Artificial Intelligence vs. Machine Learning – What’s the difference?
A plethora of sometimes confusing terms related to Data Science are currently being bandied around. It seems that almost every other person working in analytics calls themselves a Data Scientist, waxing lyrical about features and targets, NumPy, pandas, pickles and, if you are really (un)lucky, beautiful soup.
For those of us who are not actually part of a data science team, but perhaps work near one, these terms can seem pretentious, confusing or intimidating, or possibly all three! While I will not try to unpack all Data Science terminology in this post, I will look at three key concepts:
- Machine Learning (ML)
- Artificial Intelligence (AI)
- Deep Learning (DL)
… and attempt to explain what they are all about.
People tend to use these terms interchangeably, and while they are related and broadly describe a machine thinking, they are actually all quite distinct.
Artificial Intelligence (AI) is the process whereby machines mimic human thinking or behaviour. It is a very broad field and includes examples such as IBM’s Deep Blue (the machine that beat chess grand master Garry Kasparov in 1997), Social Media monitoring, virtual assistants, the Roomba iRobot and scientific research.
Andrew Moore, General Manager, AI & Industry Solutions, Google Cloud at Google, states that “artificial intelligence is the science and engineering of making computers behave in ways that, until recently, we thought required human intelligence.”
AI is a computer algorithm that displays intelligence through its decision-making abilities, with the primary goal being to increase the chance of success, not accuracy. Machine Learning and Deep Learning provide added efficiency to AI.
Where AI is defined as the ability to acquire and apply knowledge, Machine Learning (ML) is defined as the acquisition of knowledge or a skill. The key principle is that the machine uses data to ‘learn.’ There are two basic types of Machine Learning:
- Supervised Machine Learning
- Unsupervised Machine Learning.
Supervised Machine Learning
In Supervised Machine Learning, the data that the machine uses to learn is well labelled. For example, all credit applications are flagged as Approved or Declined. This data is used to build models that are then applied to unlabelled data to predict the likely outcome.
Examples of Supervised Machine Learning algorithms include logistic regression, decision trees, random forests, extreme gradient boosting as well as single-layer neural nets.
Unsupervised Machine Learning
Unsupervised Machine Learning uses data that is not labelled with a specific outcome. For example, we have applications for credit, but we do not know which ones were approved and which ones were declined. The primary goal with unsupervised machine learning is to model the underlying structure of the data to learn more about the data.
Examples of unsupervised machine learning algorithms include k-means (a clustering technique) and principal component analysis, which is a technique used to reduce the dimensionality of high-dimensional datasets without losing the information within the dataset. In this technique, variables are combined such that the underlying structures within the data are represented by these components.
Deep Learning (DL) is a subset of Machine Learning and uses some Machine Learning techniques such as neural networks to solve real-word problems. It is called Deep Learning because more than one layer is used in the neural nets.
This technique is data hungry and delivers the highest accuracy when trained on large datasets. This does not mean that some of the other techniques are not data hungry. For example, trees require large datasets in order to extrapolate effectively.
An example of Deep a Learning application is sentiment-based news aggregation. In both Machine Learning and Deep Learning, accuracy is king.
Data Science is a broad and deep field, with many areas of specialisation. AI, along with Machine Learning, will continue to make rapid advances and will definitely impact our future at an accelerated rate.
In the smaller, more specific ‘narrow AI’, which enables machines to perform very specific tasks, using models trained with data for that same specific task, we already leverage this technology every day.
This post covers some of the key differences between AI, ML and DL; check out the suggested reading links below if you would like to learn more.
About the Author
Carolyn Rohm is Head of Analytics at ADEPT Decisions (www.adeptdecisions.com)