Q. Which of the following statements with regard to Large Language Models (LLMs) used in machine learning is/are correct?
I. LLMs assign probabilities to the next possible words and then pick the one with the highest probability.
II. LLMs process data through mathematical optimization to minimise prediction errors.
III. LLMs produce unbiased outputs.
Select the answer using the code given below:
1. (a) 1 only
2. (b) 1 and 2 only
3. (c) 2 and 3 only
4. (d) 1, 2 and 3

Answer : B

Explanation:

Statement 1 is correct: LLMs work like an advanced autocomplete which means they assign probabilities to all possible next words and predict the most likely one. This is the core mechanism behind how LLMs generate text.
Statement 2 is correct: During training, LLMs use mathematical optimization (minimising a "loss function") to reduce errors in their predictions. Every time they predict wrongly, their internal settings are adjusted to do better next time.
Statement 3 is not correct: LLMs are trained on internet text written by humans, which contains biases, stereotypes, and errors. So LLM outputs are not unbiased which meant they can reflect and even amplify the biases present in their training data.

Source:

https://trajectories.substack.com/p/things-to-know-about-large-language