Artificial intelligence (AI), machine learning (ML), and deep learning (DL) are often used interchangeably. However, they are different, and many people are confused between them. AI is an umbrella term that encompasses ML and DL. ML is a subfield of AI, while DL is a subset of ML.
What is AI?
Artificial intelligence is not a new term. It was coined by John McCarthy in 1956. Since then, it has become one of the most disruptive technologies that will change the way we live in this world. Its original goal was to let the machines mimic human intelligence and generate human-like action or decision. This phrase came first to be a concept before ML became a catchphrase as a method for achieving artificial intelligence.
AI can be classified into two types: strong AI, also known as general AI or artificial general intelligence (AGI), and weak AI or narrow AI. The strong AI refers to the machines that have human intelligence and can solve any problem instead of one problem. Simply put, this type of AI is versatile and can help solve many problems. One example is Ava, a humanoid AI, in the movie Ex Machina. As a result, strong AI indicates the AI technology we are pursuing in the future.
On the contrary, the weak AI denotes that the machines or computers can only accomplish one intellectual task on their own. For example, Amazon’s online shopping recommendation system, Google’s Gmail spam filter, DeepMind’s AlphaGo, and iPhone’s Siri are all designed and programmed to perform one particular task. Generally, this AI represents the AI technologies that people are using today.
What is ML?
Machine learning refers to the ability of the machines to learn from data by themselves, just like humans learn from experience. And then, those machines can make a decision or prediction for something, based on the model generated by the input data and the learning algorithm(s) without the need to use explicitly programming method for the machines. Typically, ML can be divided into three categories: supervised learning, unsupervised learning, and reinforcement learning.
Supervised learning is a type of machine learning algorithm that utilizes labeled data to train itself. It employs input variables and an output variable to map the function. The term “supervised learning” may also refer to the idea of having someone to supervise the model training process. However, it can be split into two types: classification and regression.
Classification is often used to deal with the discrete output variable. Popular classification algorithms include:
On the other hand, a regression algorithm is usually applied when output is numerical. Examples of regression algorithm consist of:
Note that the decision tree algorithm can be used to solve either classification or regression problem, depending on the type of predicted outcome. The classification tree can be applied when the outcome is categorical. The regression tree can be adopted when the outcome is continuous.
Unsupervised machine learning does not leverage labeled data. In other words, it only receives unlabeled data that contains input variables but no output variables. Simply put, there are no correct answers for unlabeled data. The machines learn on their own to discover the hidden patterns and functions from the input data. In this case, there is no teacher to supervise the model training process.
The following lists some of the most common algorithms used in unsupervised learning:
In general, neural network algorithms can also be either supervised or unsupervised. However, in today’s neural network applications, many of them belong to supervised learning, such as Convolution Neural Networks (CNN), and Recurrent Neural Networks (RNN). Conversely, many neural network algorithms are considered unsupervised, such as Autoencoders and Restricted Boltzmann machines (RBM).
To recap, the primary difference between supervised and unsupervised learning is the use of ground truth datasets. Supervised learning leverages datasets tagged with annotation, whereas unsupervised learning does not.
Shing-Yi Lai, Technical Writer, inwinSTACK