Question. Suppose we call the kind of unrepetant boot-licking (“I apologize for the error”) that LLMs often do “chatlighting” (h/t @Dolimac
). Why exactly is chatlighting so frequent? Is it a consequence of RLHF? Do we see the same chatlighting behavior in “base models” without
ETHICS
-
Chatlighting in LLMs: RLHF’s Role in Obsequious Responses
By
–
-
ChatGPT Voice Hallucinations from Background Noise Fragments
By
–
It's quite fun leaving ChatGPT Voice running and later seeing how it hallucinated increasingly weird fragments of conversation based on snippets of background noise
-
Default Settings Change: Yes to No Paradigm Shift
By
–
I think changing default to no instead of yes is the biggest change one can do.
-
Claude generates false information based on conversational context
By
–
Claude doesn’t know that it lies it just says stuff based on conversational context, like when MetaAI claimed to have a child.
-
AI Preferences: Clarifying the Core Discussion
By
–
I think we need to back up and try to say better what this conversation is about. It's got nothing to do with human preferences. It's about AIs ending up with preferences.
-
Utility as Optimized Property in Outer Optimization Processes
By
–
We're not using "utility" as something that's supposed to descriptively surprisingly true of human preference; we're using it to describe properties that we expect to get optimized into a thinking process by any outer optimization that rewards good outer solutions.
-
Gary Marcus Five Critical Points for Elon Musk
By
–
sketched here: https://
garymarcus.substack.com/p/dear-elon-mu
sk-here-are-five-things
… -
Evolution’s Optimization Process and Reality’s Infinite Degrees of Freedom
By
–
Evolution's outer optimization process was aimed at "inclusive genetic fitness", a simple concept that humans never even formed except by postmortem reconstruction millions of years later. There's vastly more degrees of freedom in "where to steer reality" than "predicting
-
Humans Make Mistakes Too: LLMs vs Human Error Discussion
By
–
“but humans make mistakes, too” is so common a reply, I have written a discussion comparing humans and LLMs:
-
Groupthink versus Tulips: AI Market Rationality Questioned
By
–
https://
garymarcus.substack.com/p/groupthink-v
ersus-tulips-what-will?r=8tdk6
…