AWS Certified AI PractitionerFoundation ModelsMedium
A startup is developing an application that uses a foundation model to generate creative content, such as short stories, poems, and marketing taglines. Users have started complaining that the generated content, while grammatically correct, often repeats phrases, uses similar sentence structures, and lacks true originality across different prompts. Which common issue with generative foundation models is the startup encountering?
- AAlgorithmic bias.
- BUnderfitting.
- CGenerative repetitiveness.
- DHallucination.
Show answer & explanationAnswer & explanation
Correct answer: C. Generative repetitiveness.
The description 'often repeats phrases, uses similar sentence structures, and lacks true originality across different prompts' directly points to generative repetitiveness. This is a common issue where generative models, especially when not sufficiently diverse in their output or poorly prompted, tend to fall into predictable patterns.
Why the other options are wrong
- A. Algorithmic bias refers to unfair or discriminatory outcomes due to biased training data or model design, not lack of originality.
- B. Underfitting means the model is too simple and performs poorly on both training and new data, which is not the issue here as the content is 'grammatically correct' but unoriginal.
- D. Hallucination is when a model generates factually incorrect or nonsensical information, which is different from being repetitive.
Generative Repetitiveness
A common problem in generative foundation models where the output, despite being grammatically correct, lacks diversity and originality, frequently repeating phrases, sentence structures, or ideas.
- Often due to sampling strategies (e.g., greedy decoding).
- Can be mitigated by using diverse decoding methods like nucleus sampling or beam search with penalty.
- Reduces perceived creativity and usefulness of the generated content.
Memory trick: Repetitive generation lacks true creativity.