It seems there is a new action where an up or down keyboard arrow will toggle between suggested messages - and wipes the existing input contexts of the input box. This happens even several inputs into a chat with the new push that broke other stuff for others.
I go as far as shutting down all the browser productivity windows, restarting the browser with no extensions filtering or modifying things, hard-refresh the contexts, think that I can’t reproduce it on existing chats or a new input window with some pastes and start going back to work producing original words. Again BAM content destroyed with a typical editing key stroke.
Yep. Images triggers it when everything else is disabled.
Going as far as to block the monitoring of every keypress sent to chatgpt.com/backend-api/conversation/experimental/generate_autocompletions
Clearing all the local storage manually.
Still there’s a script serving up mars ducks on its own without making a request.
I even went in and killed the prosemirror version of “ask anything” input and enabled the prior input in parallel, but its events are invalid or unserved JS in this gated feature.
@OpenAI_Support This issue has a reversion and has returned.
I was just now typing along at about line three of filling up the input box on a new chat after having pressed the create image button, made the mistake of pressing the up key, and all my typing was wiped clean, replaced with a stupid scripted image prompt suggestion again.
In the 30 years that I have worked in both Web Development and Operations I have NEVER seen this type of aberration in programming. This “Create Image” button fiasco happens to me in every browser, after deactivating ALL my extensions, and in spite of applying numerous suggestions from GPT4. Once pressed, I’m screwed. Every time I enter text in the chat field and may mindlessly use my arrow keys to move up, down, left, or right, the pos injects its suggestion into the box overwriting my text. Nevermind turning it off . . . It persists . . .If I am creating images I must paste my information in the field and ohhh . . .everrrrr soooo carefully tap the “send” key. One screw-up and I lose my request and end up with some completely notorious, useless and random image. (Oh crap! I just tapped the DOWN arrow key!) Did it happen? Is my text gone?
Phewww. . . No.
To add insult to injury, This happened when I was accused of policy violation. Chat had created numerous iterations of a “Glass Removal Suction Tool” and on the 11th iteration, I see in red bold text “this image violates our content policy”. I copy the EXACT same request I’d made numerous times - this time, asking for the metal handle to be changed to a wooden handle. I hit the DOWN arrow (just to get to the bottom of the field), hit send and, VOILA . . .an image of a brown vase with marshmallow flowers on a tilt-a-whirl appears. I wonder if Open AI should throw China a couple of bucks to see if they can produce someting better?
Shared this with the team, thank you all for your input. It looks like cmd + z restores the original prompt, but I agree that an up arrow to restore would be clearest!
