What is AI? 

It can safely be said that today, we are living in the age of AI. Everywhere you look, companies are touting their most recent AI, machine learning, NLP, intelligent automation, and deep learning breakthroughs. AI has become the buzzword of all buzzwords, and major tech companies are embracing it as if it were one of man's most important discoveries, but what exactly is AI?

Historical perspective

To take a step back for a moment, two historical episodes were instrumental in the inception of Artificial Intelligence; the publication of Alan Turing's On Computable Numbers, in which he developed the theory of computation, and an AI workshop at Dartmouth College in 1956.

Three years later, Marvin Minsky and John McCarthy cofounded the Artificial Intelligence Project, which soon became one of the preeminent AI research centers in the world and would later evolve into the MIT Computer Science and Artificial Intelligence Laboratory. Minsky defined AI as “the science of making machines do things that would require intelligence if done by men.” However, despite some early successes, AI would discover that the computing world wasn't quite ready for the complexity of code AI produced. Researchers found it increasingly difficult to build models on even the most powerful computers of the time. 

"Asking a manager 'are you using AI?' in a few years from now will be like asking 'are you using the computer?'”
Jim Sterne, Director Emeritus Digital Analytics Association

A.I. Timeline

Source: OECD library

 Machine Learning 

Machine Learning (ML) is the subfield of computer science that explores the construction and study of algorithms that can learn from data. Such algorithms operate by building a model based on inputs and using that to make predictions or decisions, rather than following only explicitly programmed instructions.

ML “evolved from the study of pattern recognition and computational learning theory in artificial intelligence." It “explores the study and construction of algorithms that can learn from and make predictions on data—such algorithms overcome following strictly static program instructions by making data driven predictions or decisions, through building a model from sample inputs.

AI and ML can be used in the following ways:

  • Voice recognition
  • Voice search
  • Sentiment analysis
  • Flaw detection
  • Fraud detection
  • Recommendation engine
  • Facial recognition
  • Machine vision
  • Motion detection

Machine learning can be broken down into the following three categories:

  1. Supervised learning: The computer is presented with example inputs and their desired outputs, given by a “teach.
  2. Unsupervised learning: No labels are given to the learning algorithm, leaving it on its own to find structure in its input. Unsupervised learning can be a goal in itself (discovering hidden patterns in data) or a means towards an end (feature learning).er”, and the goal is to learn a general rule that maps inputs to outputs.
  3. Reinforcement learning: A computer program interacts with a dynamic environment in which it must perform a certain goal (such as driving a vehicle), without a teacher explicitly telling it whether it has come close to its goal or not. Another example is learning to play a game by playing against an opponent..

Deep Learning

Deep learning is a subset of machine learning. Neural networks expand into sprawling networks with a large number of sizeable layers that are trained using massive amounts of data. These deep neural networks underpin technology like NLP, speech recognition, and computer vision.

AlphaGo beats Lee Sedol in Go

Generative Adversarial Networks (GANs) in action

Benefits for business

7 Patterns of A.I.

In her Forbes article The Seven Patterns of AI, Kathleen Walch lays out a theory that, regardless of the application of AI, there are seven commonalities to all AI applications. Walch adds, “Any customized approach to AI is going to require its own programming and pattern, but no matter what combination these trends are used in, they all follow their own pretty standard set of rules. These seven patterns are then applied individually or in various combinations depending on the specific solution to which AI Is being applied.”

These patterns include:

  1. Hyperpersonalization
  2. Autonomous systems
  3. Predictive analytics and decision support
  4. Conversational/human interactions
  5. Patterns and anomalies
  6. Recognition systems 
  7. Goal-Driven Systems Pattern


The hyperpersonalization pattern, which can be boiled down to the slogan, ‘Treat each customer as an individual' is defined as using machine learning to develop a profile of each individual, and then having that profile learn and adapt over time for a wide variety of purposes including displaying relevant content, recommend relevant products, provide personalized recommendations and so on.

Autonomous Systems

These are physical and virtual software and hardware systems that can accomplish a task, reach a goal, interact with their surroundings, and achieve an objective with minimal human involvement. Where the primary objective of hyper-personalization is to treat people as individuals, the goal of autonomous systems is to streamline things with as little human interaction as possible.

Predictive analytics

Using ML and other cognitive approaches to understand how past or existing behaviors can help predict future outcomes or help humans make decisions about future outcomes based on these patterns. The idea is for AI-powered predictive analytics to helps users make better decisions. ML can be a constant process, adapting over time to provide better business results.

Conversational Pattern

This pattern allows machines to communicate as humans do. The conversational/human interaction pattern “is defined as machines and humans interacting with each other through conversational forms of interaction and content across a variety of methods including voice, text, and image forms.” The objective: enable machines to interact with humans exactly how humans interact with each other.

