What is Artificial Intelligence, Machine Learning and Deep Learning?
In business, you have probably heard people using buzzwords like “Artificial Intelligence“, “Machine Learning“, “Deep Learning” on a day-to-day basis.
Some people use them interchangeably just like they are talking about the same thing meanwhile some other people (mostly SME or tech savvy) may use them very specifically. This totally different context may confuse you about the true meaning of these terms.
In this blog post, to help you better understand these three fundamental concepts in machine learning, we are going to go through each of them one by one.
Artificial Intelligence, Machine Learning, and Deep Learning are not the same
The relationship of “Artificial Intelligence“, “Machine Learning” and “Deep Learning” can be summarised as follows:
TL;DR: They are in a hierarchical relationship: AI is the superset of ML, ML is the superset of DL.
Artificial Intelligence (AI)
AI is an umbrella term referring to a large field of study in computer science. It covers knowledge and studies which people are trying to make machine thinks like a human. According to what Stuart Russell mentioned in his all-time-classic Artificial Intelligence: A Modern Approach, AI can be broken down into 4 different aspects:
- Make computer Act humanly
- Make computer Think humanly
- Make computer Think rationally
- Make computer Act rationally
Majority of the research in the past few decades are mainly focused on the first aspect: how to make machine acting humanly.
- Natural language processing: to make computer perceive natural language like human
- Computer vision: to make computer perceive objects visually like human
and this is where machine learning falls into.
Machine Learning (ML)
ML is a subfield of AI and probably the most emerging one within the past decades. Various computer scientist discussed about the definition of ML and I am going to cover my favorite ones.
A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P, if its performance at tasks in T, as measured by P, improves with experience E.
In this definition there are three important components:
- Experience (dataset in supervised learning or environment in reinforcement learning)
- Task (the problem statement)
- Performance measure (a method to quantify how good the model is)
If a method is missing either one of them, it cannot be considered as ML.
… In a typical scenario, we have an outcome measurement, usually quantitative (such as a stock price) or categorical (such as heart attack/no heart attack), that we wish to predict based on a set of features (such as diet and clinical measurements). We have a training set of data, in which we observe the outcome and feature measurements for a set of objects (such as people). Using this data we build a prediction model, or learner, which will enable us to predict the outcome for new unseen objects. A good learner is one that accurately predicts such an outcome.
A typical example of ML problem is the handwritten digit classification.
This problem contains all three components mentioned in Tom M. Mitchell’s framework
- Experience: images of handwritten digit
- Task: to identify the digit from images
- Performance measure: the accuracy of the identification
Deep Learning (DL)
DL is a subfield of ML referring to a collection of ML methodologies which make use of a hierarchical and complex structure of mathematical models.
… to allow computers to learn from experience and understand the world in terms of a hierarchy of concepts, with each concept deﬁned through its relation to simpler concepts. By gathering knowledge from experience, this approach avoids the need for human operators to formally specify all the knowledge that the computer needs. The hierarchy of concepts enables the computer to learn complicated concepts by building them out of simpler ones. If we draw a graph showing how these concepts are built on top of each other, the graph is deep, with many layers. For this reason, we call this approach to AI deep learning.
Many recent breakthroughs in ML is actually based on DL technology.