We all hear these terms being thrown around and often used interchangeably; some of us tag along without knowing what they mean, or we might see them as buzzwords, and others claim to know — and do — what these terms really entail.
Note that the distinctions between these terms aren’t clear-cut, but this article will help to give a sense of the general uses of the terms, how they are related to one another, and how all are threaded together by data science.
IoT For All explains that Artificial Intelligence describes machines that can perform tasks resembling those of humans. So AI implies machines that artificially model human intelligence. AI systems help us manage, model, and analyze complex systems.
Media portrayal of AI can make it seem like robots will takeover the world and control humankind, but this is highly unlikely and isn’t the most imminent threat of AI. More danger lies in placing too much trust in the AI systems we build, because they have potential to make flawed predictions or draw incomplete conclusions based off of faulty data. Because of this, there is a growing need for recognizing the limitations of AI.
As BBC puts it, “A system is only as good as the data it learns from.” Of course, data is not the only input into an AI system, as there are many other driving factors that shape the design of an AI system. Seeking to drive better impacts of AI, OpenAI emerged as a nonprofit AI research company centered on the deployment of responsible, safe, and beneficial AI systems.
A notable extension of AI is something called Decision Intelligence (DI). This is a new and growing discipline that spans AI, ML, decision theory and social science. It seeks to understand and model how decisions lead to outcomes, by implementing AI and ML at larger scales in organizational decision-making and society.
Machine learning essentially is a building block for AI. By doing machine learning, you are teaching a machine to learn how to perform a task, such as image recognition, recommender systems, fraud detection, etc.
Zendesk describes ML as “Algorithms that parse data, learn from that data, and then apply what they’ve learned to make informed decisions.”
There are many different types of ML algorithms, including Linear Regression, Support Vector Machines, Naive Bayes, Decision Trees, and more. You can find a comprehensive overview of the top 10 most common ML algorithms here.
Deep learning is one of many approaches to ML. It implements an Artificial Neural Network (ANN), which has multiple layers between its input and output layers. The “deep” in deep learning refers to the many layers in a network that allows for more complex processing. Check out 3Blue1Brown’s video on neural networks here.
Google’s AlphaGo is an example of a deep learning technology; it was able to learn and master the game AlphaGo, and ultimately beat the proclaimed best players in the world. Read more about industry applications of deep learning in Forbes’s article, “Thirteen companies that use deep learning to produce actionable results.”
Data science is a broad field that spans the collection, management, analysis and interpretation of large amounts of data with a wide range of applications. It integrates all the terms above and more to summarize or extract insights from data (exploratory data analysis) and make predictions from large datasets (predictive analytics).
The field involves many different disciplines and tools, including statistical inference, domain knowledge (expertise), data visualization, experiment design, and communication. Data science helps answer the question “what if?” and it plays a crucial role in building ML and AI systems, and vice versa.
If you’re more of a visuals person/designer, take a look at 10 of the best data visualization examples from history to today.
To learn more about data science, read our post on breaking down data science here, and checkout this video by technologist and YouTuber Joma Tech.