Pattern-matching pattern

Machine learning is particularly good at identifying patterns and finding anomalies or outliers. The goal of the patterns and anomalies pattern is to use ML and other cognitive approaches to learn patterns in data and learn higher-order connections between data points to see if it fits an existing pattern or if it is an outlier or anomaly. The object of this pattern is to find what fits with existing data and what doesn’t.

Recognition systems

Image and object recognition, facial recognition, audio and sound recognition, handwriting and text recognition, and gesture detection are all examples of this well-developed pattern, which computers excel at. Google, Facebook, Apple, Samsung, Huawei and a whole host of other companies are investing heavily in recognition systems. 

Goal-Driven Systems

For decades, machines have been beating humans at easily conquered games like checkers and chess. Now, through the power of reinforcement learning and much more powerful computers, machines can beat humans at some of the most complex games imaginable, including Go and multi-player games like Dota 2. Games are just the beginning to solutions that could potentially lead to breakthroughs in solving long-hoped for goals in Artificial General Intelligence (AGI).


Cluster Analysis is a method used to classify objects into groups where those similar in characteristics are grouped together in a cluster. It is also known as segmentation analysis or taxonomy analysis. Clustering helps us look at the data as a whole so we can classify the points and logical structures based on the groupings we choose to create. Clustering has several uses, including in marketing, where it is used to identify homogenous groups of buyers; in medical science, where it is used for disease classification; in geology it is used to identify the weaknesses of earthquake-prone regions.

k-means Clustering

Spatial Analysis

Spatial data, also known as “Geospatial Data” refers to records in the dataset that have a geographic aspect. It can be the location of a store, the distance between two bus stops, or the boundaries of a state. 

Decision Trees are some of the easiest models to understand and illustrate as they are straightforward and don't require complex interpretation. Decision Tree models are good for handling complex, non-linear relationships, and outliers on the data, but it also has its disadvantages. Some of its predictions tend to be weak since it is prone to overfitting. They are not very robust, because small changes in the training data can give a larger change in the output. However, they are one of the most commonly used analytical models around. 

Decision Trees

Logistic Regression

Logistic Regression is a binary classification algorithm. Meaning, it predicts either true or false, churn or not, pass or fail and other binary categorical values. This algorithm also uses a regression function which classifies the target variable based on probability. Logistic regression fits the data to an “S” shaped logistic function also known as the Sigmoid Curve. It curves from 0 to 1, where 0 is the least probable or false, while 1 is the most probable or True. This splits the data into one of the two categories and each row of data is classified by comparing the probability to a threshold between categories. 

So Why A.I.?

In the article Artificial intelligence Unlocks the True Power of Analytics, Adobe explains the vast difference between doing things in a rules-based analytics way and an AI-powered way, including the following:

Identify customers who are at risk of defecting. The difference:

  • Rules-based analytics: You manually study reports on groups of customers that have defected and try to see patterns.
  • AI-powered analytics: Your tool automatically Identifies which segments are at greatest risk of defection.

Provide warnings whenever a company activity falls outside the norm. The difference:

  • Rules-based analytics: You set a threshold for activity (e.g., “200–275 orders per hour”) and then manually investigate whether each alert is important.
  • AI-powered analytics: The AI analytics tool automatically determines that the event is worthy of an alert, then fires it off unaided. 

Conduct a root cause analysis and recommend action. The difference:

  • Rules-based analytics: You manually investigate why an event may have happened and consider possible actions.
  • AI-powered analytics: Your tool automatically evaluates what factors contributed to the event and suggests a cause and an action. 

Evaluate campaign effectiveness. The difference:

  • Rules-based analytics: The business manually sets rules and weights to attribute the value of each touch that led to a conversion.
  • AI-powered analytics: The AI analytics tool automatically weights and reports the factors that led to each successful outcome and attributes credit to each campaign element or step accordingly. 

Identify customers who are at risk of defecting. The difference:

  • Rules-based analytics: You manually study reports on groups of customers that have defected and try to see patterns.
  • AI-powered analytics: Your tool automatically Identifies which segments are at greatest risk of defection.


We are available 24/7, on multiple social channels, or connect with us below.


505 Hennessy Road, Suite 613, Causeway Bay, Hong Kong

+852 5192 1277

Rua da Estrela, Macau

+853 6616 1033



Intellligencia is a Hong Kong- and Macau-based software consulting company that works specifically in the hospitality, gaming, fintech, esports, manufacturing, retail, sports betting, and travel industries. Although located in Asia, we work with clients as far away as North America, Mexico, India, Armenia, Australia, and the Philippines.


Address: 505 Hennessy Road, #613, Causeway Bay, Hong Kong

Phone: +852 5196 1277

Email: andrew.pearson@intelligencia.co

Website: www.intelligencia.co


Sign up to receive email for the latest information.
© Intelligencia Limited. All Rights Reserved. 2023