Artwork

Contenuto fornito da LessWrong. Tutti i contenuti dei podcast, inclusi episodi, grafica e descrizioni dei podcast, vengono caricati e forniti direttamente da LessWrong o dal partner della piattaforma podcast. Se ritieni che qualcuno stia utilizzando la tua opera protetta da copyright senza la tua autorizzazione, puoi seguire la procedura descritta qui https://it.player.fm/legal.
Player FM - App Podcast
Vai offline con l'app Player FM !

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

11:40
 
Condividi
 

Manage episode 456811163 series 3364758
Contenuto fornito da LessWrong. Tutti i contenuti dei podcast, inclusi episodi, grafica e descrizioni dei podcast, vengono caricati e forniti direttamente da LessWrong o dal partner della piattaforma podcast. Se ritieni che qualcuno stia utilizzando la tua opera protetta da copyright senza la tua autorizzazione, puoi seguire la procedura descritta qui https://it.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

478 episodi

Artwork
iconCondividi
 
Manage episode 456811163 series 3364758
Contenuto fornito da LessWrong. Tutti i contenuti dei podcast, inclusi episodi, grafica e descrizioni dei podcast, vengono caricati e forniti direttamente da LessWrong o dal partner della piattaforma podcast. Se ritieni che qualcuno stia utilizzando la tua opera protetta da copyright senza la tua autorizzazione, puoi seguire la procedura descritta qui https://it.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

478 episodi

Tutti gli episodi

×
 
Loading …

Benvenuto su Player FM!

Player FM ricerca sul web podcast di alta qualità che tu possa goderti adesso. È la migliore app di podcast e funziona su Android, iPhone e web. Registrati per sincronizzare le iscrizioni su tutti i tuoi dispositivi.

 

Guida rapida

Ascolta questo spettacolo mentre esplori
Riproduci