负责任的人工智能用户

• Design clear prompts with useful context and constraints.
• Review outputs for factual, privacy, bias, and copyright risks.
• Build a repeatable human-in-the-loop workflow.

先推荐： 人工智能基金会

1. 1

   Prompting with context

   结果： Turn vague requests into testable instructions.

   练习： Rewrite a weak prompt with audience, context, constraints, examples, and success criteria.

   Prompt and context designLanguage-model mechanics
   Prompt Engineering →ChatGPT 与大语言模型 →
2. 2

   Reviewing outputs

   结果： Detect unsupported claims and decide what requires verification.

   练习： Annotate an AI answer as verified, attributed, uncertain, or unsupported.

   Output review and error detectionBias and fairnessSource and claim verification
   人工智能幻觉 →人工智能偏见 →
3. 3

   Privacy and ownership

   结果： Choose safe inputs and understand content rights.

   练习： Create a red/yellow/green data policy for an AI workflow.

   隐私和数据处理Copyright and content provenance
   人工智能与隐私 →人工智能与版权 →
4. 4

   A dependable workflow

   结果： Combine AI assistance with checkpoints and human ownership.

   练习： Map an existing task and mark where AI may draft, check, or never decide.

   AI辅助工作流程设计工具选择与比较Automation boundaries and safeguards
   人工智能在日常生活中的应用 →人工智能写作工具 →