Understanding AI: The LM Glossary

Artificial Intelligence (AI) is a rapidly evolving field, and keeping up with its terminology can be challenging. To simplify the learning process, we've categorized the most frequently used AI terms into five main territories, each represented by a different color. This structured approach ensures a logical flow of information, helping both beginners and experts navigate AI concepts effectively.

Blue Zone: AI Model Types, Architectures, and Sizes

The blue zone covers the fundamental aspects of AI models, including their types, architectures, and sizes. Terms in this category include:

  • Large Language Model (LLM) – AI models designed to understand and generate human-like text.
  • Transformer – A deep learning architecture that powers most modern AI models.
  • Parameters – The numerical values that determine a model’s behavior and learning capabilities.
  • Fine-Tuning – The process of adjusting a pre-trained model to improve performance on a specific task.

Yellow Zone: Interacting with and Guiding AI Models

The yellow zone highlights how we engage with AI systems, including input methods and instruction techniques:

  • Prompt Engineering – Crafting effective prompts to get desired responses from AI.
  • Temperature – A setting that controls the randomness of AI-generated responses.
  • Zero-Shot Learning – When an AI model responds to a task it has never seen before.
  • Few-Shot Learning – Providing examples in the prompt to guide AI toward better results.

Purple Zone: AI Model Processing and Output Generation

This zone focuses on how AI models handle input and create outputs:

  • Tokenization – The process of breaking down text into smaller units (tokens) for AI processing.
  • Context Window – The amount of text an AI model can consider at once.
  • Logits – The raw scores AI generates before converting them into probabilities.
  • Beam Search – A technique used to generate coherent and high-quality text.

Pink Zone: Model Training, Tuning, and Optimization

AI models require extensive training and fine-tuning to perform efficiently. The pink zone covers key training-related terms:

  • Pretraining – The initial phase where AI learns from vast datasets before fine-tuning.
  • Gradient Descent – An optimization algorithm that adjusts AI parameters to minimize errors.
  • Batch Size – The number of training examples processed at once.
  • Loss Function – A metric used to evaluate how well an AI model is learning.

Green Zone: Connecting AI Models with External Knowledge (RAG)

The green zone explores methods for integrating AI models with external knowledge sources:

  • Retrieval-Augmented Generation (RAG) – A technique that enhances AI responses by fetching relevant external data.
  • Embedding – Representing words or concepts as numerical vectors for efficient retrieval.
  • Knowledge Graph – A structured way of organizing information to improve AI reasoning.
  • Vector Database – A specialized database that stores and retrieves high-dimensional data efficiently.


Comments

Post a Comment

Popular posts from this blog

Understanding Tokenization: The Cornerstone of Large Language Models (LLMs)

Image
 In the world of Large Language Models (LLMs), tokenization is a fundamental yet fascinating process. But what exactly is tokenization, and why is it so important for LLMs to function effectively? What is Tokenization? Tokenization is the process of breaking text into smaller, manageable units called tokens. These tokens can represent words, subwords, or even individual characters. For example, the word "tokenization" might be split into smaller subwords such as "token" and "ization." This step transforms raw text into a structured format that LLMs can process. Since LLMs cannot directly comprehend raw text, tokenization acts as a bridge, converting human-readable text into sequences of numbers that the model understands. Why is Tokenization Important in LLMs? 1. Facilitating Text Understanding Tokenization ensures that a language model can interpret text input by mapping tokens to numerical representations. This allows the model to "read" and pr...
Read more

Agents vs. Models: Understanding the Key Differences in AI

As artificial intelligence continues to advance, the terms models and agents are often used interchangeably. However, they serve distinct roles in the AI ecosystem. Understanding the difference between AI models and AI agents is crucial for anyone looking to harness the full potential of generative AI. What is an AI Model? An AI model is a trained system that makes predictions or generates responses based on input data. These models rely solely on the information available in their training datasets. Whether it’s a large language model (LLM) like GPT or a fine-tuned transformer, AI models execute a single inference at a time without retaining memory of past interactions unless specifically designed to do so. Key Characteristics of AI Models: 🔹 Limited Knowledge Scope : The model's understanding is restricted to the data it was trained on. 🔹 Single-Step Predictions : AI models make isolated predictions per query. 🔹 No Built-In Tool Access : Models do not inherently interact wi...
Read more