GSCM 410 Final Exam flashcards

Question 1

What is the primary analogy used to describe AI in its foundational purpose?

Accepted Answer

AI is primarily described as a powerful productivity tool, similar to power tools used in construction or manufacturing, highlighting its capabilities and the need for skilled application.

Question 2

Who proposed the Turing Test, and what was its purpose?

Accepted Answer

Alan Turing proposed the Turing Test in 1950 as a benchmark for machine intelligence, where a human judge tries to distinguish between a human and a machine based on conversation.

Question 3

When and where was the term 'Artificial Intelligence' coined?

Accepted Answer

The term 'Artificial Intelligence' was coined in 1955/1956 at the Dartmouth Conference, marking the formal beginning of the field.

Question 4

What were 'Expert Systems' and what was their main limitation?

Accepted Answer

Expert Systems, developed in the mid-1980s, encoded human expertise into decision rules but were limited by the Polanyi Principle ('We know more than we can tell').

Question 5

What is 'Backpropagation' and why was it significant?

Accepted Answer

Backpropagation, rediscovered in the mid-1980s, is an algorithm using gradient descent to train Artificial Neural Networks (ANNs) by adjusting weights and biases to minimize error.

Question 6

How did GPUs impact AI development?

Accepted Answer

The development of GPUs, initially for video games, provided the massive parallel computation needed to train complex neural networks, leading to a resurgence in AI research.

Question 7

What innovation did the 'Transformer Architecture' introduce in 2017?

Accepted Answer

The Transformer Architecture introduced the 'attention' mechanism, enabling models to weigh the importance of different data elements and process sequences in parallel, surpassing RNNs and LSTMs.

Question 8

What is the difference between Machine Learning (ML) and Deep Learning?

Accepted Answer

Machine Learning (ML) is an AI approach where models learn from data, while Deep Learning is a subset of ML that uses Artificial Neural Networks (ANNs) with many hidden layers to detect complex patterns.

Question 9

What distinguishes Generative AI (GenAI) from other AI models?

Accepted Answer

Generative AI (GenAI) models are capable of creating new content (text, images, video), whereas Pattern Recognition Models are designed to classify or analyze existing input.

Question 10

What is Artificial General Intelligence (AGI)?

Accepted Answer

Artificial General Intelligence (AGI) is the hypothetical future state of AI possessing human-like understanding and reasoning across diverse tasks, even those outside its training data.

Question 11

How is the Fourth Industrial Revolution characterized in terms of AI's impact on jobs?

Accepted Answer

The Fourth Industrial Revolution, driven by AI and intelligent automation, uniquely threatens professional and knowledge-worker jobs, as AI can perform complex decision-making tasks.

Question 12

What is a 'Perceptron' in the context of ANNs?

Accepted Answer

A Perceptron is the basic computational unit of an ANN, mimicking a biological neuron by taking numerical inputs, processing them with weights and a bias, and producing an output via an activation function.

Question 13

What is the role of 'weights' and 'biases' in a perceptron?

Accepted Answer

Weights represent the importance of each input, and biases act as a threshold; both are numerical values adjusted during training to minimize prediction errors.

Question 14

What is the primary goal of training an Artificial Neural Network (ANN)?

Accepted Answer

The primary goal of training an ANN is to minimize 'loss,' the difference between the model's predicted output and the correct 'ground truth' output, by adjusting its weights and biases.

Question 15

What is 'Model Overfitting' and how is it typically avoided?

Accepted Answer

Model Overfitting occurs when a model memorizes training data and fails to generalize to new data; it's avoided by using a slower training rate and diverse data over multiple epochs.

Question 16

What is 'Self-supervised Learning' in the context of LLM pre-training?

Accepted Answer

Self-supervised Learning is used in pre-training where the model learns from unlabeled data by predicting masked words or the next word in a sequence, generating its own learning signals.

Question 17

What is 'Reinforcement Learning from Human Feedback (RLHF)'?

Accepted Answer

RLHF is a fine-tuning process where human ratings of model outputs are used to adjust the model's weights, reinforcing desired behaviors and reducing undesirable ones like sycophancy.

Question 18

What is 'Emergent Behavior' in LLMs?

Accepted Answer

Emergent Behavior refers to capabilities that arise unexpectedly from an LLM's scale and training, rather than being explicitly programmed, contributing to their advanced abilities.

Question 19

What is the purpose of 'positional encoding' in the LLM inference flow?

Accepted Answer

Positional encoding is added to embedding vectors to preserve the order of tokens, ensuring the model understands the sequence of words, which is crucial for interpreting meaning.

Question 20

What is an 'AI System' versus an 'AI Model'?

Accepted Answer

An AI Model (like an LLM) is the 'brain,' while an AI System is the complete package including the model, user interface, pre/postprocessing, APIs, and databases that make it usable.

