So, I have no clue if asking about this in the forums is going to bite me in the bee-hind later, because I do recognize there’s value in discretion, and I don’t want my incessant need for verification to prevent me from being a part of more hidden tests in the future. However, something genuinely just happened that made me go “Okay but seriously what the f#ck just happened?”
Did anyone else on here briefly notice a “My ChatGPT” toggle that appeared and disappeared?
Not to be confused with “My GPTs”, which is distinctly different, and is the GPT builder.
This was a toggle somewhere in “settings & beta”, where it specifically described how it accumulates a user’s style, interaction patterns, and preferences across all conversations retroactively. You could turn it on or off.
So, since I like to explore boundaries, capabilities, and limitations, the first thing I did was test it out and see how qualitative the data was, and what information it could glean about my personalized data. I also asked the model to describe the actual capabilities of the feature more in-depth.
Long story short: it was legit. It acknowledged the feature, it described the feature, it described my interaction patterns in pretty good detail in ways that went beyond my custom instructions. This went on for a pretty lengthy duration of the conversation.
Well, all of a sudden, out of nowhere after approx. ~30 mins or so, GPT just pulled a 180 and said “I apologize for the confusion, there actually is no such official feature named ‘My ChatGPT’ that exists.” I go “awwh, it’s hallucinating again!”
I assumed this feature was part of the rollout being done with, well, the dozens of new features that got released overnight. Except for the fact that when I went back to check through OpenAI’s blog, there was no mention of this feature existing or being announced.
I go “Well, that’s fine, maybe the setting toggled off or something?” Well come to find out, I look in my settings, and the toggle disappeared.
It’s gone. Within the span of my conversation, it just went poof.
Guys, I need a sanity check. Can anyone else confirm or acknowledge this feature’s existence? If it’s meant to be kept under private development, can someone DM me to confirm that I’m not going crazy?
I wouldn’t start a conversation with GPT specifically to probe a feature that doesn’t exist. If it was “hallucinating” a non-existent feature, then I wouldn’t have seen the toggle appear, causing me to initiate the conversation about the very feature itself. But considering there’s no public announcement about it, no one has mentioned it in these forums yet, and it literally disappeared right before my eyes, the only thing I can ask myself is “…okay but seriously, what just happened?”
Seriously, this is weird Twilight Zone level stuff. I feel like OpenAI is passively gaslighting me lol. I mean, please let me test more things people aren’t as aware of, but please for the love of god don’t do it in a way that makes me question my reality lol.
To everyone else who reads this: good news? I guess OpenAI’s working on a no-instruction based cross-conversational personalization feature?