AI and ML Fundamentals (AIF-C01)

Question 1

A financial company wants to detect fraudulent transactions in real-time. They have a historical dataset where each transaction is already labeled as 'fraud' or 'not fraud'. Which machine learning approach is most suitable?

Accepted Answer

Supervised Learning (Classification)

Answer

Unsupervised Learning (Clustering)

Answer

Reinforcement Learning

Answer

Supervised Learning (Regression)

Question 2

A computer vision model for autonomous vehicles was trained only with images of sunny roads. In testing, it consistently fails to detect pedestrians in the rain. What term best describes the cause of this failure?

Accepted Answer

Data Bias / Lack of generalization

Answer

Overfitting

Answer

Hallucination

Answer

Excessive feature engineering

Question 3

Which metric is critical for evaluating a cancer detection medical model, where the cost of missing a positive case (false negative) could cost the patient's life?

Accepted Answer

Recall (Sensitivity)

Answer

Precision

Answer

Accuracy

Answer

Specificity

Question 4

A retail company wants to predict the exact number of units it will sell next month based on ad spend and seasonality. What type of ML model is needed?

Accepted Answer

Regression

Answer

Multiclass Classification

Answer

Clustering

Answer

Sentiment Analysis

Question 5

What is the fundamental difference between 'Parameters' and 'Hyperparameters' in an ML model?

Accepted Answer

Parameters are learned by the model from data during training; hyperparameters are manually set by the human before training.

Answer

Parameters are set manually before training; hyperparameters are learned automatically.

Answer

Parameters control hardware; hyperparameters control software.

Answer

There is no difference, they are synonyms.

Question 6

A model has 99% performance on training data but only 60% on test data. What problem does this indicate?

Accepted Answer

Overfitting

Answer

Underfitting

Answer

Model Drift

Answer

Data Leakage

Question 7

Which phase of the ML lifecycle involves transforming raw data (like text or dates) into numerical formats the model can understand (like one-hot encoding)?

Accepted Answer

Feature Engineering / Preprocessing

Answer

Data Ingestion

Answer

Model Training

Answer

Model Evaluation

Question 8

During neural network training, what is the purpose of the 'Loss Function'?

Accepted Answer

Quantify the difference between the model's prediction and the actual value to guide optimization.

Answer

Accelerate inference speed.

Answer

Split data into training and validation.

Answer

Visualize results for the client.

Question 9

Which ML system is responsible for suggesting relevant products to a user in an online store based on their browsing history?

Accepted Answer

Recommendation System

Answer

Anomaly Detection

Answer

Computer Vision

Answer

Natural Language Processing

Question 10

What does it mean for an AI model, like GPT-4 or Claude 3, to be 'multimodal'?

Accepted Answer

It can process and understand different types of data (text, images, audio) at the same time.

Answer

It can run in multiple AWS regions simultaneously.

Answer

It speaks multiple languages.

Answer

It uses multiple learning algorithms.