Making images with AI quiz

Question 1

What is a pixel, and why does it matter in AI image generation?

Accepted Answer

The smallest colored unit in a digital image

Answer

A type of AI model used specifically for image creation

Answer

A large training dataset fed into a neural network

Answer

A type of camera lens filter that sharpens photos

Question 2

What does GAN stand for, and what is its basic idea?

Accepted Answer

A pair of competing neural networks that work together to generate realistic images

Answer

A professional graphics editing tool similar to Photoshop

Answer

A type of image compression filter

Answer

A large labeled database of real photographs

Question 3

In a GAN, what is the job of the 'generator'?

Accepted Answer

Creates new images from random noise, improving over training

Answer

Stores real images in a database for the model to look up

Answer

Decides whether a given image was taken by a camera or not

Answer

Compresses finished images into smaller file sizes

Question 4

In a GAN, what is the job of the 'discriminator'?

Accepted Answer

Judges whether an image is real or was made by the generator

Answer

Generates random noise that the generator uses as input

Answer

Decides what colors to paint each pixel in a new image

Answer

Downloads and organises training images from the internet

Question 5

What is a 'prompt' in a text-to-image AI system?

Accepted Answer

A written text description that tells the AI what image to create

Answer

A type of digital brush used in AI painting software

Answer

A training image that is fed directly into the model

Answer

A button that resets the AI back to its starting state

Question 6

How does a diffusion model generate an image?

Accepted Answer

It learns to reverse a step-by-step noising process, starting from static to reveal an image

Answer

It spreads copies of an image across many servers simultaneously

Answer

It applies the same blur filter to every training image

Answer

It resizes and crops images according to user settings

Question 7

What is 'training data' and why does an AI image generator need so much of it?

Accepted Answer

The large set of example images the AI studies to learn visual patterns

Answer

Images that the AI produces after training is finished

Answer

The programming code that runs the AI model

Answer

The server hardware on which the AI model is stored

Question 8

What is a neural network, in simple terms?

Accepted Answer

A layered computer system of connected math units that learns from examples

Answer

A type of internet network used for transferring images quickly

Answer

A social network that connects AI researchers worldwide

Answer

A physical printing device that produces high-quality images

Question 9

What is a Convolutional Neural Network (CNN) and why is it good at understanding images?

Accepted Answer

A neural network that scans image patches to detect visual features like edges and shapes

Answer

A network that only works with text and cannot process visual data

Answer

A network that links multiple security cameras together

Answer

A network used exclusively for generating video-game graphics

Question 10

What is 'latent space' in an AI image generator?

Accepted Answer

A compressed mathematical space where image features are stored as numbers

Answer

The physical hard drive where finished generated images are saved

Answer

The color palette the AI chooses for each image

Answer

The internet bandwidth used while the image is being generated

Question 11

What is image-to-image (img2img) generation?

Accepted Answer

Using an existing image as a structural guide for the AI to create a modified or stylised version

Answer

Copying an image from one folder to another without any changes

Answer

Scanning a printed photograph with a digital camera

Answer

Converting an image from JPEG format to PNG format

Question 12

What is neural style transfer?

Accepted Answer

Using AI to paint the content of one image in the visual style of another

Answer

Moving an image file from one storage folder to another

Answer

Changing the resolution or dimensions of an existing image

Answer

Training an AI to name every object it spots in a photograph

Question 13

What is a Variational Autoencoder (VAE) and how does it relate to image generation?

Accepted Answer

A model that compresses images into latent space and can generate new images by decoding points in that space

Answer

A tool for displaying images in a virtual reality headset

Answer

A video-and-audio encoder used in film production

Answer

A tool that automatically validates whether a printed image meets quality standards

Question 14

What does 'noise' mean in the context of a diffusion model?

Accepted Answer

Random pixel variations that are added to images during training and removed during generation

Answer

Unwanted sound picked up by a microphone attached to the AI system

Answer

Bugs in the Python code that runs the diffusion model

Answer

Slow internet connection that interrupts the image download

Question 15

What is CLIP and how does it help text-to-image AI understand your prompt?

Accepted Answer

A model trained to match text descriptions to images, used to guide image generation from prompts

Answer

A tool that clips and auto-crops images into a square shape

Answer

A file compression algorithm for storing large image datasets

Answer

A browser extension for saving images from any website

Question 16

What is an 'embedding' in AI image systems?

Accepted Answer

A list of numbers that represents a word or image so AI can process it mathematically

Answer

Adding a visible watermark to an AI-generated image for copyright protection

Answer

Inserting an image into a word-processing document

Answer

Saving an image in a lossless compressed file format

Question 17

What is 'fine-tuning' an AI image model?

