I85PREP
⌘ K
ICDL AI Essentials study materials
ICDL AI Essentials study resources

ICDL AI Essentials practice questions

One question at a time. Build your ICDL AI Essentials understanding.

First 50 of 200 questions. Each one lists the choices, the correct answer, and an explanation. The checker at the top of this page saves progress in this browser when it is available.

Continue in the practice tool above to load more questions, 50 at a time.

Question 1What is AI?

What does the term artificial intelligence describe?

  1. Any application software that is installed on a desktop computer
  2. Specialised hardware that increases a computer's processing speed
  3. A very large database that stores millions of business records
  4. Computer systems that carry out tasks needing human intelligence Correct answer.

Explanation

Artificial intelligence describes computer systems that perform tasks normally associated with human intelligence, such as recognising images, understanding language or making decisions. The defining feature is the kind of task being performed, not the size of the database or the speed of the hardware.

Question 2What is AI?

Alan Turing's test is designed to assess a machine's ability to do what?

  1. Run faster than a human calculator
  2. Operate without electrical power
  3. Store more data than a human can
  4. Produce responses indistinguishable from a human's Correct answer.

Explanation

The Turing test involves a human judge who exchanges messages with both a machine and a person and tries to tell them apart. If the judge cannot reliably distinguish the machine's replies from the human's, the machine is said to have passed.

Question 3What is AI?

Which description best matches a narrow, or weak, AI system?

  1. It matches human ability in every domain
  2. It requires no data of any kind to function
  3. It is designed to perform one specific task Correct answer.
  4. It becomes conscious of its own existence

Explanation

Narrow AI is built for a limited purpose, such as recommending films, translating text or detecting defects on a production line. It may perform that one task extremely well, but it cannot transfer its skill to unrelated problems.

Question 4What is AI?

Which everyday feature most clearly relies on artificial intelligence?

  1. A text editor counting the words on a page
  2. A calculator adding two numbers together
  3. A clock displaying the current time in several time zones
  4. A music app suggesting songs based on your history Correct answer.

Explanation

Recommendation features learn patterns from large amounts of user behaviour and use those patterns to predict what a person may want next. That learning from data is a hallmark of AI. Arithmetic, word counting and time display all follow fixed rules with no learning involved.

Question 5What is AI?

What is machine learning?

  1. Systems that improve at a task by learning from data Correct answer.
  2. A method of manually writing every decision rule into a program
  3. A type of computer memory that stores running programs
  4. A programming language used to design web pages

Explanation

Machine learning builds models that find patterns in examples and use those patterns to make predictions on new data. Instead of a programmer writing a rule for every case, the system adjusts itself from the data it is given.

Question 6What is AI?

A system is trained on thousands of photographs that have each been marked either 'cat' or 'dog'. Which learning approach does this describe?

  1. Reinforcement learning
  2. Supervised learning Correct answer.
  3. Rule-based programming
  4. Unsupervised learning

Explanation

Supervised learning uses examples that already carry the correct answer, known as labels. The model learns a mapping from the input photograph to the label so that it can classify new, unseen images.

Question 7What is AI?

A retailer gives an algorithm years of purchase records with no labels attached, and the algorithm groups customers with similar buying habits. Which type of learning is this?

  1. Reinforcement learning
  2. Symbolic rule chaining
  3. Unsupervised learning Correct answer.
  4. Supervised learning

Explanation

Unsupervised learning explores data without any target answer and tries to discover structure on its own, such as clusters of similar records. Grouping customers by buying habits is a typical clustering task, because no one tells the algorithm which group each customer belongs to.

Question 8What is AI?

How does reinforcement learning train a system?

  1. By supplying a fully labelled dataset of correct answers
  2. By clustering unlabelled data into similar groups
  3. By rewarding desirable actions and penalising undesirable ones Correct answer.
  4. By loading a fixed set of hand-written rules into memory

Explanation

In reinforcement learning an agent acts in an environment and receives feedback in the form of rewards or penalties. Over many attempts it adjusts its behaviour to obtain more reward, much as a player improves by learning which moves lead to a win.

Question 9What is AI?

