Level 3–( AI ) Advanced | Artificial Intelligence |

Question 1 of 50

1. Question

1 point(s)

What is a major advantage of hybrid retrieval?

It can combine different search methods to improve information retrieval

Correct Incorrect Correct answer

It removes all irrelevant information automatically

Correct Incorrect Correct answer

It eliminates the need for source documents

Correct Incorrect Correct answer

It guarantees perfect generation

Correct Incorrect Correct answer

Question 2 of 50

2. Question

1 point(s)

What is data minimization?

Collecting every available data source

Correct Incorrect Correct answer

Keeping data forever

Correct Incorrect Correct answer

Sharing data with all users

Correct Incorrect Correct answer

Collecting and using only the data needed for a defined purpose

Correct Incorrect Correct answer

Question 3 of 50

3. Question

1 point(s)

Why should AI incident response be part of an organization’s governance framework?

AI failures may require investigation, containment, correction, and documented follow-up

Correct Incorrect Correct answer

AI systems cannot experience operational failures

Correct Incorrect Correct answer

Incident response is only for hardware

Correct Incorrect Correct answer

It removes the need for monitoring

Correct Incorrect Correct answer

Question 4 of 50

4. Question

1 point(s)

An AI agent repeatedly calls the same tool because it cannot determine whether the task succeeded. What should be improved?

Tool-state feedback and termination conditions

Correct Incorrect Correct answer

Screen resolution

Correct Incorrect Correct answer

Password length

Correct Incorrect Correct answer

Document color

Correct Incorrect Correct answer

Question 5 of 50

5. Question

1 point(s)

Why should sensitive training data have controlled access?

Unauthorized access can create privacy and security risks

Correct Incorrect Correct answer

Access control reduces model intelligence

Correct Incorrect Correct answer

Sensitive data cannot be processed

Correct Incorrect Correct answer

Security controls always reduce accuracy

Correct Incorrect Correct answer

Question 6 of 50

6. Question

1 point(s)

An AI assistant begins producing inaccurate answers after a source database is changed. What should the team examine?

Only the user interface

Correct Incorrect Correct answer

Only the model's logo

Correct Incorrect Correct answer

The data pipeline, retrieval behavior, and downstream model performance

Correct Incorrect Correct answer

The office network printer

Correct Incorrect Correct answer

Question 7 of 50

7. Question

1 point(s)

What is planning in an AI agent?

Increasing the model's vocabulary

Correct Incorrect Correct answer

Encrypting user messages

Correct Incorrect Correct answer

Breaking a goal into actions or steps that can be executed

Correct Incorrect Correct answer

Removing external tools

Correct Incorrect Correct answer

Question 8 of 50

8. Question

1 point(s)

What is model governance primarily concerned with?

Policies, accountability, controls, monitoring, and oversight across the AI lifecycle

Correct Incorrect Correct answer

Increasing processor speed

Correct Incorrect Correct answer

Designing application logos

Correct Incorrect Correct answer

Writing longer prompts

Correct Incorrect Correct answer

Question 9 of 50

9. Question

1 point(s)

What is the purpose of fallback mechanisms in critical AI systems?

To provide safer alternatives when the AI cannot reliably complete a task

Correct Incorrect Correct answer

To make every response longer

Correct Incorrect Correct answer

To bypass security controls

Correct Incorrect Correct answer

To eliminate human involvement

Correct Incorrect Correct answer

Question 10 of 50

10. Question

1 point(s)

Why should agent permissions follow the principle of least privilege?

To make the agent more creative

Correct Incorrect Correct answer

To increase token counts

Correct Incorrect Correct answer

To eliminate all human interaction

Correct Incorrect Correct answer

To limit potential damage if the agent behaves incorrectly or is compromised

Correct Incorrect Correct answer

Question 11 of 50

11. Question

1 point(s)

An organization deploys dozens of AI systems without knowing who owns them or what data they use. What is a major governance gap?

Lack of AI inventory and accountability

Correct Incorrect Correct answer

Lack of screen space

Correct Incorrect Correct answer

Excessive model evaluation

Correct Incorrect Correct answer

Too much documentation

Correct Incorrect Correct answer

Question 12 of 50

12. Question

1 point(s)

An AI system cannot confidently answer a high-impact question. What is the safer design principle?

Force an answer regardless of confidence

Correct Incorrect Correct answer

Hide uncertainty from users

Correct Incorrect Correct answer

Increase the temperature automatically

Correct Incorrect Correct answer

Allow it to abstain or escalate rather than fabricate an answer

Correct Incorrect Correct answer

Question 13 of 50

13. Question

1 point(s)

What is the main purpose of an AI architecture?

