What is Machine Learning’s relationship with AI?

The beginnings of AI
Oct 11, 2019

Welcome to our 12 part series on the topic of Artificial Intelligence or AI. This article is part 2 of the 12 part series. The complete series should provide an in-depth look at this highly fascinating subject and its deepening use case which evolves daily as we progress into the Fourth Industrial Revolution.

The machine and the human

What is Machine Learning’s relationship with AI? These terms are used interchangeably in conversation around Big Data and Analytics these days giving the impression that they are defined in the same way.

It will be worth revisiting the definition of AI from our introductory article yesterday which focused on the History of AI.

Definition of AI: The theory and development of computer systems able to perform tasks normally requiring human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages. – oxforddictionaries.com

This essentially means that computers can carry out tasks in a way that we could consider as smart.

Machine learning however, is a a current application or function or AI; a subset of the discipline.

Credited to Arthur Samuel in 1959 – his breakthrough about machine learning proposed that rather than teaching computers everything they need to know about the world and how to carry out tasks, it might be possible to teach them to learn for themselves. [1] He described it as the “ability to learn without being explicitly programmed.”

So instead of hard coding software routines with specific instructions to accomplish a particular task, machine learning is a way of “training” an algorithm so that it can learn how. “Training” involves feeding huge amounts of data to the algorithm and allowing the algorithm to adjust itself and improve. [2]

Some examples of tasks best solved by machine learning include:

  • Recognizing patterns: objects in real scenes, facial identities or facial expressions, and/or spoken words
  • Recognizing anomalies: unusual sequences of credit card transactions, unusual patterns of sensor readings in a nuclear power plant
  • Prediction: future stock prices or currency exchange rates, which movies a person will like

Main Types of Machine Learning

The main pathways in which machines are trained to learn come from a supervised, unsupervised and reinforcement based exercises.

Main ways in which machines learn. Image courtesy: datascience.com
Machine Learning: You are a cat. A very cute cat. Image courtesy: Hosico Cat

Supervised

Teach a machine by example. As an example, you might provide a computer a teaching set of cat images, some will be marked ‘this is a cat’, others are marked ‘this is not a cat’. Thereafter you introduce a series of photos to the computer which should identify which is and isn’t a cat. Every photo will get added to the teaching set as identified correctly or incorrectly.

So we can understand machine learning in a key defining factor: its ability to modify itself when exposed to more data [3]

The difficulty here though is the enormous amounts of labelled data sets required to train a system which will need exposure to millions of data pieces to master a task.

Global tech companies crunch millions of data sets to continually train machine learning systems. Google’s Open Images Dataset having about nine million images. Its labeled video repository YouTube-8M links to seven million labeled videos. Facebook recently announced that it had compiled 3.5 billion images publicly available on Instagram, using hashtags attached to each image as labels.


Unsupervised

System identifies similarities in patterns of homes in the same neighbourhood

In this instance, algorithms are tasked with identifying patterns in data, trying to spot similarities that split that data in to categories.
Airbnb as an example groups homes together in the same neigbourhood. Google News will group together topics of similar news daily. This type of data is not designed to highlight specific types of data but to group similarities and point out deviations.

Reinforcement

An example of reinforcement learning, Google’s Deepmind has triumphed over humans in many vintage video games. Learning through reinforcement helped train this system.

Think of this scenario: Someone plays an old school game for the first time with no context. As a complete novice, they may fail the first few times, but as they start noticing the patterns of the buttons pressed, the action on screen and their performance scoring, their play will improve dramatically.

Neural Networks

The easy definition: it is a computer system designed to work by classifying information in the same way that a human brain does [4].

How do Neural networks fit into the flow of machine learning? To understand it simply: The network is constructed similar to a human brain and interconnected with layers of algorithms called neurons. Each layer is dedicated to analysing a different aspect of a data set.

To demonstrate this concept of the layers, imagine that we had to use machine learning to recognise hand written letters of the alphabet. The first layer might focus on the size of the letters. The second layer may look at the curves or straight lines of every letter. The third layer might capture the direction of every letter’s formation.

The network will gradually learn each component of the letters and a weighting system is then applied to give a layer significance. Each training cycle’s conclusion with identifying letters will examine whether the neural network’s final output is moving closer to or further away from the desired outcome of correctly identifying the handwritten letters of the alphabet.

Neural Network’s structure. Image courtesy: hackernoon.com

Application of Machine Learning Enterprise Applications for Industry

The most important use cases of machine learning in modern day run through nearly every major field serving humans. Here is a run-through of some of the broad use case areas:

  1. Process Automation: This uses human decision but within the framework of boundaries and patterns. This works towards leaving the human free to focus on product innovation and improved customer service.
  2. Customer Service: Chatbots and virtual digital assistants are a common feature on new landscape of work. Traditional organograms might be hard-pressed to accurately pinpoint the place of these services alongside humans. What is beneficial from this process is the high amount of data processed which allows better training of the machine learning system.
  3. Healthcare: The company Medecision developed an algorithm that was able to identify eight variables to predict when diabetes patients could avoid being hospitalised.
  4. Marketing Personalisation: You are addressed by name in an email newsletter which carries all the products you looked at online yesterday. Hyper personalisation and curated products based on preference come from machine learning systems studying and serving against user behaviour.
  5. Security: ABI research analysts estimate that machine learning in data security will increase spending in analytics, big data and artificial intelligence to $96 billion by 2021.

Voice analytics, image analytics and sensor technologies fall into deep learning and will be explored in greater depth in the next article.

And finally a quick round up of machine learning algorithms:

  • Linear regression
  • Logistic regression
  • Linear Discriminant Analysis
  • Classification and regression trees
  • Naive Bayes
  • K-Nearest Neighbours
  • Learning Vector Quantization
  • Suport Vector Machines
  • Bagging and Random Forest
  • Boosting and Ad-Boost

Coming up next in part 3 of this 12 part series, we focus on AI & Deep Learning.

References:

[1] Forbes, Bernard Marr, What is the difference between Artificial Intelligence and Machine Learning? ,https://www.forbes.com/sites/bernardmarr/2016/12/06/what-is-the-difference-between-artificial-intelligence-and-machine-learning/#b624e8a2742b

[2] IOTforall, Calum McClelland, The Difference Between Artificial Intelligence, Machine Learning, and Deep Learning, https://medium.com/iotforall/the-difference-between-artificial-intelligence-machine-learning-and-deep-learning-3aa67bff5991

[3] Skymind.ai, Artificial Intelligence (AI) vs. Machine Learning vs. Deep Learning, https://skymind.ai/wiki/ai-vs-machine-learning-vs-deep-learning

[4] Forbes, Bernard Marr, What is the difference between Artificial Intelligence and Machine Learning? ,https://www.forbes.com/sites/bernardmarr/2016/12/06/what-is-the-difference-between-artificial-intelligence-and-machine-learning/#b624e8a2742b


Want to continue the conversation? Leave your questions in the comments below. Or you can email us too on talk@sociallyacceptable.co.za

Ciao for now.

Socially Acceptable is a South African media and communications company, with a focus on technology, digital marketing and public relations. With offices in Johannesburg and Durban, we provide services throughout the country and African continent.

Leave a Reply

Your email address will not be published. Required fields are marked *

Get a Quote