Why is the quality of training data so important in machine learning?

  1. Data quality matters only for image-based models
  2. Training data is used only to display final results to end users
  3. Data quality affects only the speed of the storage disk
  4. Models learn from data, so poor data yields poor predictions Correct answer.

Explanation

A model derives its internal patterns from the examples it is shown, so errors, gaps and skew in that data are carried into its predictions. This is often summarised as 'garbage in, garbage out'. Good data collection and cleaning are therefore central to reliable AI.

Question 10What is AI?

An artificial neural network is loosely modelled on what?

  1. Layers of a computer file system
  2. Circuits in a national power grid
  3. Steps in a printed instruction manual
  4. Networks of neurons in the brain Correct answer.

Explanation

Artificial neural networks use layers of connected units, often called nodes or artificial neurons, that pass signals to one another. The design was inspired by the way biological neurons connect and transmit signals in the brain, although the artificial version is far simpler.

Question 11What is AI?

Which statement best describes deep learning?

  1. A method of compressing files for storage
  2. Machine learning using neural networks with many layers Correct answer.
  3. Learning that happens only after a system is switched off
  4. Any program that runs without user input

Explanation

Deep learning is a branch of machine learning that uses neural networks containing many hidden layers. The extra layers allow the model to build increasingly abstract representations, which is why deep learning excels at tasks such as image and speech recognition.

Question 12What is AI?

Which task belongs to natural language processing?

  1. Compressing a video file for storage
  2. Translating a sentence from French into English Correct answer.
  3. Balancing the workload between two servers
  4. Detecting the edges of objects in a photograph

Explanation

Natural language processing deals with human language in text or speech form, including translation, summarisation and sentiment analysis. Machine translation requires the system to model grammar and meaning across two languages.

Question 13What is AI?

A system identifies a pedestrian from a live camera feed. Which area of AI does this illustrate?

  1. Expert systems
  2. Natural language processing
  3. Speech synthesis
  4. Computer vision Correct answer.

Explanation

Computer vision enables machines to interpret and analyse visual information such as images and video. Detecting and classifying a person in a camera frame is a classic vision task, often used in driver assistance and security systems.

Question 14What is AI?

What distinguishes generative AI from a simple classifier?

  1. It stores data without producing any output for the user
  2. It creates new content such as text, images or audio Correct answer.
  3. It works only with numerical spreadsheets
  4. It can only sort existing files into folders

Explanation

A classifier assigns an input to a category, whereas a generative model produces new material in the style of what it learned. The output might be a paragraph of text, a picture or a piece of music that did not exist before.

Question 15What is AI?

What characterises a classic expert system?

  1. It generates images from short text prompts
  2. It learns entirely from unlabelled sensor data
  3. It replaces its own hardware when a part fails
  4. It applies rules written by human specialists Correct answer.

Explanation

Expert systems capture the knowledge of specialists as an explicit set of if-then rules and apply them to a narrow domain, such as diagnosing a fault. They reason from those stored rules rather than learning patterns from large datasets.

Question 16What is AI?

A facial recognition system performs poorly on some demographic groups because those groups were under-represented in its training set. What does this illustrate?

  1. A hardware fault in the camera lens
  2. The system having too much labelled data
  3. Bias caused by unrepresentative training data Correct answer.
  4. Encryption reducing the image quality

Explanation

When part of the population appears rarely or never in the training data, the model learns less about it and its accuracy drops for that group. This is a data-driven bias rather than a mechanical defect. Balanced, representative datasets are one of the main remedies.

Question 17What is AI?

How does a machine-learning model differ from a conventional program built from fixed rules?

  1. It derives behaviour from data patterns rather than explicit rules Correct answer.
  2. It never makes mistakes on data it has not seen before
  3. It stores no data at all while it is operating
  4. It cannot be retrained once it has been deployed

Explanation

In a conventional program a developer writes the logic that turns inputs into outputs. In a machine-learning model the logic is inferred from examples during training, so the model generalises from patterns instead of following instructions written for every case.

Question 18What is AI?

A chatbot confidently states a false fact that appears in none of its sources. What is this behaviour called?

