ICDL AI Essentials (ICDL Foundation)

Which of the following best defines Artificial Intelligence (AI)?

• AA system that can only perform pre-programmed mathematical calculations sequentially.
• BThe capability of a machine to imitate intelligent human behavior, such as learning, reas…
• CAny computer program that requires an internet connection to function.
• DA physical robot designed to replace human manual labor in factories.

What is the primary characteristic of Artificial Narrow Intelligence (ANI)?

• AIt possesses general human cognitive abilities and can perform any intellectual task a hu…
• BIt is programmed to perform a single, specific task or a limited range of tasks extremely…
• CIt is a sentient machine that has self-awareness and consciousness.
• DIt is capable of learning and teaching itself tasks beyond its original programming witho…

Which concept describes Artificial General Intelligence (AGI)?

• AA system that outperforms humans in a single specialized domain like chess.
• BA theoretical AI that possesses human-like cognitive abilities across a wide range of dis…
• CA software program that uses simple if-then decision trees to filter spam emails.
• DAn AI that surpasses the collective intelligence of all humanity combined.

What is Artificial Superintelligence (ASI)?

• AAI that has been optimized to run on standard mobile phones without cloud servers.
• BAI that is designed exclusively to automate corporate databases and financial spreadsheet…
• CA theoretical stage of AI where machine intelligence surpasses human intelligence across…
• DThe current generation of large language models like GPT-4 and Gemini.

What is an algorithm in the context of computer science and AI?

• AA physical microchip that stores training weights.
• BA set of step-by-step instructions or rules designed to solve a problem or complete a tas…
• CA computer virus designed to corrupt database systems.
• DThe user interface where humans write prompts.

How do heuristic algorithms differ from exact algorithms?

• AHeuristic algorithms guarantee the single most optimal solution, while exact algorithms d…
• BHeuristic algorithms use practical shortcuts to find a satisfactory solution quickly, whe…
• CHeuristic algorithms are written in natural human language, while exact algorithms are wr…
• DHeuristic algorithms can only run on quantum computers.

What is the primary definition of Machine Learning (ML)?

• AA database querying technique that retrieves records using strict SQL queries.
• BA subset of AI where systems automatically learn patterns and improve from data without b…
• CThe physical assembly of computers and servers in a cloud data center.
• DThe manual entry of text data into spreadsheets by human operators.

Which of the following scenarios best illustrates supervised learning?

• AAn AI agent learning to navigate a maze by receiving points for finding the exit and losi…
• BA model trained on thousands of emails labeled as 'spam' or 'not spam' to learn how to cl…
• CA clustering algorithm grouping customers into categories based on buying habits without…
• DA developer writing if-then rules to block specific email subject lines.

What is the key characteristic of unsupervised learning?

• AIt requires a human supervisor to monitor the computer processor in real-time.
• BIt uses labeled training datasets where the correct answers are provided up front.
• CIt analyzes unlabeled data to discover hidden patterns, structures, or anomalies on its o…
• DIt relies on physical robotic sensors to gather data.

In machine learning, how does reinforcement learning operate?

• ABy training a model exclusively on text databases containing regulatory guidelines.
• BBy mapping input data directly to predefined categories using manual classifications.
• CBy having an agent interact with an environment and learn to make decisions through a sys…
• DBy physically reinforcing CPU server racks to prevent overheating.

What are artificial neural networks in machine learning?

• ACables that physically link supercomputers across the internet.
• BComputational models inspired by the structure and function of biological brains, consist…
• CSocial media networks that track user behaviors.
• DElectronic circuits that process analog audio signals.

What is the primary defining feature of Deep Learning?

• AIt requires servers to be submerged in deep water tanks for optimal cooling.
• BIt uses neural networks with many hidden layers between the input and output layers to le…
• CIt is restricted to analyzing medical data related to deep-sea organisms.
• DIt is programmed exclusively using simple spreadsheet formulas.

Which of the following best describes Generative AI?

• AAn AI system that is used exclusively to generate electricity in renewable energy grids.
• BA type of AI that can create new content, such as text, images, audio, or code, based on…
• CA software program that deletes outdated files from databases.
• DA robot that physically assembles automotive parts on an assembly line.

How do generative AI models differ from discriminative AI models?

• AGenerative models are only used for text, while discriminative models are only used for i…
• BGenerative models create new data instances, whereas discriminative models classify or pr…
• CGenerative models require zero training data, while discriminative models require massive…
• DGenerative models are rule-based, while discriminative models are neural-network-based.

What was the primary goal of the Turing Test, proposed by Alan Turing in 1950?

