Machine Learning, Data Science, and AI – Oh My!

Machine Learning, Data Science, and AI are three quite popular buzzwords these days. In a variety of industries and sectors, businesspeople and engineers alike are throwing these terms around like dice at a casino. But what do they?really?mean?

Machine Learning (ML)?is a software development technique that enables computers to solve problems using real-world data. It utilizes applied mathematics, statistics, and computer science in order to provide insight and inference. (Yes, you really do need to understand the mathematical and statistical theory behind these techniques in order to be effective.)

Data Science (DS)?is an interdisciplinary field that extracts knowledge from structured or unstructured data in order to derive actionable insights. Though the term itself sounds quite systematic and straightforward, data science is more of an art. Fine-tuning parameters, engineering features, and cross-validating are a delicate dance. Change any one of these things, and the result of your algorithm can drastically shift. It’s all about the correct blend of mathematical/statistical reasoning, subtle changes, and iterative improvements that really make data science effective.

Artificial Intelligence (AI)?is as simple as it sounds – intelligence demonstrated by machines (artificial) as opposed to the intelligence of a brain (typically considered the human brain). It is a broad term that encompasses machine learning and parts of data science (see diagram below). The idea of artificial intelligence has existed for several decades, with the term being coined in the 1950s. Current applications of AI include self-driving cars (Tesla), human speech simulators (Siri and Alexa), recommender systems (Netflix), and targeted advertising (Google and almost every other platform that can house ads).

Pictured below is a Venn Diagram that helps visualized these terms and how they are related to one another.

Now that you understand the high-level explanations of these buzzwords, let’s dive into some more detailed definitions commonly associated with these topics. Starting off first with Deep Learning – something that appeared in the above Venn diagram, but hasn't been defined yet.

Deep Learning?is a specific part of machine learning that focuses on artificial neural networks. This type of learning can be supervised, semi-supervised, or even unsupervised, leaving the restrictions at the door. We won’t go into the specifics of?supervised?or?unsupervised learning?as those are beyond the scope of this article, but IBM has great introductions on both of these learning methods (hyperlinked).

Machine Learning Models?are generic programs made specific by the data used to train them. They generate predictions by finding patterns extracted from data. This includes the type of machine learning algorithm (e.g. linear regression, neural network, decision tree, etc.) that will be used in the model training step of the machine learning process.

Model Training?is done through an iterative process. This determines what changes can be made to an algorithm’s weights and biases in order to get closer to the desired result. This step is done after the data has been cleaned, transformed, and engineered.

Model Evaluation?is used to estimate the generalized accuracy of a model based on data not used to train the model. This is the step that is done after model training where we can tell if the machine learning model did a good job or not. During this stage, you may find that your model didn’t perform quite like how you intended it to. In this case, you would need to go back and change parameters, pick a different algorithm, or even ensure that your data was properly cleansed and prepared.

All in all, machine learning, data science, and artificial intelligence are incredibly complex topics. While the definitions gone over in this article are helpful to gain an introductory understanding, keep in mind that these fields are still rapidly expanding, and we’ve barely scratched the surface. New techniques and methods are being discovered, and the complexity of these artistic sciences is bound to increase.?