Why AI image generation creates visual content in Prompt Engineering / GenAI - Challenge Your Understanding

Challenge - 5 Problems

🧠 Conceptual

intermediate

How does AI image generation learn to create pictures?

AI image generation models create pictures by learning from many examples. What is the main way these models learn to produce new images?

A By randomly guessing pixels until an image looks good

B By memorizing every image exactly and copying them

C By using a fixed set of pre-drawn images to combine

D By learning patterns and features from many images to create new ones

💡 Hint

Think about how learning from examples helps create something new, not just copying.

AI image generation models learn patterns and features from many images. They do not memorize or copy exactly but create new images by combining learned features.

❓ Predict Output

intermediate

Output of a simple AI image generation step

Given this simplified code snippet that generates a 2x2 pixel image with random colors, what is the shape of the output array?

Prompt Engineering / GenAI

import numpy as np
image = np.random.rand(2, 2, 3)
print(image.shape)

A (2, 2, 3)

B (3, 2, 2)

C (2, 3, 2)

D (3, 3, 3)

💡 Hint

Remember the shape format for images is height, width, and color channels.

The output shape (2, 2, 3) means 2 pixels height, 2 pixels width, and 3 color channels (RGB).

❓ Model Choice

advanced

Choosing the right AI model for image generation

You want to generate realistic images from text descriptions. Which AI model type is best suited for this task?

A Convolutional Neural Network (CNN) for image classification

B Generative Adversarial Network (GAN) trained on image-text pairs

C Recurrent Neural Network (RNN) for sequence prediction

D Decision Tree for image segmentation

💡 Hint

Think about models that can create new images and learn from text descriptions.

GANs trained on image-text pairs can generate realistic images from text prompts by learning to create images that match descriptions.

❓ Metrics

advanced

Evaluating AI image generation quality

Which metric is commonly used to measure how realistic AI-generated images are compared to real images?

A Inception Score (IS)

B Precision

C Accuracy

D Mean Squared Error (MSE)

💡 Hint

This metric uses a pre-trained image classifier to evaluate generated images.

Inception Score measures the quality and diversity of generated images using a pre-trained classifier, making it popular for image generation evaluation.

🔧 Debug

expert

Debugging a failed AI image generation training

During training of an AI image generator, the loss stays very high and images look like noise. Which issue is most likely causing this?

A The model has too few layers to learn features

B The dataset images are too small in size

C The learning rate is too high, causing unstable training

D The optimizer is set to 'adam' instead of 'sgd'

💡 Hint

Think about what causes training to fail to improve and produce noise.

A very high learning rate can cause the model to jump around the solution space and never converge, resulting in high loss and noisy images.