Topic 5 - Generative AI

Question 1

Generative AI (GenAI)

Answer 1

A type of artificial intelligence that creates new content based on patterns learned from existing data.

Question 2

Training

Answer 2

The process by which a Generative AI model learns from existing content to create a statistical model.

Question 3

Statistical Model

Answer 3

A model created during training that Generative AI uses to predict expected responses and generate new content.

Question 4

Discriminative Models

Answer 4

Traditional machine learning models that classify or predict based on labelled data.

Question 5

Generative Models

Answer 5

AI models that generate new data similar to their training data.

Question 6

Large Language Model (LLM)

Answer 6

A statistical language model, typically based on deep learning architectures, trained on a massive amount of data for generating and translating text and performing NLP tasks.

Question 7

Transformer

Answer 7

A deep learning architecture developed by Google in 2017, commonly used as the basis for LLMs.

Question 8

Vector

Answer 8

A single-dimensional array used to represent data numerically for machine processing.

Question 9

Tokenization

Answer 9

The process of breaking down input data into basic units called tokens.

Question 10

Tokens

Answer 10

Basic units of data processed by LLMs, which can be words, subwords, or characters.

Question 11

Context Length

Answer 11

The maximum number of tokens an LLM can accept as input and generate as output.

Question 12

Embeddings

Answer 12

Tokens that have semantic context, representing the meaning and relationships between words and phrases.

Question 13

Model Parameters

Answer 13

Settings that control the behaviour and output of a Generative AI model.

Question 14

Max Output Tokens

Answer 14

Specifies the maximum number of tokens allowed in the model’s response.

Question 15

Temperature

Answer 15

Controls the degree of randomness in token selection; lower values are more deterministic, higher values are more creative.

Question 16

Hallucinations

Answer 16

Nonsensical or grammatically incorrect words or phrases generated by a model.

Question 17

TopK

Answer 17

A parameter that changes how the model selects tokens for output by considering the K most probable tokens.

Question 18

TopP

Answer 18

A parameter that changes how the model selects tokens for output by selecting from the most probable tokens until their cumulative probability reaches the specified P value.

Question 19

Stop Sequence

Answer 19

A sequence of characters that tells the model to stop generating text.

Question 20

Prompt Design

Answer 20

The process of creating effective prompts to elicit desired responses from language models.

Question 21

Prompt

Answer 21

A natural language request submitted to a language model.

Question 22

Zero-shot Prompting

Answer 22

A prompt that provides no examples.

Question 23

One-shot Prompting

Answer 23

A prompt that includes a single example.

Question 24

Few-shot Prompting

Answer 24

A prompt that includes multiple examples.

Question 25

Text Classification

Answer 25

A task for models to categorize text into predefined classes (e.g., spam filtering, sentiment analysis).

Question 26

Extraction Prompts

Answer 26

Prompts designed to extract specific information from text (e.g., Named Entity Recognition, Question Answering).

Question 27

Summarization Tasks

Answer 27

Tasks that involve extracting the most important information from text to create a concise summary.

Question 28

Content Generation

Answer 28

The task of generating new text content for articles, blogs, descriptions, etc.