What is machine learning (ML)?

Machine learning (ML) enables systems to improve performance by learning from experiences and data.


Also for WindowsiOSAndroidChromebook and For Business

Machine learning may sound like a course you have to take to work at a factory, but in computing, it’s a subset of artificial intelligence (AI). In a nutshell, it allows systems to improve performance by learning from experiences and data. The AI subset is so ingrained in modern software, that many existing technologies that we take for granted would not be possible without it.

What exactly is machine learning (with examples)?

So, what is machine learning (ML) exactly, and why is it so important? Well, ML is an application of AI and a branch of computer science that allows systems to learn from experience, data, and algorithms in order to enhance accuracy. Interestingly, developers don’t have to explicitly program machines to take advantage of ML — computers can learn themselves without human action.

Modern technology like chatbots, virtual assistants, proactive antivirus software, and more rely heavily on machine learning. For example, self-driving vehicles use multiple ML algorithms based on both supervised and unsupervised learning models to decide what actions to take in order to drive as well as or better than humans.

Likewise, cybersecurity for businesses like Endpoint Detection and Response (EDR) can use machine learning to detect unknown malware and find unknown “zero-day” threats by identifying malicious patterns. This is just the tip of the iceberg regarding how machine learning will impact cybersecurity in the future.

ML and AI also allow for the communication of machines without human intervention. Such machine-to-machine (M2M) applications can help supply chain or warehouse management systems efficiently track and monitor inventory. Similarly, M2M applications help energy companies manage supply more precisely by sending collection data from energy harvesting sources to remote computers for analysis.  

What are some common machine learning methods?

Supervised machine learning

Supervised learning is when labeled datasets train algorithms under a supervisor. Supervised data is more common than other learning methods because it can be more efficient. An example of this type of ML is when algorithms can classify spam in your inbox.

Unsupervised machine learning

Unsupervised learning is when an algorithm works with unlabeled data unsupervised. It has to determine itself how to process information. Researchers can use unsupervised learning to find patterns and data groupings in unlabeled datasets without intervening. Of course, a data analyst may still need to validate unsupervised machine learning recommendations.

Semi-supervised machine learning

Semi-supervised learning strikes a balance between supervised and unsupervised learning by training algorithms with labeled and unlabeled data. Typically, the labeled data volume is smaller, while the unlabeled data volume is much larger. An application of semi-supervised learning is at a hospital, where a radiologist labels a small number of scans for diseases to help machines accurately extract relevant information from a higher volume.

Reinforcement machine learning

Reinforcement learning is like supervised learning, except the algorithm learns through trial and error and delayed rewards instead of sample data. For example, an autonomous vehicle can learn through mistakes in a training environment what decisions are undesirable. Likewise, a health care system can use it to determine optimal policies from past experiences.

What is the difference between AI and machine learning?

When looking up AI vs machine learning, you may notice that some people incorrectly use the terms interchangeably. AI is essentially an umbrella term for synthetic intelligence. Meanwhile, machine learning is an AI focus that allows machines to learn from experiences and data without someone programming or assisting them to do so. A machine that uses AI may not necessarily have ML capabilities. For example, in 1996, IBM’s Deep Blue chess-playing system used more AI and less ML to defeat Russian grandmaster Garry Kasparov by evaluating countless moves in real-time.

What industries use machine learning?  

  • Governments use machine learning for utilities, public safety, fraud detection, or border control by analyzing a massive volume of data.
  • Healthcare companies use machine learning to enhance diagnosis and treatment, and develop accurate health monitoring devices.
  • Retail companies can enhance targeted marketing goals by analyzing customer buying patterns.
  • The energy sector finds sources more efficiently with machine learning algorithms.
  • Many companies in the transportation sector, such as delivery companies, ridesharing businesses, and public transport, use computers to find ideal routes to improve services, profitability, and carbon footprint reduction.
  • The financial industry uses machine learning to enhance cybersecurity and essential data insights.

Does Netflix use machine learning?

Yes, Netflix uses ML or multiple applications. One of the most apparent ones is their content recommendation system. The system uses ML to examine your genre preferences, viewing history, and the viewing history of like-minded users to suggest what movies, films, and documentaries you can try.