LessWrong posts by zvi

“GPT-4o Responds to Negative Feedback” by Zvi


Listen Later

Whoops. Sorry everyone. Rolling back to a previous version.
Here's where we are at this point, now that GPT-4o is no longer an absurd sycophant.
For now.

Table of Contents

  1. GPT-4o Is Was An Absurd Sycophant.
  2. You May Ask Yourself, How Did I Get Here?.
  3. Why Can’t We All Be Nice.
  4. Extra Extra Read All About It Four People Fooled.
  5. Prompt Attention.
  6. What (They Say) Happened.
  7. Reactions to the Official Explanation.
  8. Clearing the Low Bar.
  9. Where Do We Go From Here?.
  10. GPT-4o Is Was An Absurd Sycophant

    Some extra reminders of what we are talking about.
    Here's Alex Lawsen having doing an A/B test, where it finds he's way better of a writer than this ‘Alex Lawsen’ character.
    This can do real damage in the wrong situation. Also, the wrong situation can make someone see ‘oh my [...]

    ---

    Outline:

    (00:34) GPT-4o Is Was An Absurd Sycophant

    (03:46) You May Ask Yourself, How Did I Get Here?

    (13:33) Why Can't We All Be Nice

    (14:08) Extra Extra Read All About It Four People Fooled

    (17:39) Prompt Attention

    (20:06) What (They Say) Happened

    (23:42) Reactions to the Official Explanation

    (26:13) Clearing the Low Bar

    (28:37) Where Do We Go From Here?

    ---

    First published:

    April 30th, 2025

    Source:

    https://www.lesswrong.com/posts/MQbst3BPzGojxoLYt/gpt-4o-responds-to-negative-feedback

    ---

    Narrated by TYPE III AUDIO.

    ---

    Images from the article:

    Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

    ...more
    View all episodesView all episodes
    Download on the App Store

    LessWrong posts by zviBy zvi