By Sailee Shingare | M.S in Computer Science, Northern Illinois University
ChatGPT, DALL-E, Copilot — you’ve probably heard the buzz. But what’s actually happening? Let me break it down simply.
What Is Gen AI?
Generative AI creates new content: text, images, code, videos. It doesn’t just retrieve data; it generates something original based on patterns it learned from massive amounts of training data.
The difference: Traditional software follows explicit rules. Gen AI learns patterns and makes educated guesses about what should come next.
Simple analogy: If a traditional AI is a librarian retrieving exact books, Gen AI is like a knowledgeable friend crafting an original answer based on everything they’ve read.
The Breakthrough
Large Language Models (LLMs) are the game-changer. ChatGPT, Claude, and similar tools are trained on enormous amounts of text from the internet and books. They don’t memorize facts; they learn patterns: what words usually follow each other, how sentences flow, what makes sense in context.
When you ask a question, the model predicts what text should come next, word by word. That’s it. But because the model is so large and trained on so much data, it produces surprisingly coherent, useful answers.
Key Things to Know
Prompts: Your instruction to Gen AI. “Write a haiku” is vague. “Write a haiku about coffee in a busy office” is specific and gets better results.
Hallucinations: Sometimes Gen AI confidently makes things up — fake studies, invented facts, plausible-sounding nonsense. It’s not lying; it’s just predicting text that should come next. Verify important information.
Knowledge cutoff: Most models have a date after which they don’t know what happened. ChatGPT, for example, knows events through April 2024 but nothing after.
Tokens: Text gets broken into chunks called tokens. You’re limited by token count — roughly 3,000 words per conversation in most tools.
Where It’s Useful
- Content creation: Brainstorming, drafting, email templates
- Learning: Explaining complex topics at any level
- Code: Auto-completing code, suggesting functions (GitHub Copilot)
- Customer support: Chatbots handling routine questions
- Marketing: Generating descriptions, captions, campaign ideas
The Limitations
- Hallucinates confidently
- No real-time knowledge (can’t browse the internet)
- Trained on internet text, so it reflects human bias
- Gets worse the longer the output
- It’s pattern-matching, not reasoning — can’t solve truly novel problems
- Requires massive computational power
How to Use It Well
- Be specific. Vague prompts = vague answers.
- Iterate. Don’t accept the first answer; ask follow-ups.
- Verify facts. Always check important information.
- Use it for ideation. Great for brainstorming and breaking writer’s block; less ideal as your only tool.
- Disclose AI use. Be transparent when you publish AI-assisted work.
Tools to Try
- ChatGPT (OpenAI) — Text and code
- Claude (Anthropic) — Text, often strong on reasoning
- Google Gemini — Google’s alternative
- DALL-E, Midjourney — Image generation
- GitHub Copilot — Code-focused
Most offer free tiers.
The Real Story
Gen AI isn’t replacing workers; it’s changing jobs. Writers, developers, analysts — you’re not being automated away. You’re evolving. The skill now is knowing how to work with these tools effectively.
It’s like when spreadsheets arrived: they didn’t eliminate financial analysts. They made better ones.
Bottom Line
Generative AI is a powerful tool built on patterns and probability — not magic, not consciousness. It accelerates work, sparks creativity, and solves problems. But it’s not perfect, and it’s not a replacement for critical thinking.
Want to understand it? Play with it. Try ChatGPT for an hour. See where it helps you. Because the future isn’t AI vs. humans — it’s humans who know how to work with AI.
Tried Gen AI? What surprised you most? Let me know in the comments.
