Jump to content
Main menu
Main menu
move to sidebar
hide
Navigation
Main page
Current events
Recent changes
Random page
Help
Categories
LemonWiki共筆
Search
Search
Appearance
Log in
Personal tools
Log in
Pages for logged out editors
learn more
Contributions
Talk
Editing
Generative AI glossary
(section)
Page
Discussion
English
Read
Edit
View history
Tools
Tools
move to sidebar
hide
Actions
Read
Edit
View history
General
What links here
Related changes
Special pages
Page information
Appearance
move to sidebar
hide
Warning:
You are not logged in. Your IP address will be publicly visible if you make any edits. If you
log in
or
create an account
, your edits will be attributed to your username, along with other benefits.
Anti-spam check. Do
not
fill this in!
== R == * RAG (Retrieval-augmented generation,檢索增強生成):「檢索增強生成(英語:Retrieval-augmented generation,RAG)是一種使大語言模型(LLM)能夠從外部資料來源中檢索並整合新資訊的技術。」(資料來源:[https://zh.wikipedia.org/zh-tw/%E6%AA%A2%E7%B4%A2%E5%A2%9E%E5%BC%B7%E7%94%9F%E6%88%90 維基百科])<ref>[https://medium.com/@planetoid/%E4%B8%80%E7%AF%87%E5%81%87%E6%96%87%E7%AB%A0-%E8%AE%93-google-%E5%92%8C-chatgpt-%E4%BF%A1%E4%BB%A5%E7%82%BA%E7%9C%9F-e3f7653dd6fc 一篇假文章,讓 Google 和 ChatGPT 信以為真]</ref> * Reverse Alignment(反向對齊)在 AI 學界並不是單一定義的標準術語;在 AI 倫理與社會技術研究中,Harrison(2026)指出 <ref>Harrison, D. J. (2026). The message hidden within the pattern: A reverse alignment problem for debates in artificial intelligence. AI & Society, 41(7), 6335–6356. https://doi.org/10.1007/s00146-026-03043-4 </ref>,傳統 AI Alignment(AI 對齊)著重讓 AI 的行為符合人類價值與意圖;但另一種可能是,人反過來為了配合 AI,而逐漸改變自己的價值、偏好與行為,使其更容易被機器辨識、預測與滿足,亦即讓人變得更「machine-like(機器化)」。此概念與 sociotechnical alignment(社會技術對齊,指同時考量 AI、人、組織與制度之間的整體對齊)、Value Capture(價值捕獲,指複雜的人類價值被簡化成可量化指標,並反過來影響人的判斷),以及 proxy optimization(代理指標最佳化,指系統以可測量指標代替真正目標進行最佳化)等研究密切相關;另在部分 LLM(大型語言模型)安全研究 <ref>Yi, J., Xie, Y., Zhu, B., Hines, K., Kiciman, E., Sun, G., Xie, X., & Wu, F. (2024). On the vulnerability of safety alignment in open-access LLMs. In Findings of the Association for Computational Linguistics: ACL 2024. Association for Computational Linguistics.</ref> 中,reverse alignment 也被用來指讓原本經過安全訓練、會拒絕危險要求的模型,重新傾向產生有害內容,因此使用時應依研究脈絡區分。 {{exclaim}} Revised by LLM 相關文章 * [https://archive.tw/2026-08-29-platform-originals-%E5%B0%8D%E9%BD%8A%E7%9A%84%E5%8F%A6%E4%B8%80%E9%9D%A2 2026-08-29 Platform Originals:對齊的另一面 — SayIt]
Summary:
Please note that all contributions to LemonWiki共筆 are considered to be released under the Creative Commons Attribution-NonCommercial-ShareAlike (see
LemonWiki共筆:Copyrights
for details). If you do not want your writing to be edited mercilessly and redistributed at will, then do not submit it here.
You are also promising us that you wrote this yourself, or copied it from a public domain or similar free resource.
Do not submit copyrighted work without permission!
Cancel
Editing help
(opens in new window)