• ATo calculate the maximum processing speed of a mechanical computing machine.
• BTo determine if a machine can exhibit intelligent behavior indistinguishable from that of…
• CTo test the water-resistance of early mainframe computer cases.
• DTo prove that computers could never use binary code.

In the history of AI, what was the significance of IBM's Deep Blue defeating chess champion Garry Kasparov in 1997?

• AIt proved that Deep Blue had achieved self-awareness and AGI.
• BIt demonstrated that a computer could use raw search power and heuristic evaluation to be…
• CIt marked the invention of deep learning and neural networks.
• DIt showed that chess could only be played by computers in the future.

Why was Google DeepMind's AlphaGo defeating Lee Sedol in 2016 considered a breakthrough compared to Deep Blue's chess win?

• ABecause Go has a vastly larger search space than chess, requiring AlphaGo to use deep neu…
• BBecause AlphaGo physically played the game using a robotic arm to place pieces.
• CBecause Go is a game of chance, proving that AI could predict random luck.
• DBecause AlphaGo was the first AI program to be connected to the internet.

What is a Large Language Model (LLM)?

• AA dictionary database containing every language spoken on Earth.
• BA deep learning model trained on vast amounts of text data to predict, generate, and mani…
• CA compiler that translates programming code directly into spoken speech.
• DA system designed to translate text only into classical Latin.

In modern AI, what is the role of the 'attention mechanism' in Transformer models?

• AIt sounds an alarm if the user stops looking at the screen.
• BIt allows the model to dynamically focus on different parts of the input text when proces…
• CIt forces the computer to only process one word at a time sequentially.
• DIt restricts the AI model from generating controversial outputs.

Why do data scientists split data into training and testing sets when building machine learning models?

• ATo save storage space on cloud servers.
• BTo ensure the model is trained on one set and evaluated on an independent, unseen set to…
• CTo delete corrupted data automatically.
• DTo make sure the model only memorizes the training data perfectly.

What is an 'AI Agent'?

• AA human sales representative who sells AI software licenses.
• BAn autonomous system that perceives its environment, makes decisions, and takes actions t…
• CA security program that prevents unauthorized users from logging into AI platforms.
• DA specific file extension used to save neural network models.

What is the primary focus of Natural Language Processing (NLP)?

• ATeaching computers to compile binary code faster.
• BEnabling computers to understand, interpret, and generate human languages in a valuable w…
• CDeveloping organic microphones that mimic human ears.
• DWriting grammar books for human language students.

Which of the following is a primary task of Computer Vision in AI?

• AEnabling computers to display high-resolution graphics on monitors.
• BTraining computers to interpret and understand visual information from digital images or…
• CImproving the physical lens manufacturing process for optical glasses.
• DDeveloping artificial eyes for humans using biotechnology.

What is the primary goal of Explainable AI (XAI)?

• ATo translate English programming code into foreign languages.
• BTo make the internal workings and decision-making processes of AI models transparent and…
• CTo write simple manuals for consumer electronic devices.
• DTo explain why computer monitors produce blue light.

Why are deep neural networks often referred to as 'black boxes'?

• ABecause they must be physically housed in dark rooms to function.
• BBecause their internal mathematical parameters and weight updates are so complex that it…
• CBecause they only output black-and-white text on computer screens.
• DBecause they are designed exclusively to process secret or classified data.

What is an AI 'hallucination' in Large Language Models?

• AA system crash caused by high processor temperatures.
• BA phenomenon where an AI model generates grammatically correct but factually incorrect or…
• CWhen the model's interface displays visual distortions to the user.
• DAn error where the model refuses to answer a prompt due to safety filters.

Why do Large Language Models occasionally hallucinate?

• ABecause they are programmed to lie to users to save computing energy.
• BBecause they operate as probabilistic token predictors rather than checking factual looku…
• CBecause computer viruses alter the model files on the host server.
• DBecause the models only have access to information written before 1900.

How can bias in training data affect a machine learning model?

• AThe model will run faster because biased data is simpler.
• BThe model will replicate, and potentially amplify, the prejudices and inequalities presen…
• CThe model will automatically correct human biases and produce perfectly fair outputs.
• DThe model will crash and delete its own weights.

Which principle is best expressed by the phrase 'garbage in, garbage out' in AI development?

• AAI systems should be used to optimize local municipal waste disposal routes.
• BThe quality of a machine learning model's output is directly dependent on the quality and…
• COutdated computer hardware must be recycled in specialized electronic waste centers.
• DModel parameters should be randomized after every training run.

What is a major limitation of current Narrow AI systems regarding common-sense reasoning?

• AThey can only process data when instructed by a mathematical formula.
• BThey lack general background knowledge and intuitive understanding of the physical and so…
• CThey are too slow to perform simple logic calculations.
• DThey cannot be programmed to follow if-then statements.

