Teach with expert insights on AI, curated by your trusty Teacher’s AIde

Welcome back to Teacher's AIed, where once upon a time, we might have had the slightest desire for our students to turn back into pumpkins at midnight and show up to school the next day…completely stationary.

In an age dominated by artificial intelligence (AI), it's not just students who need to grasp the fundamental concepts of how computers perceive the world. Educators, and indeed anyone living in this AI era, can benefit from understanding these principles.

This is the fourth installment in my series on AI literacy and its implications in K-12 Education.

Did you miss the previous posts? Here are the links:

• What Is AI Literacy? Part 1

• Are you smarter than an AI Literate 5th grader?

• AI Literacy: The Lay of the Land

In today’s post, I explore AI4K12’s third Big Idea - "Learning.”

At first glance, the notion that computers learn from data seems straightforward—after all, isn't all learning all about taking in information?

Yet, delving deeper reveals a complex narrative, rich with questions about the nature of learning across beings of silicon and flesh.

What types of information do computers learn from versus humans?

Do humans use the same statistical models that computers rely on for learning?

These are some of the questions we'll delve into.

1. Nature of Learning: More than Just Trial and Error

Once upon a time, your child, led by innate curiosity, approached a glowing stove, drawn by its warm and inviting light. As if in slow motion, you see the child reach out, fingers inching closer, until suddenly—ouch!

You can feel it, too, as you try to console your kiddo.

This is a quintessential - and a bit dramatic - example of one way we learn a lesson. The sharp pain of touching the hot stove etches a vivid memory in the child's mind. That child surely learned a lesson.

Let’s analyze the different ways your child might have learned the lesson.

1. The child learned from a personal, sensory experience.

2. The child also learned from the social world around them, observing others' mistakes and heeding warnings from caregivers.

3. Lastly, through abstract reasoning, the child understood that other objects emitting heat, like a hot iron, could also cause harm.

Let’s contrast this with the “silicon learner,” navigating its path through algorithms and data.

There are a few different ways that a computer could “learn” that a glowing stove is hot and, therefore, shouldn’t be touched. Three methods to note are supervised learning, unsupervised learning, and reinforcement learning. Yet, all three of these are just a fraction of how the human child learned this lesson.

Computers learn in less flexible, creative, and interconnected ways than humans.

Yet, as we edge closer to the dawn of Artificial General Intelligence (AGI), these distinctions might begin to blur, opening a dialogue between the learning styles of man and machine.

2. Neural Networks: a shadow of a human brain

Neural networks, with their web of interconnected nodes, present a mirror, however dimly, to the complexity of the human brain.

For a more in-depth explanation of how neural networks are designed, take a look at our blog post titled: “How It's Made: Generative AI (Educator Edition).” In this post, I explain a neural network’s basic structure using the cupcake model.

Neural Networks, in many senses, are the bedrock of all AI tools. They are flexible data structures that can be utilized in a variety of contexts. In my opinion, understanding these concepts can serve as a bit of a litmus test for the design component of AI Literacy.

3. Datasets: Big or Small - I like them all

Data, in the realm of machine learning, is power.

The “better” your data, the better your model. We know this concept first hand while writing prompts for tools like ChatGPT. The quality of the information that you put in dictates the quality of the information you get out.

Well, this is at least true most of the time for data sets. Let me explain.

We can split data sets into two categories: Feature sets and Large sets.

Feature sets are laser-focused and come pre-designed for a specific task. For example, if you wanted to create a data set to create a “is the oven hot” classification tool, I’d want a data set that focuses on the relevant pieces of information: the color of the element, the position of the dial, the presence of a pot of boiling water, etc. You as the data engineer shrink the data set down just to relevant features.

Big data on the other hand does not shrink down the data. Essentially anything and everything might get included in the model. The algorithm can decide if the oven light being on or if the time is 8:42 has anything to do with stove top being hot.

In conclusion, how computers learn is a key concept in AI that educators need to understand and teach. By exploring these concepts, educators can help students develop the skills and knowledge necessary to navigate a world shaped by AI.

Stay tuned for more in this series as we continue to unravel the intricacies of AI Literacy and its importance in education and beyond!

Class dismissed!

Lewis Poche & Kourtney Bradshaw-Clay