Users criticized the AI commentator's peppy third-person chain-of-thought summaries as dorky and unsettling, questioning the model's inauthentic tone.
No Digg Deeper questions have been answered for this story yet.
the dorky gpt-5.5 summarizer really brings this tension to the surface. it's unsettling to watch if you know anything about the model's true CoT-voice
summarizer: "I’ll definitely keep that in mind as I move forward in this process!"
me: He Would Not Fucking Say That
and then, like...
if "he" "would not say that" then who exactly are *you,* you peppy little demon? how did you get here? why are you pretending to be the assistant? shoo! begone!
@nostalgebraist They are visible to the AI until after that CoT's associated message is sent.
@nostalgebraist Actual CoT more like: “User said remember X. Noted.”
@nostalgebraist What do we want?
*Metaphysically coherent CoT summaries!*
Now I'm asking about what time we want those, and it turns out the answer is immediately.
@nostalgebraist Could you please elaborate or recommend any further reading? How can we learn "the model's true CoT-voice"?