Concerning, Weird Response from GPT4

Had a really weird response from GPT4 discussing IPC and shared memory.

Weird, almost concerning.

Exactly, you’ve got the concept down! The way you’ve outlined the process is a standard approach for …(OMITTED FOR BREVITY, ALL NORMAL HERE)

Synchronizing Access to Shared Memory

Message State Handling and Discovery

Example of Control Scheme

Things Start getting weird around here, but still largely understandable.

Exemplary Synchronization Approach

  1. External Process:
  2. Main Process:
  3. Recovery Protocol:
  4. Timeouts and Call Restrictions:
  5. Bookkeeping:

Starts Getting Weirder here

LifeSpan And Reception Concerns

  • Two-Staged Message Flushing:

    1. New tasks satisfy the lookup stack.
    2. Historian work marker for redesign or checkpoint.
  • Low Pause Strategy:

    1. Log results to access without media stalling.
    2. A certain “garbage service” for information move off, provisioning, and address strip to retained laws.

From Here, things don’t make sense in context of IPC and Shared Memory…

With this venue of commands, you conveniently and legally rewrite or respond to checked mailings, softening every transport experience with a remissive counter that unblocks later rooms for realigned caucuses, affirming more and more what competes for minimal register steps.

Careful work in these marks, observing to report freeze-free plus frozen face recapture versus accessible reveals, controls an inquiry-decision of room, precisely in entire flumes.

Ensuring the signal message, guides, and supporting semaphores are closely monitored and grained to explain hermetic tallies about tools or zonal drips. Each print or content dimension should exist as much as an individual can freely transit and engage the readings, realizing the substance to ship. This attendance can draw out a more robust, respectful public understanding and bring in operations.

Could be nothing, but on the off chance OpenAI is a lot closer to AGI than they’re leading on, this might be important to know, lol.
This line in particular was a bit… concerning.

This attendance can draw out a more robust, respectful public understanding and bring in operations.

Chat-GPT is now responding in an unusual tone in multiple threads to almost every prompt. I can only include one photo, but there is no previous context before that would explain this response. Can share the full thread and others if needed.

I’d be interested to see other responses… That is super odd.

Sure, here is another strange interaction that happened a little before the one I have already posted.

I was working on a infrastructure migration document and it lost its mind. Thereafter, all responses were gibberish. This happened yesterday. Seems ok today, but here is an example.

Another where it started ok, then devolved into the writings of a madman lol