the processing power for. From establishing more people-friendly solutions to stock market predictions, machine learning is of great benefit in this area.

  • Search engines

You may or may not have noticed that search engines are a lot better than they used to be. This is because machine learning has helped platforms such as Google better understand how we interact with search results. This includes watching how users respond to results to provide a more personalised service, as well as determining which websites are best answering common queries and ranking them in order of usefulness. Machine learning is helping us get the information we need, quicker. Businesses can look at search engine algorithms and tailor their content to perform better in search.

Are you ready to find out what machine learning processes can do for your business? Get in touch with AtoBI today.