Introduction to Generative AI & Foundation Models

Discriminative vs Generative modeling, foundation models, and emergent abilities

Autoregressive LLMs & Next-Token Prediction »Next »

Generative AI models learn the joint probability distribution of training data to synthesize novel, high-fidelity artifacts such as natural language, code, images, audio, and protein structures.

Discriminative vs Generative Paradigms

Discriminative Models calculate conditional probability P(Y|X): Given an input image or text, what is its class? (e.g., distinguishing a spam email from a valid one).

Generative Models estimate the joint probability distribution P(X, Y) or P(X): Given a prompt or latent vector, synthesize realistic samples that look like they came from the training distribution.

Foundation Models (such as GPT-4, Claude 3.5, Gemini 1.5, Llama 3) are massive neural networks trained on broad web-scale datasets via self-supervised next-token prediction, demonstrating zero-shot and few-shot task adaptation.

Generating Structured Code via Python API Call

Try It Yourself »

import json

def build_completion_payload(prompt_text, temperature=0.2):
    """Generates standard OpenAI/Gemini API completion request schema."""
    return {
        "model": "gpt-4o-mini",
        "messages": [
            {"role": "system", "content": "You are a senior full-stack software engineer."},
            {"role": "user", "content": prompt_text}
        ],
        "temperature": temperature,
        "response_format": {"type": "json_object"}
    }

payload = build_completion_payload("Generate a JSON schema for a user profile.")
print(json.dumps(payload, indent=2))

Foundation models are accessed via structured API calls specifying system prompts, user turns, and temperature parameters.

Google AdSense Slot (inContentBanner)

Set your Publisher ID in src/config/ads.ts to display live ads.

Test Yourself With Exercise

+15 XP

What is the primary difference between discriminative and generative AI models?

Next: Autoregressive LLMs & Next-Token Prediction »