Simple AI guide for eLearning. Everything you need to know

After reading this article, you’ll understand AI, machine learning, and deep learning. You’ll know how to talk to your colleagues about it, what types of machine learning models exist, and how AI can beneficially augment EdTech. Let’s get started.

Many people are worried that AI will take their jobs. The fears are valid. According to Goldman Sachs, over 300 million jobs can be displaced worldwide by 2030. In the education sector, 27% of workers are exposed.

However, AI is unlikely to replace teachers, despite the risk of automation growing from 9% in 2023 to 11% in 2025.

What can educators do to pull through? There are two approaches: try to ban AI or embrace it. Historically, technology bans are shortsighted. Just look at how it’s worked for the Luddites.

The good news is that technology will not only automate old jobs but create new occupations as well. So far, technology has created more professions than it has taken away. Today, 60% of workers have jobs that didn’t exist before 1940.

In any case, drastic changes are unlikely anytime soon.

Common limitations of AI language models

You can’t trust AI content. Language models like ChatGPT generate text that looks right, but it doesn’t mean that it’s correct. They simply choose appropriate words one by one, using statistics. They don’t understand the meaning behind the words.

It’s still banned in many schools. Many school districts prohibit AI for students and teachers. While some places like NY have already lifted the ban, the presence of AI in education is still debatable. Some look forward to the future with AI, while others steer clear of it.

It’s not as smart as you’d think. According to recent Apple insights, current LLMs can’t truly reason; they just copy reasoning patterns from their training data. Therefore, chatbots’ answers may be unreliable for tasks that require true understanding and logical reasoning.

In addition, a single ChatGPT query consumes enough electricity to power a light bulb for 20 minutes, which raises the question of the energy intensity of using AI.

The list of concerns is usually supplemented by:

• Reducing maintenance costs;
• Keeping it unbiased;
• User privacy protection;
• AI’s decision-making control.

As a result, many ethical considerations surround AI integrations – we have previously covered them in dedicated material. Consequently, models like ChatGPT may not be used to their full potential until we better understand AI.

Still, many tools are already being actively used in education – let’s look at them in detail.

Optimize administrative tasks

Not surprisingly, computers handle statistics much better than humans.

Neural networks can detect patterns that we might not notice. Imagine two playful pairs in class: Alex and John, and Mia and Sophia. Everything is fine when they’re separate, but when the four come together, chaos erupts, and they start disrupting their classmates.

In any complex system, there are hidden patterns. AI can untangle the patterns to simplify operational tasks:

• Scheduling;
• Classroom and locker layouts;
• Staff allocation;
• Enrollment;
• Transportation planning;
• Event scheduling.

Simplify assessments

AI makes grading faster. It can zip through complex exams while understanding student intent. This will free up educators to focus on more engaging tasks. In some cases, teachers need to step in, and AI can help find such situations.

Also, with ChatGPT students got a new way to cheat. How can we ensure that learners do their own homework? It’s simple: just use another AI to identify plagiarism.

A special report by Microsoft states that plagiarism and cheating are top concerns for teachers (42%) and educational authorities (24%).

However, modern AI-driven systems can both prevent falsification and assess knowledge more comprehensively – not to mention saving teachers time in creating tests and grading papers.

Assist educational processes

AI chatbots can answer any question in a snap (granted, they’re not always correct yet). They save time but also make students feel supported. Some students use ChatGPT in English classes to translate Shakespeare’s tragedies into modern English.

An AI teacher assistant can make classes more fun. One student generated a Kanye West-style rap about trigonometry.

Artificial intelligence (AI) is a technology that helps machines mimic human intelligence. AI systems can see, learn, solve problems, make decisions, etc.

Today’s narrowly focused AI is good at specific tasks, but it lacks broad cognition and consciousness of general AI. It can play chess, generate text, or forecast the weather. However, it can’t deal with the full range of human tasks.

Machine Learning (ML) is a subset of AI that relies on adaptive learning. Machine learning behavior is not explicitly programmed; the algorithms learn autonomously. That’s why such models can get better over time and not stay rigid.