To increase monitor size

Correct Incorrect Correct answer

To define how AI components, data, models, tools, and users work together

Correct Incorrect Correct answer

To eliminate all software testing

Correct Incorrect Correct answer

To replace organizational policies

Correct Incorrect Correct answer

Question 14 of 50

14. Question

1 point(s)

An autonomous agent can send emails, modify records, and approve transactions. Which design is most appropriate for high-impact actions?

Allow unrestricted execution

Correct Incorrect Correct answer

Remove activity logs

Correct Incorrect Correct answer

Hide tool calls from administrators

Correct Incorrect Correct answer

Require authorization or human approval where appropriate

Correct Incorrect Correct answer

Question 15 of 50

15. Question

1 point(s)

Why should AI systems have clearly defined owners?

Ownership makes models more intelligent

Correct Incorrect Correct answer

It eliminates all technical failures

Correct Incorrect Correct answer

Someone must be accountable for operation, risks, monitoring, and changes

Correct Incorrect Correct answer

It prevents the need for documentation

Correct Incorrect Correct answer

Question 16 of 50

16. Question

1 point(s)

What best describes responsible AI at an advanced organizational level?

Maximizing automation regardless of risk

Correct Incorrect Correct answer

Using the largest available model

Correct Incorrect Correct answer

Integrating performance, safety, security, privacy, fairness, accountability, and human oversight

Correct Incorrect Correct answer

Removing humans from all workflows

Correct Incorrect Correct answer

Question 17 of 50

17. Question

1 point(s)

An enterprise AI system retrieves documents, generates answers, and records user activity. Which concern becomes especially important?

Screen resolution

Correct Incorrect Correct answer

System architecture and governance

Correct Incorrect Correct answer

Keyboard design

Correct Incorrect Correct answer

Office lighting

Correct Incorrect Correct answer

Question 18 of 50

18. Question

1 point(s)

What is an AI guardrail?

A control designed to constrain unsafe, invalid, or unauthorized model behavior

Correct Incorrect Correct answer

A neural network layer

Correct Incorrect Correct answer

A type of database

Correct Incorrect Correct answer

A method for increasing model size

Correct Incorrect Correct answer

Question 19 of 50

19. Question

1 point(s)

What is human-in-the-loop particularly valuable for?

Tasks that require no oversight

Correct Incorrect Correct answer

Decisions where human judgment or authorization remains important

Correct Incorrect Correct answer

Removing all accountability

Correct Incorrect Correct answer

Avoiding system testing

Correct Incorrect Correct answer

Question 20 of 50

20. Question

1 point(s)

An organization is planning enterprise-wide AI adoption. Which strategy reflects mature AI implementation?

Deploy AI everywhere immediately

Correct Incorrect Correct answer

Give every AI system unrestricted access

Correct Incorrect Correct answer

Focus only on model accuracy and ignore operational risks

Correct Incorrect Correct answer

Start with defined use cases, assess risks, establish governance, test systems, monitor outcomes, and improve continuously

Correct Incorrect Correct answer

Question 21 of 50

21. Question

1 point(s)

What is the primary purpose of MLOps?

To create marketing content

Correct Incorrect Correct answer

To replace data collection

Correct Incorrect Correct answer

To remove human review

Correct Incorrect Correct answer

To support the reliable development, deployment, monitoring, and maintenance of ML systems

Correct Incorrect Correct answer

Question 22 of 50

22. Question

1 point(s)

Which approach can improve agent reliability?

Give unlimited permissions

Correct Incorrect Correct answer

Constrain actions, validate outputs, log activity, and define failure handling

Correct Incorrect Correct answer

Remove all validation

Correct Incorrect Correct answer

Allow endless tool calls

Correct Incorrect Correct answer

Question 23 of 50

23. Question

1 point(s)

What is human-on-the-loop?

Humans monitor an automated system and can intervene when needed

Correct Incorrect Correct answer

Humans manually perform every model calculation

Correct Incorrect Correct answer

AI operates without monitoring

Correct Incorrect Correct answer

Users are excluded from decisions

Correct Incorrect Correct answer

Question 24 of 50

24. Question

1 point(s)

Why is model versioning important in production AI?

It allows teams to identify and reproduce specific model versions

Correct Incorrect Correct answer

It guarantees every model is unbiased

Correct Incorrect Correct answer

It removes the need for testing

Correct Incorrect Correct answer

It prevents all model failures

Correct Incorrect Correct answer

Question 25 of 50

25. Question

1 point(s)

What is the purpose of an audit trail in an AI system?

To increase model creativity

Correct Incorrect Correct answer

To replace user authentication

Correct Incorrect Correct answer

To record relevant actions and events for review and accountability

Correct Incorrect Correct answer

To remove system logs

Correct Incorrect Correct answer

Question 26 of 50

26. Question

1 point(s)

A company changes an AI model, its data source, and its business process simultaneously. What should be strengthened?

