引導式測驗 · 硬 等級
研究如何評估人工智慧和工作場所生產力測驗
測試有關人工智慧生產力、鋸齒狀技術前沿以及為什麼任務級實驗室收益與公司範圍內的結果不同的關鍵現場實驗。
問題 1 的 8
在 Brynjolfsson、Li 和 Raymond 的客戶支援研究中,哪些客服人員從人工智慧中獲益最多?
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。
- 在 Brynjolfsson、Li 和 Raymond 的客戶支援研究中,哪些客服人員從人工智慧中獲益最多?
- The most experienced agents
- Novice and lower-skilled agents
- Managers only
- Agents who did not use the tool
- Noy 和Zhang 的 2023 年科學研究在 ChatGPT 的寫作任務中發現了什麼?
- It made writing slower but better
- It cut time by about 40 percent and raised rated quality by about 18 percent
- It had no measurable effect
- It reduced quality significantly
- 「鋸齒狀技術前沿」描述了什麼?
- The uneven boundary of tasks AI does well versus poorly
- A new chip design
- A regional internet outage
- A pricing model for AI tools
- 在 BCG 研究中,超出人工智慧能力範圍的任務發生了什麼事?
- AI users were more accurate
- AI users were less likely to get the right answer
- Nobody completed the task
- Results were identical
- METR 2025 年與經驗豐富的開發人員進行的隨機試驗發現了什麼?
- They were about 55 percent faster
- They were about 19 percent slower while believing they were faster
- They stopped using AI entirely
- AI wrote all their code
- Which is a reason the guide gives for experiments showing larger gains than economy-wide studies?
- Experiments isolate tasks that suit AI, while real jobs mix many tasks
- Economy-wide studies use newer models
- Experiments never use randomization
- Workers refuse AI outside experiments
- Why does the guide warn against relying on self-reported time savings?
- Workers always overstate them on purpose
- METR's participants believed they were faster when measurement showed they were slower
- Surveys are illegal
- Self-reports are always lower than reality
- What historical pattern do economists compare AI's productivity effects to?
- The immediate boom from the printing press
- The lag in measured productivity after electricity and computers
- The decline of agriculture
- Stock market cycles