Knowledge Basics Tech

Quick answer: Build a prompt from five parts: goal, context, constraints, output format, and quality check . Add a short example when the format is unusual, then review the result and refine one weakness at a time. A prompt does not need magic words. It needs enough relevant information for the model to understand the task, plus boundaries that make the result useful. OpenAI’s official guidance highlights instructions, examples, and context as core prompt components. The G-C-C-F-Q framework Part Question to answer Example Goal What must be accomplished? “Create a beginner checklist for backing up a phone.” Context Who, what, and why? “The reader uses Android and has never made a backup.” Constraints What limits or rules apply? “Use plain English; warn before destructive steps.” Format How should the answer be organized? “Give a table, numbered steps, and three FAQs.” Quality check How should it verify its work? “Flag facts that require current officia...

Quick answer: An AI hallucination is a confident but false or erroneous output. NIST uses the term confabulation . The answer may invent facts, citations, quotations, events, or details, so confidence and polished writing should never replace verification. “Hallucination” does not mean the model experiences anything. It is an informal label for output that appears plausible but is not grounded in correct information. What confabulation looks like A book, court case, research paper, or URL that does not exist. A real source paired with a claim it never made. A quotation assembled from the general idea rather than the original words. A product feature from the wrong version or region. A made-up detail added to complete an incomplete story. Contradictory answers given in the same conversation. Why it happens Generative language models produce likely sequences based on patterns and context; they are not automatic truth databases. When information is missing, ambigu...

Quick answer: Break the AI answer into checkable claims, find the original authoritative sources, confirm that each source exists and says what the answer claims, compare dates and context, and use a qualified human for high-stakes decisions. An AI answer can be clear, detailed, and wrong at the same time. NIST recommends reviewing and verifying sources and citations in generative-AI output. Verification is part of using the tool—not a punishment for a poorly written answer. Before entering source material, follow our AI privacy checklist . For a new task, the AI tool selection guide helps you choose what capabilities to test. 1. Identify claims that matter Underline names, numbers, dates, quotations, legal or medical rules, product specifications, software steps, prices, and “current” facts. Separate factual claims from suggestions or style. Verify the claims that could change your decision, money, health, security, or reputation first. 2. Ask for sources—but do not trust the li...

Quick answer: Start with the task, not the brand. Decide what result you need, check whether the information is safe to share, test two suitable tools with the same small example, and compare accuracy, clarity, control, cost and privacy before using one for important work. There is no single “best AI” for every job. A tool that is useful for drafting text may be weak at current research. An image tool may create attractive concepts but should not be trusted for exact diagrams. Choosing well begins with understanding the job and the risk of getting it wrong. Step 1: Define the task in one sentence Write the desired outcome before opening an AI service. Examples include “summarize my own meeting notes,” “suggest a weekly meal plan,” “explain a difficult paragraph,” or “create three visual concepts for a poster.” If the sentence contains several different outcomes, split the work. Research, writing, calculation, image creation and automation need different strengths and different ch...

Quick answer: Assume anything you type or upload to an AI service leaves your immediate device. Remove names and identifiers, replace real records with fictional examples, share only the minimum needed, review the provider’s current data controls, and follow workplace or school rules. AI tools can summarize, rewrite, classify, and brainstorm, but a convenient prompt box is still a data-entry form. NIST identifies data privacy as a generative-AI risk, including leakage, unauthorized disclosure, or inference of personal and sensitive information. Privacy is one part of choosing a suitable AI tool . After protecting your inputs, use our fact-checking steps to verify its output before relying on it. Information to keep out of ordinary prompts Passwords, one-time codes, recovery codes, API keys, and security answers. Full identity numbers, passport details, payment-card numbers, and bank information. Medical records or identifiable health details. Private addresses, precise...