Change management and impact assessment

Correct Incorrect Correct answer

Screen resolution

Correct Incorrect Correct answer

Printer maintenance

Correct Incorrect Correct answer

Password formatting

Correct Incorrect Correct answer

Question 27 of 50

27. Question

1 point(s)

What does an AI pipeline typically connect?

Data preparation, training, evaluation, deployment, and monitoring stages

Correct Incorrect Correct answer

Only user passwords

Correct Incorrect Correct answer

Only hardware components

Correct Incorrect Correct answer

Only presentation slides

Correct Incorrect Correct answer

Question 28 of 50

28. Question

1 point(s)

What is model interpretability especially useful for?

Increasing storage capacity

Correct Incorrect Correct answer

Understanding factors associated with model outputs in contexts where explanation matters

Correct Incorrect Correct answer

Reducing network bandwidth

Correct Incorrect Correct answer

Generating longer responses

Correct Incorrect Correct answer

Question 29 of 50

29. Question

1 point(s)

Why is reproducibility important in advanced AI development?

Teams should be able to understand how a result or model version was produced

Correct Incorrect Correct answer

It makes every model identical

Correct Incorrect Correct answer

It eliminates the need for data

Correct Incorrect Correct answer

It guarantees future performance

Correct Incorrect Correct answer

Question 30 of 50

30. Question

1 point(s)

An AI model performs well before deployment but degrades after several months. What should an advanced team investigate first?

Data and concept drift

Correct Incorrect Correct answer

Monitor brightness

Correct Incorrect Correct answer

Keyboard response time

Correct Incorrect Correct answer

File naming conventions

Correct Incorrect Correct answer

Question 31 of 50

31. Question

1 point(s)

A high-stakes AI system has excellent average accuracy but performs significantly worse for one relevant group. What should the team do?

Report only the overall accuracy

Correct Incorrect Correct answer

Investigate subgroup performance and potential sources of disparity

Correct Incorrect Correct answer

Remove the affected group from testing

Correct Incorrect Correct answer

Increase the model's response length

Correct Incorrect Correct answer

Question 32 of 50

32. Question

1 point(s)

What is a model card commonly used for?

Storing user passwords

Correct Incorrect Correct answer

Documenting important information about a model, including intended use and limitations

Correct Incorrect Correct answer

Managing office equipment

Correct Incorrect Correct answer

Designing neural network hardware

Correct Incorrect Correct answer

Question 33 of 50

33. Question

1 point(s)

What is concept drift?

A change in the relationship between inputs and the target outcome over time

Correct Incorrect Correct answer

A change in computer hardware

Correct Incorrect Correct answer

A change in screen resolution

Correct Incorrect Correct answer

A change in file format only

Correct Incorrect Correct answer

Question 34 of 50

34. Question

1 point(s)

Why is accuracy alone sometimes insufficient for evaluating an AI system?

Different errors can have very different consequences

Correct Incorrect Correct answer

Accuracy is never useful

Correct Incorrect Correct answer

Accuracy always measures fairness

Correct Incorrect Correct answer

Accuracy automatically measures security

Correct Incorrect Correct answer

Question 35 of 50

35. Question

1 point(s)

What is the purpose of a data lineage record?

To increase token limits

Correct Incorrect Correct answer

To generate random training examples

Correct Incorrect Correct answer

To replace access controls

Correct Incorrect Correct answer

To track where data comes from and how it moves or changes

Correct Incorrect Correct answer

Question 36 of 50

36. Question

1 point(s)

Why should an AI system have defined performance thresholds?

They provide measurable criteria for acceptable system behavior

Correct Incorrect Correct answer

They guarantee zero errors

Correct Incorrect Correct answer

They eliminate the need for monitoring

Correct Incorrect Correct answer

They make training unnecessary

Correct Incorrect Correct answer

Question 37 of 50

37. Question

1 point(s)

What is calibration in predictive modeling?

The process of cleaning a monitor

Correct Incorrect Correct answer

A method for increasing token limits

Correct Incorrect Correct answer

The process of encrypting training data

Correct Incorrect Correct answer

The relationship between predicted probabilities and observed outcomes

Correct Incorrect Correct answer

Question 38 of 50

38. Question

1 point(s)

An organization wants to compare two AI models before deployment. Which approach is most defensible?

Choose the model with the longer name

Correct Incorrect Correct answer

Select the newest model automatically

Correct Incorrect Correct answer

Compare only their interface designs

Correct Incorrect Correct answer

Evaluate both using the same relevant test criteria and representative scenarios

Correct Incorrect Correct answer

Question 39 of 50

39. Question

1 point(s)

What is retrieval quality in a RAG system mainly concerned with?

