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, where it answers user queries. Users report that the model sometimes generates plausible-sounding but factually incorrect or nonsensical information. What is this phenomenon called in generative AI?
- AOverfitting
- BBias
- CHallucination
- DUnderfitting
Show answer & explanationAnswer & explanation
Correct answer: C. Hallucination
Hallucination in generative AI, particularly large language models, refers to the phenomenon where the model generates content that is factually incorrect, nonsensical, or deviates from the provided source information, while still appearing coherent and plausible. This is a common challenge in deploying LLMs for factual tasks.
Why the other options are wrong
- A. Overfitting occurs when a model learns the training data too well, failing to generalize to new data, but not necessarily generating factually incorrect plausible statements.
- B. Bias in AI refers to systematic errors in the model's output due to biased data or algorithms, leading to unfair or prejudiced outcomes, which is different from factual inaccuracies.
- D. Underfitting occurs when a model is too simple to capture the underlying patterns in the data, leading to poor performance on both training and new data.
Hallucination (Generative AI)
The phenomenon where a generative AI model produces outputs that are plausible-sounding but factually incorrect, nonsensical, or not grounded in reality or the provided source data.
- Common challenge in Large Language Models (LLMs).
- Outputs appear coherent but lack factual accuracy.
- Can be reduced through techniques like RAG (Retrieval Augmented Generation) or fine-tuning with factual data.
- Distinguished from simply making a 'wrong prediction' by its coherent yet unfounded nature.
Memory trick: Hallucinating LLMs Make Up Facts