By contrast, non-ML AI models are rigid and rule-based. When a computer can’t execute a line of code, it freezes, not knowing what to do. If operating systems could handle unconventional situations with ML, we wouldn’t have “blue screens of death”.

Most modern AI systems apply at least some elements of ML. That’s why the terms are often used interchangeably.

But you can still find non-ML AI models in fields like robotics, speech, or image recognition.

Deep learning (DL) is a subset of ML that uses neural networks. A neural network is a set of different formulas mimicking our brain’s work.

To train the model, developers upload real-world input and output data. The AI attempts to find patterns or connections between them. Initially, these connections are random and may be incorrect.

However, as the training progresses, the connections become more accurate with each step.

For example, you want to create a neural network that predicts exam grades.

First, you need a random hypothesis. Does perfect attendance equal perfect test scores? The model tests it and compares the results to real-world data. If the results were not correct, the neural network would work the math magic to come up with better predictors of test scores.

This process will take many iterations until the developers are satisfied with the accuracy.
The more diverse your input data, the more accurate the results. To estimate exam scores, you may need to collect more than just attendance. It could be student grade history or time spent on finishing homework.

To learn more about these concepts, take a look at the dedicated material where we cover them more thoroughly.

Types of machine learning models

There are plenty of different models, but we’ll focus on the high-level types.

1. Supervised learning models

To train such a model, developers feed it with extensive input and output data. As the AI identifies patterns, it can predict outcomes of new data.

There are two main types of supervised models:

• Regression models find connections between variables.
  You can visualize such connections in a chart with X and Y axes, as in the picture above. If 1 cup of coffee costs 5 dollars, you can predict the price of 10 cups. That’s called linear regression.
• Classification models sort items into categories.
  Imagine a full fruit bowl. If I ask you to pick all the bananas out, you’ll instinctively look at the shape, color, and so on. To people, classification is simple and automatic. We don’t need to have seen a billion bananas to tell them apart. Computers do, and that’s why we still have CAPTCHAs (although bots have become better at it).

Supervised models are great when you have lots of labeled data. They are perfect for content recommendation systems, automated grading, and language learning tools.

2. Semi-supervised learning models

Semi-supervised learning sits between supervised and unsupervised learning. Developers use a small amount of labeled data combined with a large amount of unlabeled data to train the model. The AI leverages the labeled data to make educated guesses about the unlabeled data.

Two main approaches in semi-supervised models include:

• In self-training, the model initially trains on the small labeled dataset and then starts predicting labels for the unlabeled data. These predictions are added to the training set, allowing the model to refine its understanding with more data.
• In co-training, two separate models are trained on different subsets of the data. Each model labels the unlabeled data for the other, improving accuracy over time.

Semi-supervised models are ideal when acquiring labeled data is expensive or time-consuming, but plenty of unlabeled data is available.

3. Unsupervised learning models

Such a model doesn’t need human guidance. Developers feed AI with large datasets and let it find patterns on its own.

• Clustering is a popular type of unsupervised learning. It’s similar to classification, except the data is unlabeled. Instead of looking for bananas in the fruit bowl, you’d separate exotic fruits – even if you don’t know what they are.
• Dimensionality reduction diminishes the number of features in a dataset while retaining important information. It helps to simplify data and improve the efficiency of other ML models.
• Anomaly detection identifies rare or unusual data points that deviate significantly from the normal patterns. It’s useful for fraud detection, network security, and identifying defects in manufacturing processes.

Use such models when you need to identify trends, organize content, or cluster students.

4. Reinforcement learning models

These models are trained with feedback. The algorithm interacts with the environment and receives feedback – either rewards or penalties. The AI learns to maximize actions for long-term rewards.

Back to the fruit bowl example. Imagine you’re trying to pick a papaya, but you don’t know what it looks like. Each time you pick the right fruit, you get a reward, like a small treat. If you pick the wrong fruit, you get a gentle reminder to try again. That’s reinforcement learning.

Use reinforcement learning models when you have a dynamic environment. With such models, you can simulate physics experiments, create adaptive learning paths, or power your interactive learning games.

By the way, being ML’s subset, deep learning differs from it a bit. We have covered the distinctive features in this material.