Discover Artificial Intelligence and Machine Learning

There is an old saying in computer science that if a function can be written in two different ways, then it will be. The premise of this article series is to lay bare the second way. This week’s article will introduce you to both Artificial Intelligence and Machine Learning.

Two terms are often used interchangeably when discussing technology: Artificial Intelligence (AI) and Machine Learning (ML). While there’s overlap between these disciplines, they are not interchangeable words for the same concept. But first things first…

What Is AI?

Artificial Intelligence is one branch of Computer Science that has kindled since antiquity. Some say it was born with the creation of the Turing Test by Alan Turing in 1950. Others might reference the musings of Ada Lovelace in the 1800s. For our purposes, Artificial Intelligence is intelligence demonstrated by machines that exhibit characteristics of being self-aware and being able to learn autonomously or intelligently.

Machine Learning is a subset of Artificial Intelligence that gives computers the power to develop new skills without explicitly programming them. This could mean identifying spam emails in your Mac’s inbox, it could mean recognizing the difference between images of cats and dogs, or even teaching Alexa how to answer more complex questions. It does this by using algorithms for decision making.

Machine learning can be divided into three categories: supervised learning, unsupervised learning, & reinforcement learning. Supervised learning takes labelled examples (training data) and learns from them to make future decisions about unlabeled data. Unsupervised learning focuses on data that is unlabeled, using clustering to discern similarities and differences. Reinforcement learning takes input from the environment and decides which action to take next based on its series of previous actions.

While there are many nuances within each category, this is a brief overview for those just starting out. For now, let’s focus on what unites Machine Learning with Artificial Intelligence – mathematics & statistics.

Without getting too technical (for now), algorithms are used to represent any intelligent decision-making system as mathematical models. An algorithm does not need to be executed by a computer; it can be represented by symbols, images, or even physical objects like chess pieces. This leads us around to the old adage I mentioned at the beginning of the article: If we can represent a problem using math and statistics, there is likely an algorithm for solving it.

But how do we formulate these models and choose which algorithms to use? Enter Machine learning. These techniques make decisions based on existing data which means they require less human input than traditional programming. The algorithm builds its own model by recognizing patterns in data rather than following explicit instructions programmed onto it by a developer. For this reason, many first assume that humans are not needed in creating intelligent systems – this is where misconceptions about AI arise.   

Machine Learning vs Artificial Intelligence

Artificial Intelligence is concerned with making computers capable of performing tasks that would normally require human intelligence. Machine Learning gives them the ability to learn from data.

While Machine Learning is a subset of AI, there are many branches within the domain of Artificial Intelligence itself. Other types include natural language processing, speech recognition, computer vision & image processing, robotics (including self-driving cars ), game playing (which leads us back to chess), expert systems (decision-making software), and search. There’s also the related field of statistics which has many practical uses in real life.

Supervised learning: Using labelled examples to make decisions about unlabeled data Unsupervised learning: Using unlabeled data to make decisions Self-Taught Learning: Transferring knowledge from one set of problem instances to another Reinforcement learning: Interacting with a dynamic environment without specific instructions Search Techniques: Locating items according to some measure of their usefulness

While unsupervised and reinforcement learning does not require labelled training data, they also have fewer practical applications. This leaves supervised and self-taught learning as the most common types. As such, we will focus on them as they are relevant for Machine Learning.

Supervised Learning: Human Supervision is Still Needed  

The goal here is to introduce a set of samples (training data) with labels (i.e. answers) that can be used by an algorithm in making future decisions about unlabeled data (test samples). The algorithm tries different techniques in order to generalize from the few examples it has seen in order to make accurate predictions across all possible situations it might encounter. This is not a trivial task because it requires algorithmic learning (by math & statistics).

Here, we use labelled data to make decisions about unlabeled data. Supervised Learning algorithms learn from examples and experience, making them incredibly useful in many real-world situations. For example:

Data Mining: Discovering patterns in large amounts of data Clustering: Grouping together similar things Classification: Picking out clear categories within the data Regression: Predicting numerical values for continuous variables

Human input is required at various stages during supervised learning since choosing relevant features and determining what to do with labels are both expertise only we possess. As such, this is typically an iterative process involving multiple steps before structure emerges from the noise. Engineers are needed to interpret, refine, and interact with the data before it becomes useful information.

There is an old saying in computer science that if a function can be written in two different ways, then it will be. The premise of this article series is to lay bare the second way. This week’s article will introduce you to both Artificial Intelligence and Machine Learning. Two terms are often used interchangeably…