Artificial Intelligence Quiz on ML and LLMs

Exam A

• Question 1: A company uses ML models for quarterly forecasts. An Al practitioner needs to report on trained ML models, including transparency and explainability for stakeholders. The report should include partial dependence plots (PDPs).

• Question 2: A law firm wants an Al application using LLMs to extract key points from legal documents. The best solution is a summarization chatbot.

• Question 3: A company needs an ML algorithm to classify human genes into 20 categories based on characteristics. Decision trees are suitable for documenting the inner mechanisms of the model's output.

• Question 4: A company has an image classification model for plant diseases. The best metric to evaluate the model's performance is accuracy.

Exam B

• Question 1: A company is training a foundation model (FM) to improve accuracy. Increasing the number of epochs will help.

• Question 2: A company wants to evaluate the impact of an LLM chatbot on website engagement rate.

• Question 3: Amazon SageMaker Clarify provides tools to monitor the quality of ML models in production.

• Question 4: A company has a model that performs well on training data but poorly in production. Increasing the volume of training data can help.

Exam C

• Question 1: A prompting attack that directly exposes the configured behavior of an LLM is extracting the prompt template.

• Question 2: A company using Amazon Bedrock needs to review security aspects, focusing on securing data in transit and at rest.

• Question 3: A social media company needs an algorithm to evaluate the toxicity of model responses. Crowd-sourced evaluation offers the simplest method.

• Question 4: A company is testing a foundation model's security. A fuzzing technique for testing inputs to identify vulnerabilities is the correct method.

This quiz covers various applications of artificial intelligence, focusing on machine learning models, foundation models, and legal document summarization using large language models (LLMs). Explore concepts such as transparency, accuracy metrics, and algorithm classification in real-world scenarios. Test your knowledge on state-of-the-art AI techniques and their implications.