引導式測驗 · 中等 等級

英國競爭暨市場管理局 (CMA) AI 基礎模型審查測驗

了解 CMA 如何評估基礎模型市場、競爭和消費者風險,以及其原則與具有約束力的法律調查結果有何不同。

問題 1 的 8

CMA 最初的基礎模型審查的目的是什麼?

預覽每個問題8 問題

當您參加測驗時,就會出現答案和解釋。

  1. CMA 最初的基礎模型審查的目的是什麼?
    • Create a new UK AI criminal offence
    • Approve every model for public use
    • Set fixed prices for all AI services
    • Build an early understanding of markets, risks and useful principles
  2. Which set names the six principles confirmed in the 2024 update?
    • Safety, truthfulness, privacy, consent, copyright and security
    • Access, diversity, choice, fair dealing, transparency and accountability
    • Compute, training, inference, prompts, agents and chips
    • Licensing, taxation, registration, audit, identity and redress
  3. 訪問原則解決了什麼問題?
    • A guarantee of free model use
    • Access to an individual’s private account
    • Ready access to important inputs such as data, compute, expertise and funding
    • The right to avoid all regulations
  4. 為什麼 CMA 審查不應該被描述為針對每家公司的調查結果?
    • A market review identifies issues and risks but does not prove each firm broke the law
    • The CMA does not consider markets
    • All partnerships are illegal
    • The report only studies software quality
  5. 哪個問題與選擇和彈性最直接相關?
    • Does the model have a colorful interface?
    • Can a business switch providers or use more than one service?
    • Can a model write in several tones?
    • Does the system use a transformer?
  6. 透明原則強調什麼?
    • Making every model open source
    • Publishing customer data
    • Guaranteeing every answer is correct
    • Giving businesses and consumers information about risks and limitations
  7. How should a reader describe the review’s status?
    • The initial review closed, with CMA continuing wider monitoring of AI markets
    • The CMA banned foundation models
    • The review remains a statute
    • The CMA approved one exclusive model
  8. How should the six principles be used?
    • As a replacement for competition law
    • As a guarantee of market outcomes
    • As lenses for analysis, not automatic proof of an infringement
    • As technical standards for model training