Guardrails
US
・UK
A1 初級
n. pl.名詞 (複數)護欄
影片字幕
「沒事多喝水」是真的！原來讓自己不再嘴饞吃不停這麼簡單？！(How I Tricked My Brain Into Stopping Cravings)
08:25
- So I took these three points and flipped them into Clarity Guardrails,
因此，我把這三個點轉化為清晰的防護欄，
- into Clarity Guardrails, which over time became stronger
隨著時間的推移，它變得更加強大
OpenAI 代理生成器...Zapier 殺手？ (OpenAI Agent Builder... Zapier Killer?)
10:37
- This time I fancy adding some guardrails.
這一次我想加入一些 guardrails。
- Guardrails exist for building agents, such as no personally identifiable information, moderation, obviously of sensitive subjects, jailbreaking, hallucinations, and
建立代理人有防護欄，像是沒有個人身份資訊、審核、明顯的敏感話題、越獄、幻覺，
為什麼亞洲 1 萬億美元穩定幣的轉移正在重塑全球金融？ (Why Asia’s $1 Trillion Stablecoin Shift Is Reshaping Global Finance)
10:40
- Scaling stablecoins beyond niche use will hinge on trust and and clear guardrails.
穩定幣要超越小眾使用，將取決於信任和明確的監管框架。
親愛的Facebook，你是如何打破民主的|Yael Eisenstat (Dear Facebook, this is how you're breaking democracy | Yael Eisenstat)
13:31
- and they could build in guardrails to stop certain content from going viral
他們可以建立護欄，以阻止某些內容的病毒式傳播。
10分鐘讓你的一天完美開始- 早晨的動機｜成功的動機視頻。 (10 Minutes to Start Your Day Perfect! - MORNING MOTIVATION | Motivational Video for Success)
10:31
- But really it's benefiting you, you know, And so like like if you need the guardrails, then add consequences.
但實際上，它是有利於你的，你知道， 所以像如果你需要的護欄，然後添加的後果。
"我們在這裡有點被包圍了"--眾議員金辛格關於總統對國會大廈的攻擊。 ("We're Kind Of Under Siege Here" - Rep. Kinzinger On The President's Assault On The Capitol Building)
07:23
- THE GOOD THING ABOUT THE UNITED STATES IS THE GUARDRAILS, THE
美國的好處是有警戒線，有警戒線，有警戒線。
東京 - 在哪裡可以買到所有東西 | 終極指南 (Tokyo - Where to Buy EVERYTHING | The Ultimate Guide)
14:06
- I also just noticed that the guardrails here are sports themed.
我還注意到，這裡的護欄是以體育為主題的。
- Guardrails themselves are also different in almost every area of Tokyo but that feels like a different video.
東京幾乎每個地區的護欄本身也不盡相同，但這感覺就像不同的視頻。
人工智能與事實的關係很複雜。這個問題能解決嗎？ (AI has a tricky relationship with facts. Can that be fixed?)
41:27
- What we need to do is, at least at this stage, until we, you know, reach AGI, whenever that that comes, or similar levels of high AI intelligence, is to think about guardrails, to think about being very transparent about where data sources are coming from, being very clear about what content we let in and what content we don't.
我們需要做的是，至少在這個階段，直到我們達到通用人工智慧（AGI）或類似高階AI智慧的水準，我們要思考設立安全護欄，要非常透明地說明資料來源，並清楚我們允許哪些內容進入，哪些不允許。
- Brooke, we keep talking about trust in AI and guardrails, which is a term that's thrown about a lot.
Brooke，我們一直在談論AI的信任和護欄，這是一個常被提及的詞。
對無法控制的人工智能的擔憂與日俱增--兩位頂級人工智能科學家解釋了原因 (Fears around uncontrollable AI are growing — two top AI scientists explain why)
30:34
- In other words, to put guardrails, make sure that agents are not behaving in ways that we would consider immoral, for example.
換句話說，就是要設置警戒線，確保代理人不會做出我們認為不道德的行為。