Check out the links below to read some of the papers discussed in this section.
When we're talking about Generative AI, we tend to use a definition like IBM's:
"Generative AI refers to deep-learning models that can generate high-quality text, images, and other content based on the data they were trained on." (Martineau, 2023).
Modern AI platforms are built from groups of algorithms, each with distinct purposes. At a basic level, these systems use training data and machine learning algorithms to “learn” patterns, relationships, or structures. These learned patterns are stored in a model, which can then generate outputs such as text, images, or predictions. Below is a glossary of foundational terms used to describe modern AI systems. For more in-depth explanations, see the IBM guide to AI.
| Term | Definition |
|---|---|
| Core Concepts and Terminology | |
| Artificial Intelligence (AI) | A broad field of computer science focused on creating systems capable of tasks that typically require human intelligence—such as reasoning, learning, and problem solving. |
| Machine Learning (ML) | A subfield of AI focused on building algorithms that learn from data to make predictions or decisions without being explicitly programmed. |
| Deep Learning | A type of machine learning that uses deep neural networks—models with many layers—to recognize complex patterns in large datasets. |
| Neural Networks | Computational models inspired by the human brain, composed of interconnected “nodes” or “neurons” that process information in layers. These are the foundation of many modern AI systems. |
| Training Data | The dataset used to "teach" a machine learning model by exposing it to inputs and (in some cases) desired outputs. |
| Model | The resulting mathematical structure or system after training, which can be used to generate predictions or outputs based on new input data. |
| Types of Learning | |
| Supervised Learning | A type of machine learning where the training data is labeled, and the algorithm learns to map inputs to known outputs. |
| Unsupervised Learning | A method where the algorithm explores patterns or groupings in unlabeled data on its own. |
| Reinforcement Learning | A technique where an AI learns by interacting with an environment, receiving rewards or penalties based on its actions. |
| Fine-Tuning | A process of further training a pre-existing model on a smaller, task-specific dataset to improve performance in a particular use case. |
| Language and Generative Models | |
| Natural Language Processing (NLP) | A branch of AI that focuses on enabling machines to interpret, generate, and interact using human language. |
| Large Language Models (LLMs) | A type of AI model (e.g., GPT-4) trained on massive datasets of text to understand and generate human-like language. |
| Generative AI | AI systems that can create new content—such as text, images, music, or code—in response to prompts. Examples include ChatGPT, DALL·E, and Midjourney. |
| Other Key Terms | |
| Parameters | The internal values in a model that are adjusted during training and define how the model processes information. |
| Tokens | Chunks of text (often words or word fragments) used by language models to break down and process input and output. |
| Prompt | The user-provided input (text, code, etc.) that an AI model uses to generate a response. |