Artificial intelligence systems are built through a structured process involving data collection, training, and deployment to perform tasks traditionally requiring human intelligence. A simple example, such as creating a cat detector, demonstrates how AI models learn from data to identify patterns and make accurate predictions.
Key Insights
- AI models learn to detect patterns, such as identifying cats in images, by analyzing thousands of labeled examples during the training phase.
- The AI pipeline includes sequential stages such as data collection, model training, evaluation, deployment, and ongoing monitoring to ensure reliable performance.
- Generative AI, a form of deep learning, represents an advanced subset of AI capable of handling complex tasks and interacting conversationally with users.
This lesson is a preview from our AI Fundamentals for Government Employees Course. Enroll in a course for detailed lessons, live instructor support, and project-based training.
To show you how AI is used to create something useful to us, let's turn to the internet's finest use case, cats. Let's build a cat detector. So, we have a problem.
We have a bunch of images, and we want to know, is this a cat or some other type of animal? Well, to build an AI model that would be a cat detector, we would first need to feed it a bunch of examples to learn from. We'd feed it thousands of images, some of them with cats and some of them without cats. We would then select a process based on the expected level of human involvement, training time, and accuracy of the output expected.
The AI would then be trained. Off in a neural network, this AI model would look at cats and non-cat images. It would adjust its weights internally like virtual knobs to get a better telling of what's a cat and what's not a cat.
And after all that training, we would now have a workable model. Basically, this would be the AI's brain that has learned what a cat looks like. It's our cat detector.
We could now give the cat a new unseen photo, maybe of our cat at home. The model would process what it's learned and make a prediction. This is a cat, or it's not a cat.
And so, the AI model that we just built doesn't know what a cat is in the same way that we do. Instead, it's learned from all the data and all the images that we found and is able to apply it to new situations. So, it can look at pictures like these and identify with a high degree of certainty that this is, in fact, a cat and this is not.
Our cat detector example went through this pretty quickly, but now we'll review this process in more detail so that you better understand how the AI assistant you're interacting with came to be and how it works. The AI pipeline is something that refers to the structured sequence of steps or stages that are followed to build, deploy, and maintain an AI model or system. This process typically involves various phases from data collection to model training and evaluation through to deployment and ongoing monitoring.
Each stage is important in ensuring that an AI system performs well and meets the intended goals. Now, you won't typically interact with an AI until after the model's been deployed and it's in its maintenance or optimization stage. But again, we're going to go through this whole process to get a better understanding of how AI works to help you appreciate why you may want to take certain precautions when using generative AI later.
For the purposes of this fundamentals course, this visual by the Government Accountability Office is a helpful overview of the hierarchy of AI. Keep this visual in mind as we progress through the module. So again, AI is that top-level broad term for any system doing a job that typically or formally would have used human intelligence.
Machine learning is a subset of that, we could call it. It's a form of AI that specializes in a specific task. Deep learning is a subset of machine learning that gets even deeper, hence the name, with more complex tasks.
And that leads us to generative AI, which is a super-intelligent instance of deep learning that's so smart that users can converse with it. This type of AI is what we'll be most focused on in this course.
