The Evolution of Machine Learning


“Computers are able to see, hear and learn. Welcome to the future.” – Dave Waters

By now, we’ve all been introduced to machine learning in some way. Whether we’re aware of it or not, we’ve encountered AI throughout the course of our day. Netflix suggests movies we might like to see next, Google Maps recently launched bus delay forecasts for commuters, and even Hello Barbie has gotten an upgrade with natural language processing (the doll produces intelligent responses to children’s questions). So what exactly is machine learning?

Machine learning (ML) is a type of artificial intelligence that uses computer algorithms to learn from data. The most groundbreaking aspect of ML is that it’s autonomous. You don’t need to heavily monitor it or teach it to learn new things – It is designed to self-correct and become smarter as it sifts through data.

From self-driving cars to technology that helps law enforcement catch criminals, the possibilities of this technology are exponential. With this comes new security issues and even more sophisticated cyber attacks (adversarial machine learning), along with the occasional dissenter warning that AI will take all of our jobs. However, most are optimistic about how ML will change society.

So where did all of this begin?

The Dawn of Machine Learning


In 1943, the earliest known neural network was developed when neurophysiologist Warren McCulloch and mathematician Walter Pitts created an electrical model showing how neurons function. Designed to mimic how the human brain learns, modern neural networks can recognize patterns in a vast sea of information.

Alan Turing’s renowned Turing Test gauged a computer program’s ability to convince participants that they were interacting with a real human rather than a computer. Aside from a few other early AI models, machine learning capabilities remained largely unknown for the next few decades.

In 1986, another breakthrough in neural networks came when Stanford researchers employed back propagation (a multi-layered network). Building off an algorithm created in the 60s, the researchers created one of the first ‘slow learning’ models. Soon after in 1990, computer scientist Robert Schapire’s paper introduced the concept of boosting algorithms, which pushed our understanding of neural networks even further. These and other nuggets of discovery over several decades allowed machine learning to finally break through to mainstream thought in the 2000s.

As the 21st century progressed, big companies began investing in machine learning projects in rapid succession. As people began to take note of AI’s vast capabilities, Facebook’s DeepFace, Google’s DeepMind, and Amazon’s Machine Learning took off in 2014-2015.

Since then, we’ve become increasingly aware of the reality of AI as a major part of our future. With all that said, how does this pertain to the average business that wants more efficiency?

Machine Learning & the Modern Business

The most important thing to consider as AI and business are increasingly fused is automation – especially marketing automation. Phasing out as many tedious, manual tasks as possible without becoming overwhelmed by new technology is the sweet spot. Where that spot is will vary according to your budget, daily operations, type of business, preferences, and goals.

Unlike the old days when any type of AI was only for mega-brands with big budgets, machine learning is making the art of predicting more affordable. This is leveling the playing field, as being able to forecast into the future is a huge part of what makes a business profitable and sustainable. It no longer takes a massive investment to run advanced algorithms and receive answers to a variety of different operational questions. With this knowledge gap being bridged, smaller organizations have more power than ever before.

Another aspect of business that machine learning is radically shifting is customer service. This ties closely into automation as many of the customer service tools being integrated are automated tools themselves.

“Chatbots, recommendation systems, personalized message mechanisms, smart advertising targeting tools, and image recognition tools all help companies to interact with customers quickly. An automated agent/virtual assistant/chatbot is the most typical AI application for interacting with customers and answering their questions instantly.” source

Along with faster customer service, big retailers like Burberry are showing how product-based companies can use machine learning to detect counterfeits and boost sales. Loyalty and rewards programs are helping Burberry personalize the shopping experience both online and in person. Back in 2015, they announced that machine learning and AI investments had already yielded a 50% jump in repeat customers.

Yet another area where machine learning saves the day is in data management – especially for organizations with few or no staff members who are trained in data science. Machine learning can make sense of unstructured data and help you format it for strategic decision-making. In a world where it’s all too easy to collect massive amounts of data, but quite difficult to organize and interpret it, ML is a game-changer.

To get started with any kind of AI, look at where you currently stand. What big problems need to be solved? What kind of infrastructure do you have? What talent gaps exist on your team? While automation is the name of the game with ML, there’s still some management and hands-on work necessary to keep things running smoothly. Furthermore, making too many big changes at once will make it harder for ML to grow into a seamless part of company culture. For most organizations, implementing ML will be a slow and steady process that continues to yield fruit well into the future.

Like what you just read?
There’s more where that came from.