- Teacher's AIed
- Posts
- How It's Made: Generative AI (Educator Edition)
How It's Made: Generative AI (Educator Edition)
Estimated Read Time: 3 min 25 sec
Teach with expert insights on AI, curated by your trusty Teacher’s AIde

Hello! Welcome back to Teacher's AIed, where we simplify AI in Education concepts more smoothly than high school students ask their crush to be their valentine…actually, we are way smoother than that. High schoolers aren’t nearly as smooth as they think they are.

Created by DALLE
Earlier this week, I shared a few analogies that you can use to explain what Generative AI can do for you. Today, I want to share a taste of how it works.
Full transparency - I am not an expert in AI technologies. I rest my laurels in the Education camp. However, for what it is worth, I did take a graduate-level machine learning class in college. (I like to think I know enough to be dangerous.)

In sum, Generative AI tools require demystification. While they truly do seem magical with their ability to produce paragraphs and paintings at the push of a key, in the end, their language is of zeros and ones, like all computers. Let’s do a crash course in the nuts and bolts of Generative AI.
Here's what we have for you today
1. The Cupcake Model of GenAI
The Cupcake Model of GenAI
So, how does Generative AI go from prompts to Picasso or Plato?
To aid in explaining AI, I searched for an easy-to-understand yet accurate framework.
Much of what I found was either too complex or not sufficient.
Eventually, I came across a Medium blog post that offered a simplified model of the basic construction of Generative AI. It’s called the cupcake model:
This diagram shows the handful of different processes that happen between an input and an output. After a prompt is provided, the GPT preprocesses it and feeds that output into a neural network trained on vast bodies of data. The output of that step is then fine-tuned and evaluated.
I recommend that you read this short article as well. In my opinion, the author sufficiently explains — or at least sufficiently enough for an educator’s purposes — all of the components of this process except for the Neural Network stage.
Let me take a stab at explaining Neural Networks.

2. Neural Networks - Strangely Human
Neural Networks are truly fascinating! We, as educators, talk about helping students make sense of what they know and reflect on their own learning. This is exactly what Neural Networks do.
Glancing over the topic is understandable as Neural Networks are complex (i.e., what I learned in a graduate-level computer science class…this isn’t Coding 101).

Created by DALLE
However, allow me to share a bit more nuance to the neural network component of the Generative AI process in as plain English as possible.
When I think of Neural Networks, I think of the word “prediction.”
As an educator, this is a process you are quite familiar with. Students are making predictions as well. To give a literacy example, after looking at a text, students predict what letter is printed, what word those letters make up, the tone of a sentence, and the main idea of a paragraph.
We learn through recognizing patters and making predictions based on those patterns.
What makes neural networks so powerful is that they mimic the same prediction algorithms that happen in our brain’s neural networks.
What makes Generative AI - specifically LLMs like ChatGPT different - is their ability to predict not just individual words, but entire sentences and paragraphs.
This is about as far as I can bring you over email. If you’re hungry to learn more about neural networks, a well-scaffolded video is this one from Wired. AKA, it starts with explaining machine learning to an 8-year-old. If you’re ready for even more, I recommend this video for an in-depth (read: mathematical) explanation. You can even email Lewis at [email protected] if you want to continue the conversation!

How these tools work shouldn’t be kept a mystery from students. I can think of a few reasons as to why, but I’ll limit myself to just one. It’s vitally important for students to see these tools as creations of other humans. Students should feel inspired that engineers are capable of building something so powerful it feels strangely human.
If you agree that this type of knowledge should be shared with students - in appropriate and intelligible ways — consider subscribing to our newsletter. Let Teacher’s AIed help you teach with expert insights on AI. If you are already subscribed, consider sharing this post with a colleague who might be interested.

If you’re not already subscribed to Premium, make sure to subscribe to get access to all our pieces!

Class dismissed!
Lewis Poche & Kourtney Bradshaw-Clay
Reply