Questionário vinculado ao guia · Médio Nível

Questionário sobre assistentes de IA

Entenda como os assistentes combinam modelos e ferramentas, por que a memória difere do treinamento e como verificar evidências, permissões, ações e transferências.

Caminhos de guia relacionadosAssistentes de IA
Pergunta 1 de 8

O que é um assistente de IA na definição do guia?

Visualize cada pergunta8 perguntas

As respostas e explicações aparecem conforme você responde ao teste.

  1. O que é um assistente de IA na definição do guia?
    • Uma interface que ajuda uma pessoa a obter informações, criar material ou concluir tarefas
    • Um modelo com acesso ilimitado a todas as contas
    • Um banco de dados de respostas verificadas
    • Um chatbot que não pode usar ferramentas
  2. If an assistant answers from a supplied document, what is its evidence basis?
    • The document automatically retrains the model
    • The model has verified every claim in the document
    • The provided document is part of its context, but the answer still needs checking
    • The assistant must ignore the supplied text
  3. What does a saved preference in an assistant’s memory imply about training?
    • That every model parameter changed
    • Not necessarily that the underlying model was retrained
    • That the preference is public
    • That the assistant cannot forget anything
  4. What determines whether an assistant can access a connected account?
    • The model’s name alone
    • A prompt asking it to ignore permissions
    • The amount of text in the conversation
    • The product integration and permissions granted
  5. The assistant says it drafted a report but has not saved a file. How should it label the result?
    • As a draft, not a completed file
    • As a saved file because text exists
    • As an approved report
    • As a verified result
  6. What interaction feature supports recoverability?
    • Forcing a restart after every error
    • Hiding whether an action is running
    • Allowing correction, cancellation, or manual workflow
    • Executing consequential actions without review
  7. Why should a product explain whether an answer used supplied sources or tools?
    • It guarantees the answer is correct
    • The answer’s evidence basis affects how it should be checked
    • It determines the user’s account password
    • It removes the need for source links
  8. Which metric better reflects assistant usefulness than answer length alone?
    • Number of sentences
    • Tamanho do modelo
    • Conversation duration
    • Task outcome and user effort