Appearance
Topic Drill 1.1.4: Generative AI
About this drill
LO AI-1.1.4 (K2): Explain generative AI.
Lesson: Ch 1, section 1.1.4
This drill covers every point in section 1.1.4:
- What GenAI creates, and that many GenAI systems also classify and predict; outputs resemble the training data
- The three main technologies: GANs, diffusion models, transformers (self-attention, LLMs, multimodal)
- Concerns: misuse (deepfakes, misinformation, fraud; privacy, security, manipulation) balanced against benefits; employment; sustainability
- Foundation models with fine-tuning, multimodal models, and regulation such as the EU AI Act
Topic Drill: 1.1.4
Q1. Which description of generative adversarial networks (GANs) matches the syllabus?
AThey generate content by gradually adding and then removing noise from data
BThey use self-attention to generate coherent, contextually relevant text
CThey use two neural networks in competition to create highly realistic synthetic data
DThey retrieve documents from a knowledge base to enhance prompts
Q2. Which technology underpins large language models (LLM), according to 1.1.4?
ATransformer models using self-attention mechanisms
BDiffusion models
CGenerative adversarial networks
DFuzzy logic
Q3. A realistic video of a company CEO announcing a merger that never happened circulates on social media and moves the share price. Which GenAI concern does this illustrate?
ASustainability
BImpact on employment
CMultimodality
DMisuse, such as deepfakes that undermine trust in digital media
Q4. A law firm finds that GenAI now drafts most standard contracts, so it needs fewer junior staff for this work. Which GenAI concern and consequence does the syllabus describe for this situation?
AMisuse, which requires deepfake detection
BImpact on employment, which may lead to workforce disruption and require widespread reskilling
CSustainability, which requires greener infrastructure
DPrivacy, which requires anonymization of contracts
Q5. What does the syllabus say most practical GenAI tools today are based on?
ARule-based expert systems
BModels trained from scratch for each application
CFoundation models that are fine-tuned for specific applications
DSuper AI systems
Q6. Which TWO statements about GenAI are CORRECT according to the syllabus?
Select 2 options.
AGenAI systems produce outputs that resemble their training data
BGenAI is limited to producing text and images
CDiffusion models use two competing neural networks
DGenAI raises no environmental concerns once a model is trained
EMany GenAI systems also support classification and prediction
Q7. A tool accepts a photo and a typed question and answers with spoken audio. Which trend in GenAI does this illustrate?
AMultimodal models that process and generate content across text, images and audio
BNarrow AI becoming super AI
CDiffusion models replacing transformers
DRule-based reasoning
Q8. Which statement BEST reflects how the syllabus balances the risks and benefits of synthetic content?
ASynthetic content has no legitimate use and should be banned
BThe risks are purely technical and can be removed by better models
CSynthetic content only affects white-collar employment
DIt amplifies risks related to privacy, security and manipulation, but brings notable benefits in entertainment, marketing, education and research
Q9. According to the syllabus, what has the high energy consumption of large GenAI models prompted?
AA ban on training models larger than a set size
BCalls for more efficient model designs and greener infrastructure
CA move back to rule-based systems
DMandatory use of neuromorphic processors
0 / 9 answered
← 1.1.3 Types of AI Technologies · Next drill: 1.1.5 Hardware for ML Systems →