Make IT Simple
Generative AI Award · Topic 1.3
A model that predicts an exam score from hours studied, taken through every stage of the machine learning pipeline — from framing the problem to reviewing the result. Step through it below.
A Visual & Mathematical Deep Dive into Multi-Layered Neural Networks
In the study of modern Artificial Intelligence—specifically within frameworks such as the EXIN BCS Generative AI Award syllabus—Deep Learning (DL) stands out as a critical subfield of Machine Learning. Unlike traditional rule-based or shallow statistical systems, deep learning utilizes multi-layered artificial neural networks capable of modeling complex, non-linear structures in massive datasets.
At its core, deep learning stacks multiple artificial neural layers between raw input data and final predictions:
To visualize signal propagation across multiple stacked hidden layers, here is a complete, self-contained HTML5 & JavaScript project that you can embed or run directly in CodePen:
If the forward pass is how a network makes a guess, backpropagation (backward propagation of errors) is how it learns from its mistakes.
Let's calculate one complete update step for a single neuron using real numbers:
Setup Parameters:
Step 1: Forward Pass & Loss Calculation
Step 2: Chain Rule Gradient Derivation
Step 3: Weight Adjustment
| Metric | Before Pass | After Pass (w = 0.8202) |
|---|---|---|
| Weight (w) | 0.8000 | 0.8202 |
| Prediction (ŷ) | 0.8455 | 0.8508 (Closer to target 1.0) |
| Loss (L) | 0.0119 | 0.0110 (Decreased) |
Learn how an instruction requests an output, why small wording changes can reshape a response, and when to use zero-shot, one-shot, few-shot, character and structured reasoning prompts.
Explain cloud computing.
Cloud computing provides computing services over a network.
A prompt is the instruction given by a user to a generative AI model. It tells the model what output is being requested. A prompt might ask for an explanation, summary, list, recommendation or creative response. The model uses the prompt as immediate context while its transformer also considers patterns learned from training data and the content generated so far.
The prompt starts the interaction. Without a useful instruction, the model does not know the task, subject or expected form of the response.
The transformer examines the prompt, connects it with learned patterns from training data, and considers each part of the response as it is generated. The prompt therefore guides predictions about what content should come next.
Prompt engineering is the art of altering and refining prompts to reach a desired or better-quality output. It is often iterative: the user reviews the response, notices what is missing or unsuitable, and rewrites the instruction with clearer direction.
“Explain cybersecurity.”
This requests an explanation but gives little direction about audience, length or focus.
“Explain cybersecurity to a beginner in three short paragraphs and include one everyday example.”
The refined wording gives the model a clearer target.
Explain artificial intelligence to a beginner using three short bullet points.
Prompt types differ mainly in the guidance supplied to the model. Zero-shot provides no example, one-shot provides one example, and few-shot provides several. Character prompts define a viewpoint, tone or style. Structured reasoning prompts ask the model to work through a complex problem in stages. Select each type to compare its purpose and construction.
| Prompt type | Examples supplied | Main use | Key distinction |
|---|---|---|---|
| Zero-shot | None | A direct, basic task | The model receives an instruction but no demonstration. |
| One-shot | One | Show the expected pattern once | One example guides the desired form. |
| Few-shot | Several | Make a pattern clearer | Multiple examples provide stronger guidance. |
| Character | Not defined by examples | Set tone, style or viewpoint | Uses characteristics such as a character, time period or location. |
| Structured reasoning | May vary | Handle a multi-level problem | Requests a staged, organised solution. |
Choose a prompt type, enter a task and generate a model prompt. The lab shows how the same request can be framed with different kinds of guidance.
A prompt requests an output and gives the transformer immediate context.
Prompt engineering alters wording and direction to seek a desired or better output.
Zero means none, one means one, and few means several examples.
Characteristics can be based on a character, period or geographical location.
A structured prompt breaks multi-level reasoning into organised stages.
Clearer prompts can improve relevance, but generated output must still be checked.
Follow the journey from large banks of training data to a generated response—and discover how testing, transformers and human feedback improve the system.
Learn patterns from vast data sets.
Evaluate with controlled, unseen data.
Learn from perceived output quality.
Use human checks and rewards.
Respond to new data and prompts.
A useful generative AI system is not created in a single step. It begins by learning from large quantities of data, is evaluated with information it has not previously seen, and is then improved through feedback. Only after the model has been trained and tested is it ready to receive new prompts and generate responses. Select each stage below to explore its purpose and place in the process.