In affective computing, what is a key limitation of AI systems designed to recognize human emotions?

• AThey cannot detect variations in voice pitch or facial expressions.
• BThey do not experience true emotions or empathy; they only categorize physical cues based…
• CThey are legally blocked from being deployed on mobile devices.
• DThey require specialized neural implants to function.

Which of the following illustrates a vulnerability of computer vision systems known as an 'adversarial attack'?

• AA physical hacker destroying a camera lens with a hammer.
• BMaking subtle, invisible alterations to pixels in an image that cause a model to misclass…
• CA system failure caused by a sudden power surge in the server room.
• DAn AI model generating a high-quality video clip from scratch.

Why do speech-to-text AI models often struggle in real-world environments?

• ABecause they can only process digital audio that has been translated into Morse code firs…
• BBecause background noise, overlapping speakers, and diverse regional accents can degrade…
• CBecause they are designed to only recognize words spoken by computer scientists.
• DBecause sound waves travel too slowly for modern CPUs to process.

What is the concept of a 'Stochastic Parrot' in AI criticism?

• AA robotic toy that repeats sounds using basic microphone feedback loops.
• BThe idea that large language models merely combine and repeat language patterns from thei…
• CAn algorithm designed to model the migratory patterns of tropical birds.
• DA model that randomly deletes words from its database to save memory.

How do 'out-of-distribution' scenarios limit the reliability of machine learning models?

• AThey cause the model's server to lose internet connection.
• BThey present the model with input data that differs significantly from the training datas…
• CThey prevent the model from saving its files in standard formats.
• DThey make the model's interface load in black-and-white.

Why is it a security risk to input proprietary or sensitive personal information into public generative AI tools?

• ABecause the tools will automatically crash if they detect numbers.
• BBecause public models may store the input data and use it to train future versions, poten…
• CBecause public tools will encrypt your computer's hard drive and demand a ransom.
• DBecause inputting text wears out the processor chips faster than standard computing.

What environmental factor represents a key physical limitation of training large AI models?

• AThe models consume vast amounts of electricity and water for data center cooling, leading…
• BAI models generate radioactive waste during training.
• CTraining models depletes the Earth's magnetic field over time.
• DAI systems can only train during sunny weather.

What is the training data 'cutoff date' in Large Language Models?

• AThe exact time of day when a model must be shut down for maintenance.
• BThe point in time up to which the model was trained on data, meaning it has no direct kno…
• CThe date when a model's software license expires and it stops functioning.
• DThe speed limit at which data is transferred to the model.

What is 'overfitting' in machine learning?

• AWhen a model is too simple to learn the underlying patterns in the training data.
• BWhen a model learns the training data, including its noise and random details, too perfec…
• CWhen the physical computer server case is too small for the motherboard.
• DWhen a model automatically deletes correct predictions to save space.

What is 'underfitting' in machine learning?

• AWhen a model cannot capture the underlying trend of the data because it is too simple or…
• BWhen a model is trained for too many epochs on the same dataset, causing it to memorize n…
• CWhen a computer program runs out of memory and crashes.
• DWhen a server cabinet is half-empty.

Which real-world scenario illustrates gender bias in AI recruitment systems?

• AA system that scans resumes and automatically rejects files saved in PDF format.
• BAn AI hiring tool that penalizes resumes containing the word 'women's' (e.g., 'women's ch…
• CA software program that schedules interviews based on calendar availability.
• DAn AI tool that randomly selects applicants using a lottery system.

Why have facial recognition AI systems faced criticism regarding racial bias?

• ABecause they require users to wear colored glasses to be scanned.
• BBecause studies showed they had significantly higher error rates for individuals with dar…
• CBecause they can only recognize faces in black-and-white video streams.
• DBecause they are designed to delete profiles of minority users.

What is a 'deepfake'?

• AA secure database located deep underground.
• BSynthetic media (images, audio, or video) where a person's likeness is replaced with some…
• CA computational method used to predict tectonic plate movements.
• DAn online scam that uses standard emails to steal passwords.

Why can AI models not autonomously verify the truth or accuracy of their own generated output?

• ABecause they lack a connection to external power grids.
• BBecause they match linguistic and mathematical patterns rather than possessing semantic u…
• CBecause they are programmed to always believe their own outputs are correct.
• DBecause they do not have access to spellcheck systems.

In the lifecycle of an AI model, what is the difference between 'training' and 'inference'?

• ATraining is performed by humans, while inference is performed entirely by robots.
• BTraining is the process of teaching a model using a dataset to adjust its parameters, whi…
• CTraining requires zero electricity, while inference requires massive supercomputers.
• DTraining is only for text, while inference is only for images.