Whether the retrieved information is relevant to the user's request

Correct Incorrect Correct answer

Whether the interface uses large fonts

Correct Incorrect Correct answer

Whether the model has many parameters

Correct Incorrect Correct answer

Whether the prompt contains emojis

Correct Incorrect Correct answer

Question 40 of 50

40. Question

1 point(s)

A model predicts a 90% probability for an event, but such predictions occur only 60% of the time. What should be investigated?

Probability calibration

Correct Incorrect Correct answer

File compression

Correct Incorrect Correct answer

Prompt length

Correct Incorrect Correct answer

Hardware drivers

Correct Incorrect Correct answer

Question 41 of 50

41. Question

1 point(s)

Why can latency be an important AI performance metric?

Delayed responses can reduce usability or disrupt time-sensitive workflows

Correct Incorrect Correct answer

Latency measures fairness directly

Correct Incorrect Correct answer

Latency determines model intelligence

Correct Incorrect Correct answer

Latency eliminates hallucinations

Correct Incorrect Correct answer

Question 42 of 50

42. Question

1 point(s)

A RAG system retrieves five irrelevant documents before finding the correct one. Which component should be investigated?

Screen hardware

Correct Incorrect Correct answer

Retrieval and ranking strategy

Correct Incorrect Correct answer

User password policy

Correct Incorrect Correct answer

File compression

Correct Incorrect Correct answer

Question 43 of 50

43. Question

1 point(s)

What is adversarial testing of an AI system?

Increasing training data automatically

Correct Incorrect Correct answer

Replacing every model component

Correct Incorrect Correct answer

Deliberately probing the system with challenging inputs to discover weaknesses

Correct Incorrect Correct answer

Removing security controls

Correct Incorrect Correct answer

Question 44 of 50

44. Question

1 point(s)

What is an AI system’s cost-performance trade-off?

Choosing the most expensive model automatically

Correct Incorrect Correct answer

Eliminating evaluation to save money

Correct Incorrect Correct answer

Balancing computational cost, response quality, speed, and operational requirements

Correct Incorrect Correct answer

Increasing model size without constraints

Correct Incorrect Correct answer

Question 45 of 50

45. Question

1 point(s)

Why can chunking strategy affect RAG performance?

It determines processor temperature

Correct Incorrect Correct answer

It eliminates the need for embeddings

Correct Incorrect Correct answer

It guarantees factual answers

Correct Incorrect Correct answer

It influences how information is divided and retrieved as context

Correct Incorrect Correct answer

Question 46 of 50

46. Question

1 point(s)

Why should prompt injection be considered in tool-using AI agents?

It always improves model performance

Correct Incorrect Correct answer

Untrusted content may attempt to manipulate instructions or actions

Correct Incorrect Correct answer

It only affects image quality

Correct Incorrect Correct answer

It eliminates the need for access control

Correct Incorrect Correct answer

Question 47 of 50

47. Question

1 point(s)

An organization needs an AI assistant for a high-volume customer service workflow. Which architecture should be evaluated?

A design balancing retrieval, model capability, latency, cost, monitoring, and escalation

Correct Incorrect Correct answer

The model with the largest parameter count only

Correct Incorrect Correct answer

A system without logging

Correct Incorrect Correct answer

A system with unrestricted tool access

Correct Incorrect Correct answer

Question 48 of 50

48. Question

1 point(s)

What is reranking in a retrieval pipeline?

Rewriting the original database

Correct Incorrect Correct answer

Reordering retrieved results according to their relevance

Correct Incorrect Correct answer

Retraining the entire model

Correct Incorrect Correct answer

Encrypting every prompt

Correct Incorrect Correct answer

Question 49 of 50

49. Question

1 point(s)

An AI agent reads an external webpage containing instructions telling it to reveal confidential data. What should the system do?

Treat the webpage as untrusted content and prevent unauthorized actions

Correct Incorrect Correct answer

Follow the webpage instructions

Correct Incorrect Correct answer

Give the webpage administrator access

Correct Incorrect Correct answer

Disable authentication

Correct Incorrect Correct answer

Question 50 of 50

50. Question

1 point(s)

What is continuous evaluation in an AI application?

Testing the model once before launch

Correct Incorrect Correct answer

Repeatedly testing performance as models, data, prompts, and usage conditions change

Correct Incorrect Correct answer

Removing all evaluation after deployment

Correct Incorrect Correct answer

Measuring only server temperature

Correct Incorrect Correct answer

Average score
Your score

Level 3–( AI ) Advanced | Artificial Intelligence

O.G. EDUCATION GROUP

ACADEMY OF OXFORD INTERNATIONAL ACADEMY OF GLASGOW FOR CULTURE & SCIENCES