AWS Certified AI PractitionerAI/ML and Generative AI FundamentalsMedium
An AI/ML team has developed a large language model (LLM) and deployed it for customer service. Users complain that the LLM sometimes generates plausible-sounding but factually incorrect responses, especially when queried about obscure topics or details not explicitly present in its training data. What is the term for this phenomenon in generative AI?
- AHallucination
- BBias
- COverfitting
- DUnderfitting
Show answer & explanationAnswer & explanation
Correct answer: A. Hallucination
The phenomenon where a generative AI model, particularly an LLM, produces outputs that are plausible-sounding but factually incorrect or nonsensical, especially when it lacks sufficient information, is known as 'hallucination'. This is a common challenge in generative models. Overfitting, bias, and underfitting refer to different types of model performance issues.
Why the other options are wrong
- B. Bias refers to systematic errors in the model's predictions, often due to skewed training data, leading to unfair or inaccurate outcomes for certain groups.
- C. Overfitting occurs when a model learns the training data too well, including noise, and performs poorly on unseen data.
- D. Underfitting occurs when a model is too simple to capture the underlying patterns in the training data, resulting in poor performance on both training and unseen data.
Hallucination (Generative AI)
The phenomenon where a generative AI model produces outputs that are plausible-sounding or grammatically correct but are factually incorrect, nonsensical, or not grounded in the input or training data.
- Often occurs when models lack sufficient information or are prompted with ambiguous queries.
- A significant challenge in ensuring the factual accuracy of generative AI outputs.
- Can be mitigated by techniques like retrieval augmented generation (RAG).
Memory trick: Generative models can hallucinate facts.