Image with light blue background. On the left is a symbol with a brain and a computer hard drive representing machine learning and on the right is a robotic arm representing AI

The term “artificial intelligence (also known as AI)” may jar up visions of robots helping in the kitchen and self-driving cars. Although sci-fi fans and tech enthusiasts are patiently waiting for that day to arrive, AI and its subsidiary, machine learning, are already being used in our daily lives. Although often used interchangeably, machine learning is actually a subgroup of artificial intelligence.

To get down to it, machine learning refers to the algorithms and technologies that train a system to respond positively or negatively to commands, teaching the machine to make decisions and identify problems. AI is the general ability of computers to mimic human thinking and perform tasks.

What exactly is Machine Learning?

A tool to achieve artificial intelligence, machine learning uses algorithms to memorise patterns and apply each new understanding to make better and better decisions. Programmers use machine learning to see just how far they can go to improve the cognition and responsiveness of a system.

How does Machine Learning become Artificial Intelligence?

The goal of turning a simple machine into a system capable of mimicking human action and thought, starts by feeding said machine data and commands, using penalization as part of the training process. Take, for example, the many times you’ve gone onto a website only to be confronted with the test of your life: you must choose only the images that contain a car. You can rest easy knowing you’re not the only one who must endure this anxiety-inducing task; a machine had to learn it before you, taking a penalty for each time it chose anything but the car. The machine is trained by building paths based on a correct or wrong answer. Penalised paths become weaker, and the machine eventually creates paths leading to the right action, in this case, choosing the car. This type of machine learning is known as image recognition.

Image recognition also comes into play when tagging your friends on social media, and more importantly, identify whether an x-ray shows evidence of cancer. In fact, modern healthcare has greatly been improved by machine learning. For example, doctors combine facial recognition with machine learning to scan photos to see which phenotypes correlate with rare diseases, potentially leading to life-saving preventive measures. Machines can even help structure the best treatment plan and formulate a diagnosis.

Finally, machine learning can predict disease patterns to help prevent hospital readmissions or let staff know that things may get busier.

Machine learning can also be found when you settle in, ready to consume a family-sized bag of chips on your own, and Netflix is already waiting with a list of recommended movies. Another example is when you’re in the market for new music and turn to Spotify for suggestions. The listening service uses collaborative filtering - similar to Netflix - by making predictions on what the person would like based on past user experience and positive or negative responses to past suggestions. The company is also using reinforcement learning, a more complex form of machine learning which uses a pre-assembled and tested strategy to help the machine take sequential action to fulfil a specific goal. In Spotify’s case, the technology would use reinforcement learning to help listeners discover new artists and genres they wouldn’t have necessarily found without Spotify nudging them to do so.

Other types of machine learning include systems that detect fraud and cyber security attacks at financial institutions. Used in statistical arbitrage, machine learning can help predict market trends and keep a large number of securities safe and growing in value.

And what about AI?

At what point does machine learning interact with AI? And why are the above examples considered machine learning but not AI? The truth is, the above examples could fall into the AI category as well. Still, we can’t talk about them without first talking about the machine learning that makes them tick. The machine learning technologies used in the above examples become AI when it is finalised and ready to be deployed. Artificial intelligence refers to the field of developing computers and robots that can behave like and go beyond the limitations of humans.

Some forms of artificial intelligence start with machine learning - inputting data and gaining the desired result; the result is then applied to artificial intelligence. For example, Apple’s Siri or Amazon’s Alexa are two of our favourite AI pals who are there to answer our questions or schedule doctor appointments and dance lessons. Both use voice recognition technology, a type of machine learning that has been perfected (well, almost) and then implemented into phones, laptops, and speakers as AI. It also comes in handy when using the voice-to-text feature on your phone.

Another example of machine learning-based artificial intelligence is customer service chatbots which answer frequently asked questions to cut down the workload on human agents. Take also predictive text, from forming complete sentences in emails and documents to responding to funny Facebook posts with what your phone thinks should come after, “I went to jail because…” predictive text is AI that uses forecasting to guess your next best word.

While all machine learning will eventually wind up as some sort of AI (unless, of course, the project is scrapped), not all AI stems from machine learning. Symbolic logic – rules engines, expert systems, and knowledge graphs all fall under the AI umbrella but aren’t machine learning.

The difference between machine learning-based AI and other types is that AI backed by machine learning is dynamic and doesn’t depend on human interaction to better itself, making it more reliable and streamlined. Humans interact with AI and machine learning in various ways throughout the day, from your Google Home letting you know today’s forecast to your self-driving car taking you to work tomorrow (maybe).

In summary, it's important to remember that although similar, machine learning powers AI (though not all AI) and is the engine behind what makes your Netflix selection more accustomed to your liking and provides the building blocks to your iPhone’s Siri. AI is the field of computers performing human tasks and mimicking human thought to predict patterns, from bank fraud prevention to YouTube’s watch next.

Where to next?

Futurists may latch on to the idea that AI will soon take over every aspect of our lives; our world will likely stay human-controlled, at least for the near future. However, AI will likely become more prominent in our generation. For example, self-driving cars are in the works, although they may take some time to perfect. The medical field will become even more accurate, diagnosing diseases and developing drugs at a faster pace, and virtual nurse assistants can monitor patients. If making dentist appointments makes you nervous, Google is currently working on a robot that can do that for you. Some of our previous students have used machine learning and AI to predict the outcomes of sports games, classify breast cancer tissue, see how public transit impacts those with reduced mobility, and more. There is a risk, too, of AI replacing specific jobs sooner than we think, but it depends on who you ask. For the time being, we can expect processes to be more streamlined, some more customer service bots, and a lot more advancements in the medical, financial, and cyber security fields. Word’s still out on that kitchen robot, however.

If you’re itching to know more, our bootcamps can launch you into the field of machine learning, AI, data, cyber security, and more in just 12 short weeks. With wraparound support, tuition help, and lifelong career services, bootcamps are the launching pad for future automated vehicle designers.