What is a key limitation of large language models like those used in GenAI?

Question

BlackTom AI · Accepted Answer

Large language models generate outputs probabilistically and can “hallucinate” — producing fluent but incorrect or incomplete answers.

1) Relevant knowledge points and concepts
- LLMs predict the next token based on patterns learned from training data; they do not have guaranteed factual grounding or true understanding. 
- Common limitations: hallucinations (plausible-sounding but wrong content), sensitivity to prompt/context, training-data gaps and biases, and lack of built-in verification or real-world experience. 
- These limits mean human review, testing, and domain expertise remain essential.

2) Step-by-step reasoning
- Because LLMs use statistical pattern matching rather than deterministic reasoning, they sometimes output statements that appear coherent yet are factually wrong or incomplete. 
- When asked for specialized engineering, medical, or legal answers, they may omit critical constraints or produce subtly incorrect calculations or code. 
- Therefore the key limitation being tested is their propensity to create plausible-looking but incorrect or nonsensical outputs.

3) Calculations/intermediate steps
- Not applicable here because the question is conceptual rather than numerical.

4) Why the correct answer is correct
- Option b directly describes hallucination and the probabilistic nature of model outputs, which is a widely documented limitation: LLMs can confidently present misinformation or incomplete solutions in a fluent form, requiring human verification.

5) Why the other options are incorrect
- a: False — LLMs can and often do produce working code snippets, though those may need correction. 
- c: False — Models do not truly understand engineering context as a human expert does; they lack real-world grounding and judgment. 
- d: False — LLMs do not eliminate the need for testing, validation, or engineering judgment; in fact, their outputs increase the need for careful review.

6) Conclusion
- The best choice is b: They can produce plausible-looking outputs that are incomplete, incorrect, or nonsensical. This matches the provided answer.

What is a key limitation of large language models like those used in GenAI?单项选择题

类似问题

Which of the following best describes Artificial Intelligence (AI)?

Which of the following is not a major force driving the rapid evolution of AI?

It is safe to assume that any content you receive from an AI-based tool is (check all that apply)

Microsoft CEO, Satya Nadella, said that "As __________ gets more efficient and accessible, we will see its use skyrocket, turning it into a commodity we just can't get enough of." What is he referring to?

Which of the following is an example of a high-velocity automated decision process?

Which of the following is an intelligent personal assistant?

更多留学生实用工具

智能学习助手

风格化写作助手

论文查重助手

文献引用助手

课堂转译助手

课堂笔记助手

Quiz搜索助手

学校历年真题

智能刷题助手

智能匹配练习

希望你的学习变得更简单