AWS Certified AI Practitioner AIF-C01 Practice Test 17
Question 160
A help center must choose an FM for short multilingual support replies. After selecting the recommended approach, which evidence MOST DIRECTLY demonstrates that the actual required outcome has been achieved?
- A component test invokes this alternate approach successfully: Choose cheapest regardless of language support
- Pilots meet target-language accuracy, latency and cost needs
- A reviewed design record approves this competing approach: Select by maximum output tokens alone
- A system inventory lists this competing approach as active: Always choose the largest parameter count
Explanation
Correct Answer
B:Pilots meet target-language accuracy, latency and cost needs. This checks the business result rather than just a configured feature.
Incorrect Answers
A:Price cannot compensate for missing languages A component invocation does not validate the necessary end-to-end outcome.
C:Long output capability does not establish short multilingual accuracy Design approval is not evidence that the alternative meets the actual requirement.
D:Size alone does not prove practical fit An active setting cannot independently prove the requested business outcome.
Question 161
A data scientist wants to experiment, train and deploy a custom time-series predictor. During a design review, someone proposes Use Bedrock Guardrails to perform custom training. Which technical objection is specifically relevant to this proposal?
- SageMaker AI supports the broader ML lifecycle
- Guardrails filter inputs and outputs rather than train models
- Config audits resource settings, not ML training
- Polly synthesizes speech rather than training arbitrary predictors
Explanation
Correct Answer
B:This is the actual failure mode of the proposed approach: Use Bedrock Guardrails to perform custom training. Guardrails filter inputs and outputs rather than train models
Incorrect Answers
A:SageMaker AI supports the broader ML lifecycle This explains the recommended design, but it is not the requested diagnosis of Use Bedrock Guardrails to perform custom training.
C:This limitation concerns Use AWS Config as feature engineering and model hosting; it does not diagnose the proposed Use Bedrock Guardrails to perform custom training approach.
D:This limitation concerns Use Polly as the general model-training platform; it does not diagnose the proposed Use Bedrock Guardrails to perform custom training approach.
Question 162
A data scientist wants to experiment, train and deploy a custom time-series predictor. A pilot uses this approach instead: Use Polly as the general model-training platform. Which limitation MOST DIRECTLY explains why the requirement is not met?
- SageMaker AI supports the broader ML lifecycle
- Polly synthesizes speech rather than training arbitrary predictors
- Config audits resource settings, not ML training
- Guardrails filter inputs and outputs rather than train models
Explanation
Correct Answer
B:This is the actual failure mode of the proposed approach: Use Polly as the general model-training platform. Polly synthesizes speech rather than training arbitrary predictors
Incorrect Answers
A:SageMaker AI supports the broader ML lifecycle This explains the recommended design, but it is not the requested diagnosis of Use Polly as the general model-training platform.
C:This limitation concerns Use AWS Config as feature engineering and model hosting; it does not diagnose the proposed Use Polly as the general model-training platform approach.
D:This limitation concerns Use Bedrock Guardrails to perform custom training; it does not diagnose the proposed Use Polly as the general model-training platform approach.
Question 163
A customer expects identical exact wording for every creative-answer request. Before approving a rollout, an architect suggests Add an embedding database to force identical prose. Which mismatch should be documented for this precise design?
- Sampling and model conditions can vary completions
- Retrieval does not force identical generation
- Generative outputs are not universally deterministic
- Length is not a reproducibility guarantee
Explanation
Correct Answer
B:This is the actual failure mode of the proposed approach: Add an embedding database to force identical prose. Retrieval does not force identical generation
Incorrect Answers
A:Sampling and model conditions can vary completions This explains the recommended design, but it is not the requested diagnosis of Add an embedding database to force identical prose.
C:This limitation concerns Promise byte-identical text from every call; it does not diagnose the proposed Add an embedding database to force identical prose approach.
D:This limitation concerns Assume longer prompts alone guarantee equality; it does not diagnose the proposed Add an embedding database to force identical prose approach.
Question 164
A help center must choose an FM for short multilingual support replies. Before approving a rollout, an architect suggests Select by maximum output tokens alone. Which mismatch should be documented for this precise design?
- Size alone does not prove practical fit
- Price cannot compensate for missing languages
- Real workload requirements determine model fit
- Long output capability does not establish short multilingual accuracy
Explanation
Correct Answer
D:This is the actual failure mode of the proposed approach: Select by maximum output tokens alone. Long output capability does not establish short multilingual accuracy
Incorrect Answers
A:This limitation concerns Always choose the largest parameter count; it does not diagnose the proposed Select by maximum output tokens alone approach.
B:This limitation concerns Choose cheapest regardless of language support; it does not diagnose the proposed Select by maximum output tokens alone approach.
C:Real workload requirements determine model fit This explains the recommended design, but it is not the requested diagnosis of Select by maximum output tokens alone.
Question 165
A support agent must recall prior steps for the same customer without exposing other customers. Before approving a rollout, an architect suggests Discard all prior context despite multi-turn requirements. Which mismatch should be documented for this precise design?
