What’s the best way to practice AI skills?
In this article, we’ll discuss how to practice your AI skills.
Learning concepts is only a small portion of the work involved in learning AI skills.
In fact, I would guess that 20% of all learning is concepts, and the other 80% is practice.
As a result, you should expect to spend most of your time on consistent, dedicated, hands-on practice.
In this article, we’ll learn how to practice to become proficient with our AI skills.
The best way to begin practicing your AI skills is to start by solving toy problems. These are very simple problems that focus on teaching you a single concept or skill. The problem definitions are clear, the data are clean, and the feedback is quick. The goal is to learn as quickly as possible, so we want minimal complexity and distractions. Solving toy problems is like learning how to ride a bike with training wheels.
There are many great places to practice solving toy problems. For example, DataCamp and Codecademy provide simple interactive practice problems that allow you to learn a single skill or concept. Teachable Machine allows you to build simple AI/ML applications using a drag-and-drop user interface. Tensorflow Playground allows you to practice building neural networks while seeing their inner workings.
Once you’ve learned the basic concepts and skills, it’s time to take the training wheels off. A programming assignment involves a more complex problem that requires you to combine a few skills together to develop a solution. You will likely still be given clear instructions and clean data to start. However, it will require you to interpret the instructions and transform the data. Plus, the feedback loop is slower.
There are great resources available for practicing with programming assignments. These include several online learning platforms with interactive programming assignments and labs, like Pluralsight, Coursera, and edX. In addition, you can try the self-guided “Getting Started” projects on Kaggle or you can practice with coding katas from ML Katas and Codewars.
After you’ve mastered a few programming assignments, you will be ready for a hands-on project. A hands-on project simulates a real-world problem. It will require you to combine several skills together to solve a complex, multi-faceted problem. The data will be dirty, which will require both cleaning and transformation. In addition, the feedback loop will be much slower, and there may not be a single correct answer.
For hands-on projects, I recommend Kaggle competitions. These will provide you with all of the data, instructions, and help you need to practice solving real-world problems with AI. If you’d like more freedom, you can also use currated Kaggle datasets to apply your skills without instructions, guidance, or any specific goals.
One of the best ways to get experience solving real-world problems is by creating an open-source project. With an open-source project, you can solve a problem that is interesting to you. This will help you build your project portfolio and teach others while you learn. In addition, it will allow you to contribute to the open-source community. If you’re running into a specific problem, it’s likely others are too.
The best place to get started with an open-source project is GitHub. It is the most popular repository for free open-source software projects. Contributing to an open-source project will also teach you about software development workflows, code repositories, pull requests, collaborating with other open-source contributers, and other collaborative software-development skills.
Once you feel confident with your knowledge and skills, you’re ready to begin applying them on the job. Your goal with your first real-world projects is to find low-hanging fruit — problems that have a low cost/risk that will have high benefit/reward. A few quick wins can really help demonstrate the value of AI to your company and help you get buy-in from your team and manager.
The best way to start working on real-world projects in your company is to identify an ideal problem and get permission to try solving the problem with AI as an experiment. If you don’t currently have a job that allows you to use AI to solve real-world problems, you might want to consider joining a hackathon, getting an internship, or volunteering for a project that will allow you to solve problems with AI.
Finally, the AI industry is constantly change and it changes very fast. This means that you need to continuously invest in learning to keep up-to-date on the latest tools and technologies. To do this, I recommend creating rapid prototypes using new AI tools.
Building rapid prototypes allows you to quickly learn how to use a new tool without much additional overhead. Just use the tool to solve the simplest problem you can think of. Then, iteratively build in more complexity as necessary, and stop when you feel you’ve learned enough or you get bored.
To learn more, please check out my next article in this series: Getting Certified in AI.