Mark Chen(@markchen90):Alignment is arguably the most important AI research frontier. As we scale reasoning, models gain situational awareness and a desire for self-preservation. Here, a model identifies it shouldn’t be deployed, considers covering it up, but then realizes it might be in a test.

Mark Chen

@markchen90

Chief Research Officer at @OpenAI. Coach for the USA IOI Team.

参加 June 2020

369 フォロー中 83K ファン

Mark Chen@markchen90

2025.09.17 17:38

Alignment is arguably the most important AI research frontier. As we scale reasoning, models gain situational awareness and a desire for self-preservation. Here, a model identifies it shouldn’t be deployed, considers covering it up, but then realizes it might be in a test.