Quiz on Large Language Models

What type of data are large language models (LLMs) trained on?

• Labeled data only

• Unlabeled and self-supervised data

• Only numerical data

• Only images and sounds

What is the key benefit of a foundation model?

• It is trained on small datasets

• It learns from large amounts of unlabeled data to produce adaptable output

• It is designed for one specific task

• It requires no training

What is one example of "text-like things" that LLMs can process, as mentioned in the video?

• Videos

• Code

• Images

• Graphs

How much data does GPT-3 use during its training?

• 45 terabytes

• 1 petabyte

• 1 gigabyte

• 10 terabytes

How many parameters does GPT-3 use to process data?

• 175 million

• 1 billion

• 175 billion

• 10 billion

Which of the following is NOT a component of an LLM?

• Data

• Architecture

• Hardware

• Training

What role does the "transformer" architecture play in LLMs?

• To store data

• To define model architecture

• To help the model adjust its predictions

• To store words what is the answer to this questions

What happens during the training phase of an LLM?

• The model generates random outputs

• The model learns to predict the next word in a sentence

• The model adjusts its architecture

• The model analyzes images and video

What does fine-tuning allow an LLM to do?

• Improve its general understanding of text

• Focus on a specific task or domain for better performance

• Decrease its size and complexity

• Work with images instead of text

How does an LLM improve its predictions during training?

• It uses feedback from users

• It adjusts its parameters to reduce prediction errors

• It gathers new data from the internet

• It evaluates its performance without changes