AI Literacy list page — category filters and plain-language AI guides.
🧠 AI Literacy
AI is changing how we invest and work. Plain-language guides — no jargon required.
Where AI Bias Actually Comes From
AI bias isn't one thing. It seeps in differently at each of the three training stages, and both adding more data and correcting for it tend to surface new problems of their own.
Read →Multimodal AI — How It Understands Images, Audio, and Video
Images and audio get converted into coordinates the same way text does, then fed into the same model. That explains both why it's so expensive and why it's weak at precise tasks like counting.
RAG, Fully Explained — How an Internal Document Chatbot Is Built
RAG doesn't retrain a model — it retrieves relevant documents per question and hands them over. How it works via embeddings and a vector DB, why hallucination drops but doesn't vanish, how it differs from fine-tuning, and its security weak point.
Building Your Own Prompt Templates — Cleaning Up Recurring Work
Writing a fresh prompt every time makes context and format inconsistent. How to separate the fixed frame from the fill-in slots, three conditions for a stable template, and three ready-to-use examples.
Who Owns What AI Writes — AI and Copyright
That the author must be human is settled in both Korea and the US. AI-only output has no copyright; whether training data is infringement is still being litigated. What is settled and what is not.
How Reasoning Models Are Different — How "Thinking" AI Works
A reasoning model writes out its working as tokens before answering. Those tokens feed back in and help on hard problems, but it is slow and costly. When to use it and when to stick with a regular model.
AI and Your Privacy — Where Does What You Type Actually Go
Whether it trains the model and whether a human reads it are different questions. The riskiest input isn't your own data — it's someone else's. What to check before pasting in company material.
On-Device AI — What's Different About AI That Runs on Your Own Device
AI that finishes entirely on your device, no server round trip. How quantization compresses the model and NPUs save battery — and what actually works today.
Writing With AI — Prompts for Drafting, Editing, and Summarizing
Using one prompt for drafting, editing, and summarizing is the shared cause of failure — the three need opposite latitude. Supply material, lock the scope, set the discard rule.
Tokens and the Context Window — Pricing, Limits, and Long Documents
A context window isn't memory — it's a bowl refilled every turn. Why Korean costs more, why doubling length quadruples compute, and how to read a "1M token" headline.
Prompt Injection and AI Security — Before You Hand Work to an Agent
An attack that manipulates AI through instructions hidden in external content. Why the model can't tell commands from data, direct vs. indirect injection, and a checklist before handing work to an agent.
Getting AI to Read Your Documents — PDFs, Spreadsheets, and Long Reports
Long documents lose their middle, and tables break apart during extraction. What actually happens with scanned PDFs, table-heavy files, and financial statements — plus a checklist by document type.
Why AI Gets Numbers Wrong — Calculations, Statistics, and Financial Data
AI doesn't calculate numbers — it predicts them. Why error rates climb with digit count, what code execution does and doesn't fix, and the patterns to watch for in financial data.
What Is an AI Agent? — How It Differs From a Chatbot, and Why It's Suddenly Everywhere
The line between a chatbot and an agent is loop structure, not capability. Broken into three parts — function calling, planning, the loop — and why agents drive up inference demand.
How an AI Model Gets Built — Pre-training, Fine-tuning, RLHF
Which stage costs the billions, and which one sets the tone? Following the three stages in order shows why GPU clusters exist.
Researching With AI — A Search, Summarize, Verify Workflow
"Summarize this" gets you something plausible but unverified. A four-stage loop — gather, summarize, cross-check, source-verify — with the prompts for each.
ChatGPT vs Claude vs Gemini — A 2026 Hands-On Comparison
Sorted by real use, not benchmark scores: long documents, code, Korean writing quality, recency, and price — five axes compared head to head.
50 AI Terms — Tokens, Parameters, RAG, Agents, All in One Place
The 50 AI terms flooding news and earnings calls, sorted into five clusters — each with a one-line definition and where you'll actually run into it.
Prompt Engineering Basics — 7 Principles That Double Answer Quality
Asking well is a set of rules, not a knack. Role, context, format, examples, constraints, verification, iteration — each with a bad-vs-good side-by-side.
Why AI Hallucinates — And 5 Ways to Catch It
Hallucination isn't a bug — it's a byproduct of probabilistic prediction. The three weak spots (numbers, citations, recency) and five verification steps you can apply today.
How Does ChatGPT Actually Answer? — LLMs Explained Without Metaphors
Tokenization → embeddings → attention → next-token probability. Why GPUs are needed, and why wrong answers sound so confident.