Artificial Intelligence: A Breakdown of Must-Know Buzzwords

Introduction

In the field of artificial intelligence now a days we are hearing various buzzwords. In this blog post we will familiar ourselves with AI subfields and how utilizing their set of techniques and theories solve particular types of problems.

Here we will familiarize ourselves with Machine Learning (ML), Deep Learning, Neural Network (NN), Natural Language Processing (NLP), Computer Vision (CV), AI Agent/ Agentic, RAG AI AI, CAG AI and Multi Model.

Artificial Intelligence

AI is making computers think and act like humans. Basically, it's about designing algorithms and systems so that machines can learn from data, adapt to new inputs and execute functions that mimic human cognitive abilities like language understanding, image interpretation and logical reasoning. In the other word AI learns from examples to help humans solve problems.

          .----------------.
          |   [  AI  ]     |
          |  Neural Net    |
          |  Connections   |
           '------||------'
                  ||
        .---------||---------.
        |        ||         |
        |  Data  ||  Learning|
        | Stream ||   Flow   |
        '---------||---------'
                  ||
           .------||------.
          |   Adaptive   |
          |    Systems   |
           '------------'

Following is some of the classic subfields.

Machine Learning

Focuses on algorithms that allow computers to learn patterns from data to make predictions or decisions. It has many branches such as supervised learning, unsupervised learning and reinforcement learning.

        .---------||---------.
        |        ||         |
        |  Data  || Features |
        | Input  || Extraction |
        '---------||---------'
                  ||
           .------||------.
          |  Model Selection |
          |  & Evaluation    |
           '------||------'
                  ||      
           .------------------.
          |[ Machine Learning ]|
          |  Model Training    |
          |  & Optimization    |
           '------||------'
                  ||        
           .------||------.
          |   Predictions   |
          |  & Deployment   |
           '------------'

Deep Learning

Deep Learning focuses on training neural networks with multiple layers to learn patterns and make decisions without human intervention. The "deep" in deep learning refers to the multiple layers in these neural networks.

        Input Layer
    .----------------.
    | Feature Input  |
    '----------------'
            ||
  .------------------.
  | Hidden Layer 1   |   <-- Extracts basic patterns
  '------------------'
            ||
  .------------------.
  | Hidden Layer 2   |   <-- Learns deeper features
  '------------------'
            ||
  .------------------.
  | Hidden Layer 3   |   <-- Identifies complex relationships
  '------------------'
            ||
  .------------------.
  | Output Layer     |   <-- Makes final decision/prediction
  '------------------'

Deep Learning: Multiple hidden layers allow the network to learn complex patterns

Neural Network

Neural Network is a computational system inspired by the structure and function of the human brain also the foundation of Deep Learning. It consists of interconnected neurons/nodes that process and learn from data.

        Input Layer
    .----------------.
    | Feature Input  |
    '----------------'
            ||
  .------------------.
  | Hidden Layer 1   |   <-- Extracts basic patterns
  '------------------'
            ||
  .------------------.
  | Hidden Layer 2   |   <-- Learns deeper features
  '------------------'
            ||
  .------------------.
  | Output Layer     |   <-- Makes final decision/prediction
  '------------------'

Neural Networks are the core of Deep Learning allowing AI to learn from data and improve over time.

Natural Language Processing (NLP)

Focuses on enabling communication between humans and machines. NLP combines linguistics and machine learning to help AI process human language in a way that is meaningful and useful.

        User Input
    .----------------.
    | "Hello, AI!"   |
    '----------------'
            ||
  .------------------.
  |  Tokenization    |   <-- Splitting text
  '------------------'
            ||
  .------------------.
  |  POS Tagging     |   <-- Identifying grammar
  '------------------'
            ||
  .------------------.
  |  Sentiment Analysis |   <-- Understanding emotion
  '------------------'
            ||
  .------------------.
  |  AI Response      |   <-- Generating reply
  '------------------'

Computer Vision (CV)

Enables computers to "see" and interpret visual information from the world, much like humans do i.e. helps machines interpret the world visually.