Training gives the model its initial capability by exposing it to vast data sets. The model identifies patterns that can later be used to form new responses.
Testing uses controlled, unseen data. Because the examples were not used for training, they provide a more meaningful assessment of how the model handles unfamiliar input.
Inferencing occurs when a user supplies new data and a prompt to the trained and tested model. The output may be a prediction, recommendation or another generated response.
| STAGE | MAIN INPUT | WHAT HAPPENS | MAIN PURPOSE |
|---|---|---|---|
| Training | Vast data sets | The model learns patterns from information. | Develop the model's initial capability. |
| Testing | Controlled, unseen data | The model's responses and performance are assessed. | Evaluate accuracy and performance. |
| Reinforcement learning | Information about output quality | The model learns which kinds of responses are considered better. | Improve future outputs. |
| RLHF | Prompts, human checks and rewards | Human operators review responses and reward correct ones. | Use human judgment to guide improvement. |
| Inferencing | New data and a prompt | The trained and tested model generates a response. | Put the model into practical use. |
Data is the foundation of generative AI. The model depends on training data to learn patterns and on separate test data to evaluate whether that learning can be applied successfully. The two data types have different purposes and should not be confused.
Training data supplies the enormous banks of information from which the AI model learns. Rather than storing a ready-made response for every possible prompt, the model learns patterns that help it construct a suitable response when prompted.
Test data is a separate bank of controlled information used to assess the model after training. It provides a fairer evaluation because the model has not previously learned from these examples.
If the same information were used for both training and testing, the evaluation would not clearly show how well the model responds to genuinely new examples. Keeping test data separate allows performance to be assessed using information that did not contribute to the model's learning.
Simple analogy: training data is similar to the material a candidate studies, while test data is similar to new questions used to assess whether the candidate understood the subject. Studying the exact test answers would weaken the value of the assessment.
| DATA TYPE | WHEN USED | ROLE | KEY EXAM PHRASE |
|---|---|---|---|
| Pre-training data | At the beginning | Provides the first broad bank of information. | Before refinement or fine-tuning |
| Training data | During model development | Teaches and further focuses the model. | Used to train the model |
| Test data | After or apart from training | Assesses model performance and output. | Controlled and unseen |
A transformer is a deep learning architecture that processes data in parallel using "attention mechanisms" to understand context and relationships between words.
Transformers are central to the way modern generative AI constructs language. They improve the model's ability to predict what should come next in a response—not only the next word, but also likely phrases, sentences and paragraphs. Repeating this prediction process allows the model to produce connected responses that may run into thousands of words.
A transformer uses the prompt and the response produced so far as context. It considers possible continuations, identifies a likely next element and then repeats the operation as the response grows.
The prediction is influenced by the user's prompt and by the words already generated in the response.
The transformer helps estimate which word, phrase or sentence is most likely to follow in that context.
The newly selected content becomes part of the context, and the prediction process repeats to form a longer response.
A long response is not normally produced as one indivisible block. It is developed through continuing predictions, with each new part building on what came before. This continuing use of context enables responses containing many connected sentences and paragraphs.
However, the transformer is predicting plausible continuations; it is not guaranteeing that every statement is factually correct. Therefore, response length, detail and confident wording should never be treated as automatic evidence of accuracy.
Initial training gives a model broad capability, but its responses still need refinement. Human feedback helps demonstrate what a desired response looks like and which model-generated responses are judged correct. The syllabus identifies two related but distinct approaches: supervised fine-tuning and reinforcement learning from human feedback.
| METHOD | WHAT THE HUMAN DOES | WHAT THE MODEL RECEIVES | MAIN IDEA |
|---|---|---|---|
| SFT | Creates the desired response to a prompt. | A prompt paired with a high-quality human-created response as training data. | Show the model what a desired answer looks like. |
| RLHF | Poses prompts, checks model responses and rewards correct responses. | Feedback based on human judgment of response quality. | Reinforce responses that humans consider correct. |
A human writes a clear desired answer for a prompt. The prompt and answer are then used as a training example for the model.
The model answers a prompt. A human checks that response and provides a reward when the response is considered correct.
Feedback and fine-tuning are ongoing. Repeated adjustment explains why users may observe continuing improvement in generative AI responses.
Training data teaches patterns; unseen test data assesses performance.
SFT supplies a human-created desired response; RLHF rewards good responses.
Training develops the model; inferencing uses it with new data.
Transformers enable long responses, but those responses can still be wrong.