Cracking the AI & Machine Learning Code: Essential Insights for Beginners (Part 1)

If you want to understand AI and machine learning, but don’t know where to start, I’ll take you through resources and paths I’ve found helpful to have a better understanding of these topics. Usually people want to dive right into understanding what a large language model is. Unfortunately, that’s a backwards way of learning because you end up skipping a lot of the basics.?Large language models were built on foundational machine learning tools; so you'll need to start at the foundation first.

Take Andrew Ng’s Machine Learning Course on Coursera

My first piece of advice if you want to better understand AI and machine learning is to start with the machine learning course by Andrew Ng on Coursera. Yeah, it’s several years old, but it’s still the best resource I’ve seen. Even if you don’t have a math background, it’s still understandable. It will take you a week or two to get through his full machine learning specialization and do all the exercise. Most importantly, he teaches what’s behind the math which is important so that you don’t just learn the tools and hit a “button” without knowing what it is doing exactly.?

Experiment with Data Sets and Tools

Now, it’s time to start experimenting with real world data and tools. I find it’s more intuitive to start with supervised learning problems, because they tend to map to questions we have in our daily lives (i.e. how much will this type of house cost? What are the factors that most predict whether someone will have breast cancer?)?

Start with a regression problem. You can find a good data set on predicting housing prices on Kaggle here: housing price data set. (If you’re not familiar with Kaggle, it’s a repository for datasets and machine learning competitions). Download the dataset and then use a Jupyter notebook to load the data, create your Python code to see how well your model can predict a house’s price. (Hint: If you need help writing the code, you can ask ChatGPT- I’ll write in a future post best practices on how to do this.)

After you feel comfortable with regression, you can try classification problems. Kaggle has a good data set on Titanic survivors that fits this. You can find it here. Kaggle provides a whole tutorial on how to load and submit your code.??

Use ChatGPT or Gemini as a Mentor

When you get stuck on any AI or machine learning terminology that you don’t understand, you can always use ChatGPT or Gemini to clarify. For example, if you don’t understand what hyperparameter tuning means, you can always ask either ChatGPT or Gemini: “Explain what hyperparameter tuning means. Use language a 6th grader would understand”.

You can also follow up this prompt with something more specific such as “how would I test hyperparameter tuning for a machine learning model in a jupyter notebook? Explain this in language a 12th grader would understand.” The key to effectively using ChatGPT or Gemini to help with your machine learning is to type exactly what you want to know, how you want to know it (do you want to see code, an explanation or both) and at what level of AI / machine learning fluency.?

I find ChatGPT and Gemini give pretty different answers to the same questions. So it's useful to query both and see how they frame it. When it comes to writing actual machine learning code, I find ChatGPT to be far superior!

I’ll follow up in an upcoming post with exactly how to use ChatGPT to write your machine learning code so that you can focus on honing your machine learning skills and not your Python syntax.