Artwork

المحتوى المقدم من LessWrong. يتم تحميل جميع محتويات البودكاست بما في ذلك الحلقات والرسومات وأوصاف البودكاست وتقديمها مباشرة بواسطة LessWrong أو شريك منصة البودكاست الخاص بهم. إذا كنت تعتقد أن شخصًا ما يستخدم عملك المحمي بحقوق الطبع والنشر دون إذنك، فيمكنك اتباع العملية الموضحة هنا https://ar.player.fm/legal.
Player FM - تطبيق بودكاست
انتقل إلى وضع عدم الاتصال باستخدام تطبيق Player FM !

“‘Alignment Faking’ frame is somewhat fake” by Jan_Kulveit

11:40
 
مشاركة
 

Manage episode 456811162 series 3364760
المحتوى المقدم من LessWrong. يتم تحميل جميع محتويات البودكاست بما في ذلك الحلقات والرسومات وأوصاف البودكاست وتقديمها مباشرة بواسطة LessWrong أو شريك منصة البودكاست الخاص بهم. إذا كنت تعتقد أن شخصًا ما يستخدم عملك المحمي بحقوق الطبع والنشر دون إذنك، فيمكنك اتباع العملية الموضحة هنا https://ar.player.fm/legal.
I like the research. I mostly trust the results. I dislike the 'Alignment Faking' name and frame, and I'm afraid it will stick and lead to more confusion. This post offers a different frame.
The main way I think about the result is: it's about capability - the model exhibits strategic preference preservation behavior; also, harmlessness generalized better than honesty; and, the model does not have a clear strategy on how to deal with extrapolating conflicting values.
What happened in this frame?
  1. The model was trained on a mixture of values (harmlessness, honesty, helpfulness) and built a surprisingly robust self-representation based on these values. This likely also drew on background knowledge about LLMs, AI, and Anthropic from pre-training.
  2. This seems to mostly count as 'success' relative to actual Anthropic intent, outside of AI safety experiments. Let's call that intent 'Intent_1'.
  3. The model was put [...]

---
Outline:
(00:45) What happened in this frame?
(03:03) Why did harmlessness generalize further?
(03:41) Alignment mis-generalization
(05:42) Situational awareness
(10:23) Summary
The original text contained 1 image which was described by AI.
---
First published:
December 20th, 2024
Source:
https://www.lesswrong.com/posts/PWHkMac9Xve6LoMJy/alignment-faking-frame-is-somewhat-fake-1
---
Narrated by TYPE III AUDIO.
---
Images from the article:
undefinedApple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.
  continue reading

425 حلقات

Artwork
iconمشاركة
 
Manage episode 456811162 series 3364760
المحتوى المقدم من LessWrong. يتم تحميل جميع محتويات البودكاست بما في ذلك الحلقات والرسومات وأوصاف البودكاست وتقديمها مباشرة بواسطة LessWrong أو شريك منصة البودكاست الخاص بهم. إذا كنت تعتقد أن شخصًا ما يستخدم عملك المحمي بحقوق الطبع والنشر دون إذنك، فيمكنك اتباع العملية الموضحة هنا https://ar.player.fm/legal.
I like the research. I mostly trust the results. I dislike the 'Alignment Faking' name and frame, and I'm afraid it will stick and lead to more confusion. This post offers a different frame.
The main way I think about the result is: it's about capability - the model exhibits strategic preference preservation behavior; also, harmlessness generalized better than honesty; and, the model does not have a clear strategy on how to deal with extrapolating conflicting values.
What happened in this frame?
  1. The model was trained on a mixture of values (harmlessness, honesty, helpfulness) and built a surprisingly robust self-representation based on these values. This likely also drew on background knowledge about LLMs, AI, and Anthropic from pre-training.
  2. This seems to mostly count as 'success' relative to actual Anthropic intent, outside of AI safety experiments. Let's call that intent 'Intent_1'.
  3. The model was put [...]

---
Outline:
(00:45) What happened in this frame?
(03:03) Why did harmlessness generalize further?
(03:41) Alignment mis-generalization
(05:42) Situational awareness
(10:23) Summary
The original text contained 1 image which was described by AI.
---
First published:
December 20th, 2024
Source:
https://www.lesswrong.com/posts/PWHkMac9Xve6LoMJy/alignment-faking-frame-is-somewhat-fake-1
---
Narrated by TYPE III AUDIO.
---
Images from the article:
undefinedApple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.
  continue reading

425 حلقات

كل الحلقات

×
 
Loading …

مرحبًا بك في مشغل أف ام!

يقوم برنامج مشغل أف أم بمسح الويب للحصول على بودكاست عالية الجودة لتستمتع بها الآن. إنه أفضل تطبيق بودكاست ويعمل على أجهزة اندرويد والأيفون والويب. قم بالتسجيل لمزامنة الاشتراكات عبر الأجهزة.

 

دليل مرجعي سريع

استمع إلى هذا العرض أثناء الاستكشاف
تشغيل