Which of the following statements with regard to Large Language Models (LLMs) used in machine learning is/are correct?

1. LLMs assign probabilities to the next possible words and then pick the one with the highest probability.

2. LLMs process data through mathematical optimization to minimise prediction errors.

3. LLMs produce unbiased outputs.

Statement 1 is CORRECT: At their core, LLMs are advanced statistical text-prediction engines. When generating text, they calculate a probability distribution over their entire vocabulary for what the next token (word or piece of a word) should be, given the preceding context. While they don't always pick the absolute highest probability word (they use parameters like "temperature" to introduce creativity), the fundamental mechanism relies on assigning these probabilities.

Statement 2 is CORRECT: During the training phase, LLMs use mathematical optimization algorithms (such as Stochastic Gradient Descent) to adjust billions of internal weights. The goal of this process is to minimize a "loss function," which mathematically measures the difference between the model's predictions and the actual data, thereby minimizing prediction errors over time.

Statement 3 is INCORRECT: LLMs do not produce unbiased outputs. Because they are trained on massive datasets scraped from the internet, books, and articles, they inherently absorb and replicate the human biases, stereotypes, and prejudices present in that training data. Eliminating bias in AI remains one of the biggest ongoing challenges in machine learning research