Q. 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.
Select the answer using the code given below :
Exp) Option b is the correct answer.
Statement 1 is correct: Generative autoregressive LLMs fundamentally operate as probabilistic text predictors. Given a prompt, the model calculates a probability distribution across its vocabulary for the next token and selects the next word based on this probability.
Statement 2 is correct: During training, LLMs utilize mathematical optimization algorithms (such as Gradient Descent) to iteratively adjust internal weights. This process continuously minimizes a loss function to drastically reduce token-level prediction errors.
Statement 3 is incorrect: LLMs are trained on vast, human-generated internet datasets. Consequently, they inherently absorb and replicate historical, systemic, and cultural biases present in the source text. Achieving completely unbiased output is factually incorrect and remains an unresolved challenge in AI.
Source:) https://www.ibm.com/think/topics/large-language-models
https://www.ibm.com/think/topics/machine-learning
https://www.pib.gov.in/PressReleasePage.aspx?PRID=2071446®=3&lang=2
https://www.thehindubusinessline.com/opinion/how-llms-could-widen-digital-divide/article69295542.ece
https://www.thehindu.com/education/mathematicians-explain-ais-intelligence-its-all-about-patterns-not-thinking/article70670543.ece
