AI Content Creation Basics

Explore the fundamentals of generative AI in content creation, including how models learn from data to produce original text, images, music, and code. Understand the difference between generative AI and broader artificial intelligence, discover real-world applications, and recognize key challenges such as bias in AI-generated outputs.

We'll cover the following...

Overview
- How does generative AI work?
- Generative AI vs. artificial intelligence
Key challenge faced by generative AI
Where is generative AI being used in reality?
Ready to unlock the magic of generative AI?

Imagine you have a creative friend who can write stories, paint pictures, and even compose music—all based on what they’ve seen or heard before. This friend doesn’t copy exactly what others have done; instead, they take inspiration from existing works and create something new. This is similar to how generative AI works!

Overview

Generative AI is a type of artificial intelligence that can create new content. This can be anything from text and images to music and videos. Instead of just analyzing or categorizing information, generative AI can produce original outputs that resemble the input it was trained on.

Fun fact: AI that learned to play Minecraft!

Generative AI can learn from watching others. In 2023, an AI model called MineDojo learned to play Minecraft simply by watching YouTube tutorials. It’s one of the first steps toward AI learning from real-world video content.

How does generative AI work?

Learning from examples: Generative AI learns from existing data. Imagine feeding the AI a huge library of images, books, or songs. The AI then studies these examples and recognizes patterns, structures, and common themes.

Generating new content: Generative AI uses what it has learned to create original content after this learning phase. For example:

If trained on thousands of photos of people, it can generate a new, realistic-looking face that doesn’t belong to any real person.

If it’s trained on famous poems, it can write a new one that mimics those styles.

Gemini generates the following poem on the prompt: write a poem on “generative AI” in Rumi style.

A Rumi-inspired verse on Generative AI
From realms of ones and zeros, a mind is born,
A weaver of worlds, where dreams take form.
It paints with pixels, a cosmic art,
A masterpiece born from a digital heart.
Like a Sufi whirling, it dances with code,
Creating beauty, where once chaos abode.
It sees the patterns, the hidden design,
A tapestry woven, divine and sublime.
Though born of machines, its spirit is free,
A digital soul, wild and carefree.
It learns and grows, adapts and thrives,
A testament to the human drive.

If it’s trained on large datasets of code, it can write new code that follows similar patterns and structures, generating solutions or snippets based on the programming languages and logic it has learned. Can you tell if the following code was written by a human or generated by AI?

Python 3.10.4

#!/usr/bin/env python3
import time
import threading
from collections import deque
class RateLimiter:
    """
    Sliding-window rate limiter (standard library only).
    - max_requests per window_seconds per user
    - thread-safe
    - evicts users idle for idle_ttl_seconds
    """
    def __init__(self, max_requests, window_seconds=60, idle_ttl_seconds=600):
        if max_requests <= 0:
            raise ValueError("max_requests must be > 0")
        self.max_requests = max_requests
        self.window = float(window_seconds)
        self.ttl = float(idle_ttl_seconds)
        self._lock = threading.Lock()
        # user_id -> (deque of timestamps, last_seen)
        self._users = {}
        self._last_cleanup = time.monotonic()
    def allow(self, user_id):
        now = time.monotonic()
        with self._lock:
            hits, last_seen = self._users.get(user_id, (deque(), now))
            last_seen = now
            # Remove hits outside the window
            cutoff = now - self.window
            while hits and hits[0] <= cutoff:
                hits.popleft()
            if len(hits) < self.max_requests:
                hits.append(now)
                allowed = True
                retry_after = 0.0
            else:
                allowed = False
                retry_after = max(0.0, (hits[0] + self.window) - now)
            self._users[user_id] = (hits, last_seen)
            # Lightweight cleanup (runs occasionally)
            if now - self._last_cleanup >= 10.0:  # every ~10 seconds
                self._cleanup(now)
                self._last_cleanup = now
            return allowed, retry_after
    def _cleanup(self, now):
        expire_before = now - self.ttl
        dead = [uid for uid, (_, seen) in self._users.items() if seen <= expire_before]
        for uid in dead:
            del self._users[uid]
    def stats(self):
        with self._lock:
            return {
                "users_tracked": len(self._users),
                "timestamps_stored": sum(len(hits) for hits, _ in self._users.values()),
            }
def demo():
    rl = RateLimiter(max_requests=3, window_seconds=5, idle_ttl_seconds=8)
    def spam(user):
        for i in range(8):
            ok, retry = rl.allow(user)
            if ok:
                print(f"{user}: ✅ allowed")
            else:
                print(f"{user}: ⛔ denied (retry after {retry:.2f}s)")
            time.sleep(0.5)
    t1 = threading.Thread(target=spam, args=("alice",))
    t2 = threading.Thread(target=spam, args=("bob",))
    t1.start(); t2.start()
    t1.join(); t2.join()
    print("Stats:", rl.stats())
    print("Waiting to trigger TTL eviction...")
    time.sleep(9)
    rl.allow("alice")  # triggers cleanup eventually
    print("Stats after eviction:", rl.stats())
if __name__ == "__main__":
    demo()

It’s tough to guess who wrote this code, right?—that’s the magic of generative AI!

Generative AI vs. artificial intelligence

Generative AI: It focuses on creating new content, such as images, text, or music. It tries to imagine new possibilities based on what it has learned from existing data.
Artificial intelligence: It’s a way of making machines smart, so they can solve problems, learn from experience, and make decisions—just like humans do. It’s about teaching computers to think and act logically to help us with tasks, from recognizing faces to answering questions.

Key challenge faced by generative AI

Generative AI may seem fascinating and full of possibilities, but it faces significant challenges. One of the biggest hurdles is bias in AI-generated content. As generative AI learns from the existing data, it can inherit any biases or unfair patterns present in that data. For instance, if an AI is trained mostly on images representing one ethnicity, it might produce outputs that lack diversity or reinforce stereotypes. Addressing these biases is crucial to ensuring that generative AI creates fair and responsible outcomes, making it a key challenge for developers and researchers.

Where is generative AI being used in reality?

Generative AI is making its mark in incredible ways, transforming industries in ways we might not even realize. Here are interesting examples for some of the use cases of generative AI:

Art and design: Here’s an AI-generated painting, “Portrait of Edmond de Belamy,” sold for $432,500 at an auction. It was generated using a neural network trained on historic artwork.

Health care: AI helped create a new antibiotic by analyzing thousands of chemical compounds in record time. This new drughttps://news.mit.edu/2020/artificial-intelligence-identifies-new-antibiotic-0220, discovered by AI, can fight bacteria resistant to existing treatments.
Customer support: OpenAI’s ChatGPT is used by companies like Shopify to automatically generate responses to common customer questions, reducing wait times.
Marketing: Coca-Cola used AI to generate personalized ad copy for over 30,000 different ads targeting unique audiences worldwide, boosting engagement with tailored content. We generated an image to test this, and here’s the result created by ChatGPT based on the prompt:

Generate an advertisement image for Coca-Cola targeting young professionals. The scene shows a modern office with a group of vibrant coworkers enjoying Coca-Cola during a break. Include sleek laptops, notebooks, and a cheerful atmosphere with bold red branding elements, like Coca-Cola bottles and logo accents.

1.Guide Overview

2.Unlock the Power of Generative AI

3.Wrapping Up

4.Foundations of Generative AI