        Image Input
    .----------------.
    |  Captured Image |
    '----------------'
            ||
  .------------------.
  |  Preprocessing   |   <-- Noise reduction, resizing
  '------------------'
            ||
  .------------------.
  |  Feature Extraction  |   <-- Identifying edges, textures
  '------------------'
            ||
  .------------------.
  |  Object Detection  |   <-- Locating objects in image
  '------------------'
            ||
  .------------------.
  |  Classification  |   <-- Labeling objects
  '------------------'
            ||
  .------------------.
  |  AI Decision Making  |   <-- AI response (e.g., stop sign detected)
  '------------------'

AI Agent/ Agentic AI

An AI agent is a computer program that can sense its surroundings, think and take actions to achieve specific goals. Similar to human making choices based on what they see and hear.

      User or Environment
    .--------------------.
    |   Input Data       |  <-- AI gets information (text, images, sound)
    '--------------------'
              ||
    .--------------------.
    |   AI Thinking      |  <-- AI processes & decides what to do
    '--------------------'
              ||
    .--------------------.
    |   AI Takes Action  |  <-- AI responds or does a task
    '--------------------'
              ||
    .--------------------.
    |   AI Learns        |  <-- AI improves over time
    '--------------------'

On the other hand, Agentic AI is a type of artificial intelligence that is designed to act as an agent i.e. performing tasks or making decisions to solve problems and achieve goals.

      User Gives a Goal
    .----------------------.
    |   "Plan my trip"     |  <-- AI gets the goal
    '----------------------'
              ||
    .----------------------.
    |   AI Plans Steps     |  <-- AI figures out what needs to be done
    '----------------------'
              ||
    .----------------------.
    |   AI Takes Action    |  <-- AI books tickets, suggests hotels
    '----------------------'
              ||
    .----------------------.
    | AI Learns & Improves |  <-- AI remembers and refines future plans
    '----------------------'

RAG AI

It is an AI technique where instead of relying on what it was trained on, it searches for new relevant facts from external knowledge sources (e.g. databases, documents, or the web) and then generates a more accurate answer.

      User Asks a Question
    .----------------------.
    |  "Latest AI trends?" |  <-- User Input
    '----------------------'
              ||
    .----------------------.
    | AI Searches Sources  |  <-- Retrieves updated info
    '----------------------'
              ||
    .----------------------.
    | AI Generates Answer  |  <-- Uses retrieved data
    '----------------------'
              ||
    .----------------------.
    | AI Gives Response    |  <-- Provides accurate answer
    '----------------------'

CAG AI

Unlike RAG AI which is performing real-time retrieval of documents instead CAG preloads all relevant knowledge into the model's context. This preloaded data is stored in a key-value (KV) cache, allowing the AI to generate responses instantly without the latency or potential errors introduced by retrieval processes.

+-----------------+
|  Start          |
+-----------------+
         |
         v
+-----------------+
|  Define Dataset |  (e.g., HR policies, manuals)
+-----------------+
         |
         v
+-----------------+
|  Process Data   |  (Encode/compress data into cache)
+-----------------+
         |
         v
+-----------------+
|  Preload Cache  |  (Store in model’s context/memory)
+-----------------+
         |
         v
+-----------------+
|  User Asks      |  (e.g., “What’s the vacation policy?”)
|  Question       |
+-----------------+
         |
         v
+-----------------+
|  Generate       |  (Use preloaded cache, no retrieval)
|  Response       |
+-----------------+
         |
         v
+-----------------+
|  End            |
+-----------------+

Multi-Modal AI

Multimodal AI refers to artificial intelligence systems that can process, understand, and generate outputs using multiple types of data—called "modalities"—such as text, images, audio, video, or even numerical data. Here,

• "Multi" means multiple (more than one).
• "Modal" refers to modes of data (text, images, audio, video).
• So, Multi-Modal AI can process and combine different types of information.

      User Provides Input
    .----------------------.
    |  Text + Image + Audio |  <-- Different data types
    '----------------------'
              ||
    .----------------------.
    |  AI Processes Data    |  <-- Combines all input types
    '----------------------'
              ||
    .----------------------.
    |  AI Links Information |  <-- Connects text, image, or sound
    '----------------------'
              ||
    .----------------------.
    |  AI Generates Response|  <-- Intelligent answer using all inputs
    '----------------------'

Conclusion

AI holds immense potential to drive innovation and solve complex problems, but its success will depend on our ability to navigate its ethical implications and integrate it responsibly into society.