Advanced AI and Data Science

1. What is a 'recurrent neural network' (RNN) designed to handle?

• A. Sequential data, such as time series or text ✓
• B. Only image data
• C. Only static data
• D. Only database queries

💡 RNNs are designed to process sequential data, like time series or natural language, by maintaining memory of previous inputs.

2. What is 'transfer learning' in machine learning?

• A. Applying knowledge gained from one task to a different but related task ✓
• B. Training a model from scratch every time
• C. A type of database
• D. A type of encryption

💡 Transfer learning applies knowledge already learned from one task to help solve a different but related task.

3. What is the 'Turing Test' designed to evaluate?

• A. Whether a machine can exhibit intelligent behavior indistinguishable from a human ✓
• B. The speed of a computer
• C. The storage capacity of a computer
• D. The security of a network

💡 The Turing Test, proposed by Alan Turing, evaluates whether a machine can exhibit behavior indistinguishable from a human's.

4. What is 'few-shot learning'?

• A. A machine learning approach where a model learns to perform a task from only a small number of examples ✓
• B. Training with a massive amount of labeled data only
• C. A type of database
• D. A type of encryption

💡 Few-shot learning enables a model to learn a new task effectively from only a small number of training examples.

5. What is 'backpropagation' in neural networks?

• A. An algorithm for training neural networks by propagating error gradients backward through the network ✓
• B. A type of data storage
• C. A type of encryption
• D. A type of database query

💡 Backpropagation trains neural networks by computing error gradients and propagating them backward to adjust weights.

6. What does an 'attention mechanism' allow a neural network to do?

• A. Focus on relevant parts of the input data when making predictions ✓
• B. Increase processing speed only
• C. Reduce memory usage only
• D. Encrypt data

💡 An attention mechanism allows a neural network to dynamically focus on the most relevant parts of its input when making predictions.

7. What is a 'generative adversarial network' (GAN)?

• A. A system of two neural networks contesting with each other to generate realistic data ✓
• B. A type of database
• C. A type of encryption
• D. A single neural network

💡 A GAN consists of two neural networks, a generator and discriminator, competing against each other to produce realistic synthetic data.

8. What does 'gradient descent' optimize in machine learning?

• A. The parameters of a model to minimize a loss function ✓
• B. The speed of data transfer
• C. The size of a database
• D. The security of a network

💡 Gradient descent is an optimization algorithm that adjusts model parameters to minimize a loss function.

9. What does 'overfitting' indicate about a machine learning model?

• A. It performs well on training data but poorly on new, unseen data ✓
• B. It performs poorly on all data
• C. It is too simple to capture patterns
• D. It requires more data to run

💡 Overfitting occurs when a model learns training data too specifically, hurting its ability to generalize to new data.

10. What does 'model interpretability' refer to?

• A. The degree to which a human can understand the cause of a model's decision ✓
• B. The speed of a model
• C. The size of a model
• D. The cost of training a model

💡 Model interpretability describes how easily a human can understand why a model made a particular decision or prediction.

11. What does 'explainable AI' (XAI) aim to achieve?

• A. Making AI decision-making processes understandable to humans ✓
• B. Making AI models faster
• C. Making AI models smaller
• D. Making AI models cheaper to train

💡 Explainable AI aims to make the decision-making processes of AI systems transparent and understandable to humans.

12. What is a 'transformer' architecture in AI, notably used in models like GPT?

• A. A neural network architecture relying on self-attention mechanisms, particularly effective for language tasks ✓
• B. A type of database
• C. A type of hardware
• D. A type of encryption

💡 The transformer architecture, relying on self-attention mechanisms, revolutionized natural language processing and powers models like GPT.

13. What is a 'loss function' in machine learning?

• A. A function that measures how well a model's predictions match the actual outcomes ✓
• B. A function that deletes data
• C. A function that encrypts data
• D. A function that speeds up training only

💡 A loss function quantifies the difference between a model's predictions and the actual, correct outcomes.

14. What does 'data augmentation' involve?

• A. Artificially expanding a training dataset by creating modified versions of existing data ✓
• B. Deleting unnecessary data
• C. Encrypting data
• D. Compressing data

💡 Data augmentation artificially expands a training dataset by creating modified variations of existing data samples.

15. What is a 'convolutional neural network' (CNN) primarily used for?

• A. Processing grid-like data such as images ✓
• B. Processing only text data
• C. Processing only audio data
• D. Managing databases

💡 CNNs are specialized neural networks particularly effective at processing grid-like data, such as images.

16. What does 'bias' refer to in the context of AI and machine learning models?

• A. Systematic errors that result in unfair outcomes, often reflecting biases in training data ✓
• B. A type of hardware issue
• C. A type of encryption
• D. A type of database index

💡 AI 'bias' refers to systematic errors leading to unfair outcomes, often stemming from biased or unrepresentative training data.

17. What is 'federated learning'?

• A. A machine learning approach where models are trained across decentralized devices without sharing raw data ✓
• B. Training a model on a single centralized server only
• C. A type of encryption
• D. A type of database

💡 Federated learning trains machine learning models across many decentralized devices, without requiring raw data to be centrally shared.

18. What is 'feature engineering' in machine learning?

• A. The process of selecting and transforming variables to improve model performance ✓
• B. Writing the model's code
• C. Training a model
• D. Testing a model's accuracy

💡 Feature engineering involves selecting, creating, and transforming input variables to improve a model's predictive performance.

19. What is 'hyperparameter tuning' in machine learning?

• A. Adjusting the configuration settings of a model to optimize its performance ✓
• B. Cleaning the training data
• C. Writing the model's code
• D. Testing the model on new data

💡 Hyperparameter tuning involves adjusting a model's configuration settings, not learned from data, to optimize performance.

20. What does 'AGI' stand for in AI discussions?

• A. Artificial General Intelligence ✓
• B. Automated General Interface
• C. Advanced Graphic Interface
• D. Applied General Intelligence

💡 AGI stands for Artificial General Intelligence, referring to hypothetical AI with human-level cognitive abilities across many domains.