Bergur Thormundsson
Research expert covering technology in the Nordics, AI and smart cities
Machine learning - statistics & facts
Interesting as its title might be, machine learning is the subset of artificial intelligence (AI) that does the least learning. Nonetheless, it is one of the most ubiquitous types of AI. Machine learning is a type of software that aims to automate and simplify processes with simple programs. It is ever-present in the modern world, being used in nearly all industries. Automatic responses to queries, automated stock trading, computer vision, recommendation engines, and customer service are some examples of the applications of machine learning.
Machine learning is divided into three types of learning systems: supervised; unsupervised; and semi-supervised. Supervised machine learning uses labeled datasets to easily classify incoming information to be more comprehensible for human consumption. An example of this is the automatic categorization of emails into spam or relevant emails. Unsupervised machine learning uses unlabeled datasets to identify similarities or differences between datasets. An example of this is separating customers based on interests. Meanwhile, semi-supervised machine learning is a hybridization of the two and allows specifically labeled datasets to categorize unlabeled data.
What kind of AI is machine learning?
Of the forms of AI used, machine learning is the simplest, but that makes it also one of the most useful. Other AI subsets include deep learning, neural networks, and natural language processing. Machine learning uses datasets to categorize incoming information and provides solutions based on those narrow categorizations. The amount of data required to run machine learning programs is vast, but nearly all are categorized by programmers, making it more reliant on human involvement.Machine learning is divided into three types of learning systems: supervised; unsupervised; and semi-supervised. Supervised machine learning uses labeled datasets to easily classify incoming information to be more comprehensible for human consumption. An example of this is the automatic categorization of emails into spam or relevant emails. Unsupervised machine learning uses unlabeled datasets to identify similarities or differences between datasets. An example of this is separating customers based on interests. Meanwhile, semi-supervised machine learning is a hybridization of the two and allows specifically labeled datasets to categorize unlabeled data.
Challenges to machine learning
The main challenges to the increased use of machine learning are ethical and personal. Prime among these is privacy, as the datasets retain personal data for further categorization purposes. Data that is available on servers is also at risk of being hacked. Another concern is automation in workplaces, leading to a reduced number of jobs and relocations. Technical superiority will make a difference between societies, with more developed economies reaping a greater benefit from the increased automation. How companies and industries respond to these challenges will have a significant impact on their capability in the field.Machine learning remains the largest AI segment
This particular subset of AI is not going anywhere, being needed for any smaller automation task that is influenced by AI. Machine learning is therefore likely to remain a major segment of the technology market in the coming years.
