AWS Certified AI PractitionerFoundation ModelsMedium
A startup is building an application that uses a foundation model to generate creative content, such as marketing slogans and short stories. They notice that while the model often produces excellent results, it occasionally generates output that is factually incorrect or nonsensical, despite the prompt being clear. This phenomenon is commonly referred to as:
- ABias Amplification
- BOverfitting
- CUnderfitting
- DHallucination
Show answer & explanationAnswer & explanation
Correct answer: D. Hallucination
Hallucination in foundation models, particularly Large Language Models (LLMs), refers to the generation of plausible-sounding but factually incorrect or nonsensical information. This is a common challenge when models generate creative content without a strong grounding in truth.
Why the other options are wrong
- A. Bias amplification refers to the model exacerbating biases present in its training data, leading to unfair or discriminatory outputs.
- B. Overfitting occurs when a model learns the training data too well, failing to generalize to new data.
- C. Underfitting occurs when a model is too simple to capture the underlying patterns in the data.
LLM Hallucination
The phenomenon where a large language model generates plausible-sounding but factually incorrect, nonsensical, or unfaithful information.
- Common in generative models
- Can be hard to detect
- Often mistaken for factual accuracy
Memory trick: HALLUCINATIONS are like seeing things that aren't really there.