  1. Data augmentation
  2. Hallucination Correct answer.
  3. Model compression
  4. Overfitting to the user's prompt length

Explanation

Hallucination describes a generative system producing fluent but fabricated information as though it were factual. Because the output is generated from learned patterns rather than looked up in a verified source, it can sound convincing while being wrong. Checking important claims against reliable sources is a standard safeguard.

Question 19What is AI?

Why should a human review AI output before it is used in a high-stakes decision?

  1. Models can be confidently wrong, and human judgement is a safety check Correct answer.
  2. Models never reflect bias present in their training data
  3. AI systems cannot process large volumes of information
  4. Human review removes the need for any training data

Explanation

AI models can produce plausible but incorrect results and cannot reliably signal when they are unsure. A human reviewer adds context, accountability and judgement, which matters greatly in areas such as medicine, lending or recruitment.

Question 20What is AI?

What does the term 'AI winter' refer to?

  1. A shutdown procedure for neural networks
  2. The seasonal drop in chatbot usage during summer
  3. A period when funding for AI research fell sharply Correct answer.
  4. A cooling system used in large data centres

Explanation

AI winter describes periods in the history of the field when expectations were disappointed and public and private funding for research shrank. These cycles show that progress in AI has not been smooth and has often depended on renewed interest and investment.

Question 21What is AI?

Which is a typical use of a virtual assistant on a smart speaker?

  1. Answering spoken questions and controlling smart home devices Correct answer.
  2. Replacing the operating system automatically
  3. Generating electricity for the whole household
  4. Repairing damaged hardware inside the device

Explanation

Virtual assistants combine speech recognition and natural language processing to understand spoken requests and carry out actions such as answering a question or switching on a light. They act on instructions rather than performing physical repairs or supplying power.

Question 22What is AI?

An AI system that assigns each incoming email to the category 'spam' or 'not spam' is performing what kind of task?

  1. Regression
  2. Clustering
  3. Classification Correct answer.
  4. Compression

Explanation

Classification predicts which of a fixed set of categories an item belongs to. Because the possible outputs are the two predefined labels spam and not spam, the task is binary classification.

Question 23What is AI?

What core ability allows an AI system to recognise a face in a photograph it has never seen?

  1. Reading the image's file name for clues
  2. Randomly guessing until it happens to succeed
  3. Following a printed list of every human face
  4. Recognising patterns learned from many previous examples Correct answer.

Explanation

Face recognition models learn general patterns, such as the arrangement of eyes, nose and mouth, from many labelled examples. They then compare those learned patterns with a new image, which is why they can handle faces that were not in the training set.

Question 24What is AI?

Before training a model to detect defective products, staff mark each image as 'defective' or 'acceptable'. What is this step called?

  1. Data encryption
  2. Data labelling Correct answer.
  3. Inference
  4. Model deployment

Explanation

Data labelling attaches the correct answer to each training example so that a supervised model can learn from it. Accurate labelling is essential because the model will reproduce the distinctions that the labels encode.

Question 25What is AI?

Why do many organisations publish guidelines for the responsible use of AI?

  1. To guarantee that models never make errors
  2. To increase the processing power of their servers
  3. To reduce the amount of training data required
  4. To set expectations for fairness and accountability Correct answer.

Explanation

Responsible AI guidelines set out how systems should be fair, transparent, safe and accountable, and who is answerable when something goes wrong. They shape design and deployment decisions rather than affecting computing resources or model accuracy directly.

Question 26What is AI?

Which statement best describes artificial intelligence?

  1. Software that stores records in tables and retrieves them on request
  2. Systems that perform tasks normally needing human intelligence Correct answer.
  3. Hardware that moves physical objects without any human control
  4. Programs that follow a fixed sequence of steps without variation

Explanation

AI refers to computer systems able to carry out tasks that we associate with human intelligence, such as recognising images, understanding language, or making decisions. The defining point is the kind of task performed, not the physical form of the machine. A program that merely executes a fixed sequence of steps, or a database that stores records, does not adapt or reason.

Question 27What is AI?

Which term describes an AI system built to perform one specific task, such as translating text or recognising speech?