- No session state breaks follow-up references
- Session-scoped memory maintains conversational continuity without cross-user leakage
- Shared raw history violates isolation
- Sampling randomness cannot enforce access control
Explanation
Correct Answer
A:This is the actual failure mode of the proposed approach: Discard all prior context despite multi-turn requirements. No session state breaks follow-up references
Incorrect Answers
B:Session-scoped memory maintains conversational continuity without cross-user leakage This explains the recommended design, but it is not the requested diagnosis of Discard all prior context despite multi-turn requirements.
C:This limitation concerns Put every user's conversation in one unrestricted context; it does not diagnose the proposed Discard all prior context despite multi-turn requirements approach.
D:This limitation concerns Use model temperature to isolate sessions; it does not diagnose the proposed Discard all prior context despite multi-turn requirements approach.
Question 166
An internal search tool must find semantically related policies even when the wording differs. A pilot uses this approach instead: Compare character counts as semantic relevance. Which limitation MOST DIRECTLY explains why the requirement is not met?
- Temperature controls generation rather than retrieval
- Names miss paraphrased content
- Embedding vectors represent useful semantic relationships
- Documents of equal length may be unrelated
Explanation
Correct Answer
D:This is the actual failure mode of the proposed approach: Compare character counts as semantic relevance. Documents of equal length may be unrelated
Incorrect Answers
A:This limitation concerns Use sampling temperature as an index; it does not diagnose the proposed Compare character counts as semantic relevance approach.
B:This limitation concerns Rely only on exact filename search; it does not diagnose the proposed Compare character counts as semantic relevance approach.
C:Embedding vectors represent useful semantic relationships This explains the recommended design, but it is not the requested diagnosis of Compare character counts as semantic relevance.
Question 167
An internal search tool must find semantically related policies even when the wording differs. During a design review, someone proposes Rely only on exact filename search. Which technical objection is specifically relevant to this proposal?
- Embedding vectors represent useful semantic relationships
- Names miss paraphrased content
- Documents of equal length may be unrelated
- Temperature controls generation rather than retrieval
Explanation
Correct Answer
B:This is the actual failure mode of the proposed approach: Rely only on exact filename search. Names miss paraphrased content
Incorrect Answers
A:Embedding vectors represent useful semantic relationships This explains the recommended design, but it is not the requested diagnosis of Rely only on exact filename search.
C:This limitation concerns Compare character counts as semantic relevance; it does not diagnose the proposed Rely only on exact filename search approach.
D:This limitation concerns Use sampling temperature as an index; it does not diagnose the proposed Rely only on exact filename search approach.
Question 168
An insurance assistant must interpret an uploaded photo together with a text instruction. A pilot uses this approach instead: Use text-only inference without a visual description. Which limitation MOST DIRECTLY explains why the requirement is not met?
- File names are not evidence of damage in pixels
- Multimodal input processing covers both evidence types
- Text-only input cannot directly inspect the photo
- Speech recognition does not process picture content
Explanation
Correct Answer
C:This is the actual failure mode of the proposed approach: Use text-only inference without a visual description. Text-only input cannot directly inspect the photo
Incorrect Answers
A:This limitation concerns Use a filename-based regression score as the visual interpretation; it does not diagnose the proposed Use text-only inference without a visual description approach.
B:Multimodal input processing covers both evidence types This explains the recommended design, but it is not the requested diagnosis of Use text-only inference without a visual description.
D:This limitation concerns Use audio transcription as image understanding; it does not diagnose the proposed Use text-only inference without a visual description approach.
Question 169
A freight chatbot must quote rates that change daily. During a design review, someone proposes Increase maximum output length instead of retrieving prices. Which technical objection is specifically relevant to this proposal?
- Pretraining does rather than guarantee future price knowledge
- More generated words cannot fetch current data
- Tone does not supply factual freight-rate records
- Training snapshots may be stale
Explanation
Correct Answer
B:This is the actual failure mode of the proposed approach: Increase maximum output length instead of retrieving prices. More generated words cannot fetch current data
Incorrect Answers
A:Pretraining does not guarantee future price knowledge This explains the recommended design, but it is not the requested diagnosis of Increase maximum output length instead of retrieving prices.
C:This limitation concerns Rely only on a persuasive persona instruction; it does not diagnose the proposed Increase maximum output length instead of retrieving prices approach.
D:This limitation concerns Trust pretraining as current tomorrow; it does not diagnose the proposed Increase maximum output length instead of retrieving prices approach.
Question 170
A fixed-label fraud predictor is being replaced by a much larger text-generating FM without justification. A pilot uses this approach instead: Require all AI workloads to generate free text. Which limitation MOST DIRECTLY explains why the requirement is not met?
- Image generation does not provide a reliable binary score
- A constrained labeled classification may not need generative complexity
- Prediction is valid AI without content generation
- Size does not guarantee a fit for tabular fraud decisions
Explanation
Correct Answer
C:This is the actual failure mode of the proposed approach: Require all AI workloads to generate free text. Prediction is valid AI without content generation
Incorrect Answers
A:This limitation concerns Generate synthetic fraud illustrations as the decision itself; it does not diagnose the proposed Require all AI workloads to generate free text approach.
B:A constrained labeled classification may not need generative complexity This explains the recommended design, but it is not the requested diagnosis of Require all AI workloads to generate free text.
D:This limitation concerns Choose the largest FM without comparative tests; it does not diagnose the proposed Require all AI workloads to generate free text approach.
