Artificial Intelligence (AI) is a field of computer science dedicated to simulating human intelligence in machines. It allows computers to perform tasks that typically require human cognition, such as visual perception, speech recognition, and decision-making.
To understand the landscape, it’s helpful to see how these terms nested:
Artificial Intelligence
├── Machine Learning
│ ├── Deep Learning
│ │ ├── Generative AI
│ │ ├── Foundation Models
│ │ └── Large Language Models (LLMs)
│ └── Classical ML (e.g., SVMs, decision trees)
└── AI Agents
└── Agentic AI (autonomous, goal-driven agents)
ML is a subset of AI where systems learn from data to make decisions or predictions without being explicitly programmed for every scenario.
Classic Machine Learning Types:
- Supervised Learning: Learning from labeled data (e.g., "This image is a cat").
- Unsupervised Learning: Finding hidden patterns in unlabeled data (e.g., grouping customers by behavior).
- Reinforcement Learning: Learning through trial and error to maximize a reward (e.g., an AI winning a chess game).
Deep Learning (DL) is a specialized subset of Machine Learning that is inspired by the structure and function of the human brain. While traditional machine learning often requires a human to manually "extract" features (like telling the computer to look for ears or a tail to identify a cat), deep learning models can learn these features automatically from raw data
At the heart of deep learning are Artificial Neural Networks. These are layers of interconnected "neurons" (mathematical functions) that process information.
- Input Layer: Receives the raw data (e.g., pixels of an image).
- Hidden Layers: This is where the "Deep" in Deep Learning comes from. There can be dozens or even hundreds of these layers. Each layer identifies increasingly complex patterns—the first might find edges, the next shapes, and the final layers recognize faces or objects.
- Output Layer: Provides the final prediction or classification.
Different types of deep learning models are used for different jobs:
- CNNs (Convolutional Neural Networks): Best for images and video.
- RNNs (Recurrent Neural Networks): Best for sequential data like text or speech.
- Transformers: The modern gold standard for language understanding and generation.
Generative AI is a branch of Deep Learning (DL) focused on creating new content rather than just classifying existing data.
Traditional AI is like a judge: it looks at a photo and decides, "That is a cat." Generative AI is like an artist: you tell it, "Draw me a cat wearing a space suit," and it creates a brand-new image that never existed before.
- Generative Adversarial Networks (GANs): Two networks (a "Generator" and a "Discriminator") compete to create realistic data.
- Variational Autoencoders (VAEs): Compresses data and then reconstructs it to generate new versions.
- Transformers: Large language models like GPT (Generative Pre-trained Transformer) use self-attention mechanisms to generate human-like text.
- Diffusion Models: They start with "noise" and gradually refine it into a high-quality image.
A Foundation Model is a massive AI model trained on a vast amount of data (terabytes) that can be adapted (fine-tuned or prompted) for many downstream tasks — such as chatbots, translation, coding, search, image generation, and more.
When these models focus on text, we call them Large Language Models (LLMs).
| Company | Foundation Model | Modality | Public Access |
|---|---|---|---|
| OpenAI | GPT-4o / GPT-5 | Text, Code, Image, Audio | API (OpenAI, Azure) |
| Google DeepMind | Gemini 1.5 (successor to PaLM 2) | Text, Code, Image, Video, Audio | API (Gemini Pro) |
| Anthropic | Claude 3/4 (Opus, Sonnet, Haiku) | Text | API (Claude.ai) |
| Meta | LLaMA 3 | Text | Open weights |
| Cohere | Command R+ | Text + RAG | API |
| IBM + Hugging Face | Granite | Text | Enterprise |
| xAI (Elon Musk) | Grok | Text | X (Twitter) |
| NVIDIA | Nemotron | Text | Enterprise & fine-tuning |
| AWS (Amazon) | Titan family | Text, Embedding, Image |
An LLM is a type of AI trained on a massive scale of text data (trillions of words) to understand, summarize, and generate human-like language.
What makes it “large”?
- It’s trained on huge amounts of text (books, articles, code, websites).
- It has billions of parameters (internal values that help it learn patterns in language).
How does an LLM work?
- It doesn’t “think” like a human.
- It predicts the next word (token) based on context.
By doing this repeatedly, it forms coherent sentences and ideas.
An AI agent is a system that observes its surroundings, makes decisions, and performs actions to reach a goal.
Traditional AI typically operates as a passive tool that provides "one-shot" responses to specific prompts; it is stateless, meaning it treats every request as a brand-new interaction without remembering what happened before. In contrast, AI Agents are goal-driven and stateful, maintaining a continuous "memory" of past actions to handle complex, multi-step reasoning. While normal AI is limited to the knowledge it was trained on, Agents are semi-autonomous and capable of using external tools—like searching the web, sending emails, or running code—to actively complete a task from start to finish.