The Connection Between Data Science, Machine Learning and Artificial Intelligence

Is there any connection between Data Science, Machine Learning and Artificial Intelligence? If yes, then in what order should I learn them?

This is an example of the confusion that arises when people take an academic approach to learning a real-world skill. Schooling works against you in two ways. 1) it teaches you nothing about the value of failure (tries to instill the opposite) and 2) it teaches by starting with the fundamentals.

I won’t go into the failure piece since this is very well established in technology and business. Just know that everything you want to learn lives inside your failures, not your successes. As for the core concepts piece, let me explain why this approach, while sounding practical, doesn’t work in the real world. Fundamental concepts are, by definition, disconnected from context. This is in fact their power. By remaining agnostic to any specific problem or application they can act as units of understanding, helping us rest our assumptions about our world onto something more concrete than best guesses.

But in the absence of context, fundamental concepts are unable to be applied. Concepts never work in isolation. When we go to solve real-world problems there is a strong need to understand how various concepts work together, how they will be consumed, and the trade-offs that exist between using a mix of approaches to deliver on a singular goal. We don’t really understand a fundamental concept if we have never seen it play out in a real-world setting. In short, if you don’t have a bird’s eye view of what you are working towards your ‘understanding’ about how things work is severely limited.

So if we don’t start with the fundamentals how can we learn the right concepts and know what to focus on? You have to flip the academic approach on its head. You have to learn concepts by first jumping into problems, not knowing how to solve them. You have to realize that your learning is a byproduct of solving problems, and that the concepts you need to know are born out of your struggle in solving challenges.

There is No Place to Start, There is Just Starting

I always get asked “what book should I read” or “what language should I focus on” or “what are the main concepts I need to know.” People who ask these questions are missing the point. There is no formula. You are better off reading the “wrong” book, learning the “wrong” language, and knowing the “wrong” concepts than you are not starting because of your constant anticipation around what the right approach should be. If you read the wrong book, you’ll have a much deeper appreciation for why the content of that book didn’t align with real world problems. If you learn the wrong language you’ll deeply understand what it was about that language that didn’t work on data-intensive problems. Going down the wrong path is worth its ‘weight’ in gold. If I were to give you the supposed formula for success, your knowledge would be so restricted you would fail on the job.

As for the terms Data Science, Machine Learning and AI, nobody can blame you for being confused. It’s unfortunate these terms get tossed around so casually by those more interested in hype than solving problems. Forget the term “AI” for now. The only practical application today for “AI” is using machine learning. Machine learning is the technology you are expected to understand. Data Science is likely how you wish to apply that machine learning. Data Science is the real-world application of machine learning, with the goal of creating products people use.

Everyone wants to be prepared to enter the work force. They want to show their prospective employer that they have the right skills, and the deep understanding to go along with their abilities. But skills and understanding do not come from following a set of steps. They come from solving challenges. You need to give yourself projects that convert raw data into software features that automate learning. This is the goal of the Data Scientist. Everything else will follow from your struggles in trying to make this happen.

So choose a project right now. Which project? Well, assuming you’re interested in data science for more than just a bigger paycheck, you must have problems you want to solve. There’s your project. Now build a web application to solve that problem. How do you build it? Google is your friend. Start basic, ask questions on Stack Overflow, get it up and running, take it step by step. Does this require machine learning? Ask Google. Has anyone tried solving this problem using machine learning? Are there papers? Is there anything on Kaggle? Are there blog posts about using machine learning to solve this problem? What language do I use? What language did the blog post use? How do I set up my environment to use that language? Google is your… get the point.

While it might seem natural to think that a set plan will get you your skills faster, the truth is there is no quicker way to learn what is needed than jumping in head first and struggling through the problem. Focus on the problems you are trying to solve, build real-world applications to solve them, learn to love what failure teaches you, and watch how fast everything you need to know falls in place.

For more inspiration see,

The Only Skill you Should be Concerned With

This post was originally published as a response on Quora.

Founder Kedion, Ph.D. Computational Chem, builds AI software, studies complexity, host of NonTrivial podcast.