Machine learning is a subset of artificial intelligence that enables computers to capture data and interpret it using algorithms and techniques like pattern recognition and trend detection.
A major goal in the research in this field is to build computers capable of acquiring knowledge on their own, of improving their performance with practice, without being explicitly programmed. Machine learning shouldn’t be confused with data mining or knowledge discovery in databases, even though they use similar technologies.
The term “machine learning” was first coined in 1959, by Arthur Samuel. The Samuel checkers-playing program was the first successful self-learning program that proved the early fundamental concepts of artificial intelligence. This follows the publication in the academic journal Mind, of the study “Computing machinery and intelligence”. The article referred to the Turing test which is an investigation of whether machines can think. An important concept in artificial intelligence is the extent to which machines can replicate human understanding and capacity to reason. If a human judge is engaged in a conversation with a computer without knowing it, and the human doesn’t notice the difference, the computer passes the test.
Tom M. Mitchell, machine learning expert and professor at Carnegie Mellon University, predicted the development of machine learning and the extensive influence it will have in the future.
Banks and financial institutions use machine learning to predict market performance and to prevent fraud. Public earnings documents combined with information about the stock price and what happens in the market make possible stock performance prediction. Many trading firms use machine learning algorithms to predict and execute trades at massive volume. Firms that use proprietary systems have a clear competitive advantage since no independent trader can consume the amount of information available, at the speed with which algorithms can execute a trade.
Machine learning is a growing trend in the healthcare industry. With the rise of wearable devices and sensors, doctors can now determine patients’ health in real time. The use of technology to better interpret data collected from the patient, correlated with large datasets of healthcare data leads to improved diagnoses and treatment. A major concern is finding ways to use the vast amount of data available while protecting its confidentiality. Luckily, we have found ways to apply algorithms, without human intervention, to analyse data without breaching confidentiality contracts.
Machine learning has brought significant advantages to business. Companies can now fine-tune their marketing customisation based on the data they have from the customers. More than ever, companies create the products they think their clients want, products that would solve their problems and that would make them come back for more. In marketing, everything can and will be customized. From communication campaigns, text messages, emails, the platform where the customer is active, to the way is approached on social media. Machine learning makes online marketing an entirely customizable experience.
The recommendations we get from retailers like Amazon, streaming services like Netflix and what we see on our daily Facebook feed are all ran by algorithms that analyse past data, compare it with millions of data from other users and serves suggestions of what we might like to buy or watch. One of the fascinating things about machine learning is its ability to self improve over time.
Online search is probably one of the most famous applications of machine learning. Every time we search for something online, the search engine monitors our behaviour on the page. If we don’t click on the top links, Google assumes the results weren’t relevant enough; it learns from this mistake and works to refine the results. RankBrain is an artificial intelligence system that’s being used to offer optimal search results. Rooted in machine learning, Rank Brain uses an advanced understanding of semantics and how people search and applies those learnings to future search results. RankBrain is not being programmed to respond to certain situations in a specific way, but it can upgrade itself, to make connections between ambiguous queries and specific terms to understand user intentions better. The average user of the search engines may not have noticed a difference in the search results, but there are significant developments in search queries which could influence the future development of search algorithms.