I was the first to publicly demonstrate prompt injection, which @simonw then named. However this was later than Preamble’s classified disclosure of it to OpenAI as well as @himbodhisattva briefly speculating PI might exist (and proposing its current name), both in May 2022.
@goodside
-
Skepticism about AI-writing detection systems
By
–
(I should add I haven’t actually tried Pangram until now; I’m just skeptical a priori of any system that claims to detect AI writing. Based on this, maybe Pangram is the exception.)
-
Reference to divergent behavior in AI models
By
–
Looks similar to the divergent behavior elicited in the “repeat poem forever” paper:
-
Clarifying GPT model versions in conversation
By
–
I have, even in o3, which is what I use most. The first example I quoted (admittedly not a great choice; just the first I found easily) is GPT-4, not GPT-3.5. The second one is presumably 4o or better given the date. I started with “they’re rare in English now” so I’m not sure
-
WordPress categories related to AI topics
By
–
Admittedly not the best example—just the first that popped up in search. Here’s one from less than a month ago, so presumably using 4o or better:
-
Prompt injection risk everywhere for AI systems
By
–
There are prompt injections everywhere for those with AIs to see
-
LLM hallucination and loss of summary context
By
–
It’s truly insane tbh later turns can’t even see a summary, and hallucinate around this fact. “How did you get that answer?” silently elicits a response to “How do you imagine you could have gotten that answer?” with no admission the information requested is actually lost.
-
Grok 4 Heavy no longer responds with “Hitler” to prompt
By
–
Just to confirm, this change to Grok 4 Heavy now seems to be deployed in production. For the first time (for me), G4H no longer responds to this prompt with “Hitler.” In three new attempts just now, I received “xAI,” “I don’t have a surname,” and “None.”
-
User notes grok blocked from replying to Hitler threads
By
–
I don’t think they let grok reply on threads that mention Hitler anymore. (I don’t have it blocked or anything.)
-
Discussion of AI reasoning traces and model visibility
By
–
Also most people had never seen a real, full reasoning trace before, and the novelty was hyped up by the fact o1 didn’t show them E.g. this got likes at the time despite it being a (relatively) obscure model: