How does Artificial Intelligence actually work?
If you were to look up AI on google you will give you a sentence like, “Data is trained on some mathematical models and algorithms, which then try to find patterns and make predictions.” While this is technically correct, I plan to give you more context and explain how AI really works through this article.
In this blog, I aim to cover various AI technologies, companies, and explain algorithms in a way that's easy to understand. I’ll break down the steps that go into building AI at a high level. Though the steps I mention are general, I plan to dive into more detail over time.
How Data plays a part in Artificial Intelligence?
Without data, there is no Artificial Intelligence. The first step is getting the data, but it needs to be clean before we can use it effectively. Let's take an example:
As you can see, there are a few issues with the dataset above. Some values are missing, the Customer ID column has text where a number should be, and both the Date of Birth and Account Created columns have different date formats. In the Email column, there are errors like double “@” symbols and missing ".com" extensions.
Why can’t we use this dirty data to build an AI?
It’s simple. Imagine when you were a child, someone taught you that “lying and cheating is a good thing.” As a child, you might believe this without questioning it, and you would start lying and cheating because you thought it was correct. In this scenario:
Data Input: "Lying and cheating is a good thing."
Data Output: You start lying and cheating other people.
See the issue? If we are taught wrong things, we act in the wrong way. The same principle applies to AI. If we train an AI model with dirty data, it will learn incorrect patterns and make flawed predictions.
For example, if a numeric column contains text (like "Ninety" instead of a number), the model may crash or misinterpret the values. Similarly, missing values can lead to inconsistent and biased predictions.
This is why data cleaning is essential. The data must be consistent before we move on to building an AI model. There are additional steps after cleaning the data, but I won’t get too technical in this article. I'll cover those details in future posts.
Selecting a model
Choosing the right model depends on the type of data you have. Machine learning models are generally classified into three categories:
Supervised Learning
Unsupervised Learning
Reinforcement Learning
Selecting a model depends on the data and the problem you're trying to solve. This topic deserves its own article, which I plan to write, but for now, just know that the choice of model is critical to the success of the AI system.
Training the model
Next, we train the model using the cleaned data. Typically, the dataset is split into two parts, such as 70% for training and 30% for testing (or 80%/20%). The first part is used to train the model so it can learn to recognize patterns on its own. As training progresses, the model becomes more accurate in detecting patterns and making predictions.
The remaining data is used to test the model's accuracy, ensuring it performs well on unseen data. How exactly does this work? Don’t worry, I’ll cover these steps in more detail in future articles as the blog gets more technical.
Model Deployment
Finally, we deploy the model into a production environment where it can be used by end-users or integrated into systems.
This was a high-level overview of how AI works. We will dive deeper into each of these topics as the blog progresses. My goal is to explain AI concepts in a simple, easy-to-understand way, and I look forward to sharing more detailed insights in future articles.