  1. General artificial intelligence
  2. Superintelligent artificial intelligence
  3. Narrow artificial intelligence Correct answer.
  4. Strong artificial intelligence

Explanation

Almost all AI in use today is narrow AI: it is designed for a particular task and performs poorly outside that scope. General and strong AI refer to hypothetical systems with broad, human-like competence across many domains. Superintelligence goes further still, describing a system far beyond human ability.

Question 28What is AI?

The Turing test was proposed as a way to judge whether a machine can do which of the following?

  1. store more information than a human brain can hold
  2. move around a physical environment without human help
  3. learn from data without human-provided labels
  4. reply in a way that cannot be told apart from a human Correct answer.

Explanation

Alan Turing suggested that if a machine's written replies to questions were indistinguishable from a person's, we would have reason to call it intelligent. The test therefore targets conversational behaviour rather than physical movement, memory capacity, or unsupervised learning. It evaluates observable responses, not inner consciousness.

Question 29What is AI?

Which of these is an everyday application of artificial intelligence?

  1. A photo app that groups pictures by the person shown Correct answer.
  2. A light switch that turns a bulb on and off
  3. A wall clock that shows the correct time in one room
  4. A calculator that adds two entered numbers

Explanation

Grouping photographs by the person in them requires a model to detect and match faces, which is a learned perceptual skill. A clock, a switch, and a calculator all perform fixed, pre-defined operations with no learning or pattern recognition. AI is characterised by handling variability that fixed logic cannot easily cover.

Question 30What is AI?

How does an AI chess program typically choose its next move?

  1. It evaluates many possible positions and picks the most promising one Correct answer.
  2. It asks a human operator which move should be played next
  3. It looks up the exact game in a stored table of past matches
  4. It follows a hard-coded list of moves written for every opening

Explanation

Game-playing AI searches through possible future positions and scores them against an evaluation function, then selects the strongest line it can find. Storage of complete games cannot cover the enormous number of reachable positions. Hard-coded openings cover only a small early part of the game, so search is what carries it through.

Question 31What is AI?

Once training is complete, a model receives a fresh input and produces an output. What is this stage called?

  1. Random sampling
  2. Data labelling
  3. Cluster analysis
  4. Inference Correct answer.

Explanation

Inference is the phase in which a trained model applies what it has learned to new, unseen inputs. Labelling belongs to data preparation before training, and clustering is a training technique rather than a deployment stage. Sampling is a way of choosing data, not of producing an answer.

Question 32What is AI?

A voice assistant converts your spoken request into written words. Which AI capability does this use?

  1. Image segmentation
  2. Route planning
  3. Speech recognition Correct answer.
  4. Product recommendation

Explanation

Speech recognition, also called speech-to-text, turns an audio signal into a sequence of words by matching acoustic patterns to language. Image segmentation works on pictures, and route planning computes paths through a map. Recommendation systems suggest items based on preference data.

Question 33What is AI?

An online shop suggests items by comparing your past orders with those of similar shoppers. What is this called?

  1. A recommendation system Correct answer.
  2. Predictive maintenance
  3. Process automation
  4. Speech synthesis

Explanation

A recommendation system analyses behaviour across many users to predict items an individual is likely to want. It relies on patterns in preferences rather than on fixed rules. Speech synthesis produces audio, process automation replays routine steps, and predictive maintenance forecasts equipment failure.

Question 34What is AI?

Chess engines were once celebrated as advanced AI, yet many people today regard them as ordinary software. What does this illustrate?

  1. AI can only be developed inside research laboratories
  2. People stop calling a task AI once computers solve it well Correct answer.
  3. AI systems always become more capable as they age
  4. Chess is a task that computers cannot really perform

Explanation

The boundary of what counts as AI shifts over time: once a capability becomes routine and well understood, it is often reclassified as ordinary computing. This moving boundary is sometimes called the AI effect. The phenomenon reflects changing expectations, not any change in the software's ability.

Question 35What is AI?

A website tool answers typed questions about orders and returns in ordinary language. Which AI application is this?

