The AI confession neither has internal monolog or access to the thinking tokens.
LLMs are incapable of introspection, they can’t playback their attention weights, or review them, or recall what they thought.
Even thinking tokens are just a reinforcement learning based loop to anneal the models thinking back to a solution. And again, Claude hides the thinking tokens so they don’t get used for model distillation.
This article, and all the articles like it, are pandering bullshit written by morons hoping to fool morons. It is a fiction written by the model to contrast <bad thing happened> to the chat log and system prompt.
The AI confession neither has internal monolog or access to the thinking tokens.
LLMs are incapable of introspection, they can’t playback their attention weights, or review them, or recall what they thought.
Even thinking tokens are just a reinforcement learning based loop to anneal the models thinking back to a solution. And again, Claude hides the thinking tokens so they don’t get used for model distillation.
This article, and all the articles like it, are pandering bullshit written by morons hoping to fool morons. It is a fiction written by the model to contrast <bad thing happened> to the chat log and system prompt.
Good day.