Accepted Answer

Continuing to train an already-trained model on a smaller, specialised dataset

Answer

Adjusting the brightness, contrast, and sharpness of an image by hand

Answer

Fixing bugs in the neural network's source code

Answer

Manually selecting the sharpest images from a batch of generated outputs

Question 18

What is a 'deepfake'?

Accepted Answer

An AI-generated image or video that falsely shows a real person saying or doing something they never did

Answer

A high-resolution underwater photograph taken at great depth

Answer

A depth-of-field camera filter that blurs the background of a portrait

Answer

A secure encrypted backup of important personal photos

Question 19

Why are deepfakes considered ethically harmful?

Accepted Answer

They can spread false information and seriously harm real people's reputations and safety

Answer

They consume too much computer memory and slow down devices

Answer

They make generated images look unrealistically colorful

Answer

They slow down internet speeds for all users on a network

Question 20

What is AI image 'inpainting'?

Accepted Answer

AI intelligently filling in a masked or removed section of an image so it looks natural

Answer

Physically painting on a canvas using a robot arm guided by AI

Answer

Applying a uniform color filter across an entire image

Answer

Adding a decorative frame or border around a photograph

Question 21

What is AI image upscaling (super-resolution)?

Accepted Answer

Using AI to increase image resolution by predicting and adding realistic fine detail

Answer

Reducing an image's file size so it loads faster on a website

Answer

Rotating an image to the correct upright orientation automatically

Answer

Converting a color image to black and white

Question 22

What is a 'checkpoint' in AI model training, and why is it useful?

Accepted Answer

A saved snapshot of the model's learned weights at a particular point during training

Answer

A quality check where a human reviewer rates each generated image

Answer

A security password needed to access the AI model online

Answer

A special image file format designed for AI training pipelines

Question 23

What is the difference between image classification and image generation?

Accepted Answer

Classification identifies what is in an existing image; generation creates entirely new images

Answer

Classification makes images bigger; generation makes them smaller

Answer

Classification works only on videos; generation works only on still photos

Answer

They are the same process but run at different hardware speeds

Question 24

What is data augmentation and why do AI image trainers use it?

Accepted Answer

Creating extra training examples by applying random flips, crops, and brightness changes to existing images

Answer

Compressing training images to reduce storage requirements on a server

Answer

Automatically downloading more images from the internet to expand the dataset

Answer

Manually labelling each training image with a text description

Question 25

What is 'overfitting' in an AI image generation model?

Accepted Answer

When the model memorises training images instead of learning general visual patterns

Answer

When the AI generates images with file sizes that are too large

Answer

When the AI crashes because it runs out of graphics memory

Answer

When a generated image contains too many colors and looks garish

Question 26

What is 'sampling' in a diffusion model, and why does the number of steps matter?

Accepted Answer

The step-by-step process of removing noise to produce a final generated image

Answer

Taking a small thumbnail crop from a generated image for a preview

Answer

Randomly selecting which training images to show the model in each batch

Answer

Showing the generated image to a sample group of users for feedback

Question 27

Name three popular text-to-image AI tools and explain what they have in common.

Accepted Answer

DALL-E, Midjourney, and Stable Diffusion — all generate new images from text prompts

Answer

DALL-E, Google Maps, and Photoshop — all edit existing photographs

Answer

GPT-4, DALL-E, and AutoCAD — all used by professional architects

Answer

Stable Diffusion, TensorFlow, and PyTorch — all training frameworks, not image tools

Question 28

Why does generating AI images require so much computing power?

Accepted Answer

Because billions of math operations must be performed across many network layers and many de-noising steps

Answer

Because every generated image must be automatically uploaded and stored in the cloud

Answer

Because the AI searches the internet for reference photos before drawing

Answer

Because the AI must compare every pixel against a large reference database manually

Question 29

What copyright concerns arise from training AI image models on internet images?

Accepted Answer

Using copyrighted images for training without permission, and questions of who owns AI outputs, are unresolved ethical and legal issues

Answer

AI-generated images are always free to use because a computer made them, not a human

Answer

Copyright law only applies to text documents, not to images

Answer

AI image generators automatically purchase licences for every image in their training set

Question 30

How are GANs used to generate realistic faces of people who don't exist?

Accepted Answer

By using a generator and discriminator trained on real portraits until the generator can produce convincing new faces

Answer

By scanning and storing real people's faces in a searchable database

Answer

By recognising and tagging known faces in existing photo libraries

Answer

By blurring real faces just enough that they appear to be different people

Test yourself