  1. An autonomous vehicle
  2. A robotic arm
  3. A conversational agent Correct answer.
  4. A vision system

Explanation

A conversational agent, or chatbot, interprets natural-language input and produces a relevant reply. It depends on language understanding rather than on perception or movement. Autonomous vehicles, vision systems, and robotic arms operate on sensor data and physical action.

Question 36What is AI?

A fitness app uses AI to analyse users' heart-rate and sleep data. Which concern does this raise most directly?

  1. The app may run too slowly on older phones
  2. The app may not support every screen size
  3. The app may consume too much battery power
  4. Sensitive personal health data may be collected or misused Correct answer.

Explanation

Health data is among the most sensitive categories of personal information, so collecting it through an AI service raises questions about consent, storage, and possible misuse. Performance and display issues are technical quality matters rather than ethical ones. The ethical risk arises from the nature of the data, not from how fast the app runs.

Question 37What is AI?

An autonomous vehicle causes a collision while driving itself. Which question relates to accountability?

  1. Which programming language was used to build it?
  2. How much did it cost to manufacture the vehicle?
  3. How many sensors does the vehicle carry?
  4. Who is answerable for harm caused by the system's decision? Correct answer.

Explanation

Accountability asks who bears responsibility for the consequences of an AI system's actions, whether that is the developer, the operator, or the owner. Sensor counts, programming languages, and manufacturing costs are technical or commercial details. They help explain how the system works but do not assign responsibility for harm.

Question 38What is AI?

A medical advice system stores facts such as 'fever often accompanies flu' and draws conclusions from them. Which AI area does this illustrate?

  1. Knowledge representation and reasoning Correct answer.
  2. Computer vision and image analysis
  3. Speech synthesis and voice output
  4. Robotic process automation

Explanation

Knowledge representation is about encoding facts and relationships in a form a machine can use, and reasoning is the process of deriving new conclusions from them. Systems of this kind, including classic expert systems, work from explicit rules rather than learned statistical patterns. Images, speech, and routine process steps are not involved in the example.

Question 39What is AI?

A factory arm adjusts its grip when parts arrive in different shapes and positions. Which combination does this require?

  1. Robotics together with computer vision Correct answer.
  2. Speech recognition and voice output
  3. Recommendation systems and user profiling
  4. Natural language processing and translation

Explanation

The arm must perceive the part's shape and location, which is a computer vision task, and then act physically, which is robotics. Combining perception with controlled movement lets it adapt to variation instead of repeating one fixed motion. Language and profiling techniques play no role in gripping objects.

Question 40What is AI?

Which belief belongs to the 'strong AI' position?

  1. AI should never be applied to medicine
  2. AI is useful only for one narrow task at a time
  3. AI systems are always more reliable than people
  4. A properly programmed machine could truly understand, not just imitate Correct answer.

Explanation

Strong AI holds that a suitably designed machine could have genuine understanding or mind, rather than merely simulating intelligent behaviour. The opposing weak AI view treats the machine as a powerful simulator with no real comprehension. The other options concern scope, application policy, or reliability and do not address this philosophical claim.

Question 41What is AI?

Which statement correctly contrasts symbolic AI with machine learning?

  1. Machine learning handles only numbers, while symbolic AI handles only written text
  2. Symbolic AI uses hand-written rules, while machine learning infers patterns from data Correct answer.
  3. Symbolic AI cannot represent any facts, while machine learning stores facts in tables
  4. Symbolic AI learns from data, while machine learning relies on explicit rules written by experts

Explanation

Symbolic AI encodes knowledge as explicit rules and facts that a reasoning engine applies, whereas machine learning adjusts a model from examples so that patterns are induced from data. The two approaches differ in where the intelligence comes from: human-authored logic versus statistical fitting. Modern systems often combine both.

Question 42What is AI?

A smartphone unlocks when it recognises its owner's face. Which AI capability does this rely on?

  1. Machine translation
  2. Image recognition Correct answer.
  3. Route planning
  4. Speech synthesis

Explanation

Face unlocking depends on image recognition: the camera captures a picture and a model compares it with stored representations of the owner's face. This is a perceptual classification task. Speech synthesis, translation, and route planning operate on entirely different kinds of input.