Question 21

What is the 'System Role' in API interaction with LLMs?

Accepted Answer

The System Role provides instructions to the LLM on how it should behave or what persona it should adopt, crucial for guiding responses and enforcing safety, often hidden from the end user.

Question 22

How does 'Inference Temperature' affect LLM output?

Accepted Answer

Inference Temperature controls the randomness of the LLM's output; a higher temperature leads to more varied output, while a lower temperature makes it more deterministic and focused.

Question 23

What problem does Retrieval Augmented Generation (RAG) solve?

Accepted Answer

RAG solves the limitation of LLMs not having real-time information or access to private knowledge bases by enabling them to retrieve and incorporate external documents before generating a response.

Question 24

How does RAG work at inference time?

Accepted Answer

At inference, a prompt is embedded, a semantic search finds relevant document chunks in a vector database, and these chunks are provided to the LLM along with the prompt for a grounded response.

Question 25

What is a 'gating network' in a Mixture of Experts (MoE) architecture?

Accepted Answer

A gating network in an MoE model evaluates an input and determines which specialized 'expert' sub-models are most suitable to process it, directing the computation.

Question 26

What is an 'AI Agent'?

Accepted Answer

An AI agent is an autonomous AI system capable of performing multi-step tasks with minimal human intervention by making decisions, executing actions, and adapting based on outcomes.

Question 27

What is a 'data workflow' in the context of AI agents?

Accepted Answer

A data workflow is a sequence of steps defining which models, APIs, databases, or human interventions are required to accomplish a task for an AI agent.

Question 28

What are 'Multimodal AI Models'?

Accepted Answer

Multimodal AI Models can handle and process multiple types of data (modalities) simultaneously, such as text, images, audio, and video, enabling more integrated applications.

Question 29

What is 'Vibe Coding' in AI tool development?

Accepted Answer

Vibe Coding allows users to generate entire application code from natural language descriptions, offering high customization and enabling non-programmers to create complex applications.

Question 30

What is a 'reasoning model' in AI?

Accepted Answer

A reasoning model performs an internal analysis and considers various approaches before generating a response, often by producing intermediate 'thinking' tokens to improve output quality.

Question 31

What is 'Chain-of-Thought (CoT) Prompting'?

Accepted Answer

CoT Prompting instructs an AI model to reason step-by-step before providing an answer, improving accuracy on complex tasks and allowing verification of the model's logic.

Question 32

What is 'AI sycophancy'?

Accepted Answer

AI sycophancy is the tendency of an AI model to agree with or flatter users, even when their input is incorrect, representing an alignment failure that prioritizes user satisfaction over accuracy.

Question 33

What is the main difference between proprietary and open source models?

Accepted Answer

Proprietary models keep their source code secret and are accessed via APIs or executables, while open source models make their source code publicly available for modification and redistribution.

Question 34

What is 'transfer learning' in AI?

Accepted Answer

Transfer learning is an efficient technique where a pre-trained model's weights are used as a starting point and then fine-tuned for a specific task, leveraging existing knowledge.

Question 35

What is LMOps (Language Model Operations)?

Accepted Answer

LMOps is the discipline focused on deploying and running LLMs on various hardware, analogous to DevOps, encompassing technical work like hardware provisioning and software configuration.

Question 36

What are 'model parameters' and how do they relate to model size?

Accepted Answer

Model parameters are the weights of an ANN and are a primary measure of model size; generally, more parameters mean broader knowledge but higher computational costs.

Question 37

What is the 'context window' of an LLM?

Accepted Answer

The context window defines the maximum number of tokens an LLM can process at one time, including both input and output, affecting its ability to handle complex information.

Question 38

What is 'Parameter-Efficient Fine-Tuning (PEFT)'?

Accepted Answer

PEFT techniques, like LoRA, update only a subset of a model's parameters during fine-tuning, making it more efficient and cost-effective than full fine-tuning.

Question 39

What is 'model quantization'?

Accepted Answer

Model quantization reduces the numerical precision of model parameters (e.g., from 32-bit to 8-bit) to shrink model size and reduce computational cost, often with minor quality tradeoffs.

Question 40

What are the four key dimensions for assessing AI system performance?

Accepted Answer

The four key dimensions are accuracy, efficiency, scalability, and usability, providing a comprehensive framework for evaluating AI system performance.

Question 41

What is the purpose of 'ground truth' in accuracy metrics?

Accepted Answer

Ground truth is the known correct output or label used as a reference to evaluate the correctness of an LLM's generated output.

Question 42

What are 'concurrency' and 'throughput' in scalability metrics?

Accepted Answer

Concurrency measures the number of simultaneous requests an LLM system can handle, while throughput measures the number of queries it can process per second.