Question 43What is AI?

In machine learning, what is a model?

  1. A chart that shows how training progressed over time
  2. A trained artefact that turns inputs into outputs or predictions Correct answer.
  3. A collection of raw images stored on a disk
  4. A physical replica of a robot used for testing

Explanation

A model is the result of training: a set of learned parameters that maps an input, such as an image or a sentence, to an output such as a label or a score. It is distinct from the data it was trained on and from any visualisation of the training process. The term does not refer to a physical object.

Question 44What is AI?

Which statement about AI and human intelligence is accurate?

  1. AI can outperform people on narrow tasks without general intelligence Correct answer.
  2. AI systems cannot be faster than humans at any task
  3. AI systems are conscious in the same way people are
  4. AI systems feel emotions in much the same way humans do

Explanation

Current AI can exceed human performance in specific domains, such as certain games or image classification, while lacking the broad, flexible intelligence people apply across situations. High performance on a narrow task is not evidence of consciousness or feeling. Nor does specialisation mean machines are always slower than humans.

Question 45What is AI?

Which statement about AI systems is true?

  1. They always produce correct results
  2. They cannot run on mobile devices
  3. They never need any training data
  4. They can assist with tasks but may still make mistakes Correct answer.

Explanation

AI systems are probabilistic tools that support human work, and their outputs can be wrong, especially when inputs differ from what they saw in training. They generally depend on data to learn from, and many run on phones. Claiming perfect accuracy or a ban on mobile deployment misstates how these systems behave.

Question 46What is AI?

A model scores almost perfectly on the examples it was trained on yet performs poorly on new records. Which explanation fits best?

  1. It was trained with unsupervised learning methods
  2. It has seen too little data to memorise anything at all
  3. It has memorised the training examples rather than learned general patterns Correct answer.
  4. It has been released to far too many users

Explanation

This gap between training performance and new-data performance is the signature of overfitting: the model has captured noise and specific details of the training set instead of the underlying regularity. The remedy is usually more or more varied data, simpler models, or regularisation. Generalisation, not memorisation, is the goal of training.

Question 47What is AI?

A bank uses AI to estimate how likely each loan applicant is to repay. Which kind of task is this?

  1. Image segmentation
  2. Predictive modelling Correct answer.
  3. Machine translation
  4. Speech synthesis

Explanation

Estimating a future outcome such as repayment from historical records is predictive modelling, a core use of machine learning in business. The model learns relationships between applicant characteristics and past outcomes, then applies them to new cases. The other options concern pictures, audio, and language conversion.

Question 48What is AI?

Why is explainability particularly important for an AI system that supports medical diagnosis?

  1. It removes the need to collect any training data
  2. It reduces the amount of electricity the system uses
  3. Clinicians must be able to justify and check decisions that affect patients Correct answer.
  4. It makes the trained model much smaller

Explanation

In medicine, a decision can affect a person's health, so a clinician needs to understand the reasoning behind a suggestion before acting on it. Explainability supports verification, informed consent, and professional accountability. It is a transparency property, not an efficiency measure, and it does not replace the need for data.

Question 49What is AI?

Which statement describes a recognised limitation of current AI systems?

  1. They may fail when a situation differs from their training data Correct answer.
  2. They cannot process images of any kind
  3. They cannot run on ordinary computers
  4. They are unable to store data of any kind

Explanation

Models learn from the distribution of their training examples, so unfamiliar situations can lead to unreliable output. This is why testing on varied data and keeping humans in the loop matters. Modern AI does process images, does store and use data, and frequently runs on ordinary hardware.

Question 50What is AI?

What does 'AI alignment' refer to?

  1. Aligning image pixels before a model processes them
  2. Designing systems to act in line with human values and intentions Correct answer.
  3. Sorting training data into neatly ordered columns
  4. Connecting several computers into a single network

Explanation

Alignment is the research and governance goal of making an AI system pursue the outcomes people actually want, rather than a literal or unintended interpretation of its instructions. It covers specification of goals, robustness, and oversight. The other options describe image preprocessing, data preparation, and networking.