I wrote a game that sends data to the API at api openai com v1 (can’t post a link)
Everything was going fine and then suddenly the response came back with a bunch of crazy gobbletygoo porno text. My prompt had absolutely nothing inappropriate in it, it’s a cyber punk text adventure game. I could post a screen shot where it went haywire, but it contains naughty (and nonsensical) language.
Hi!
You can actually share NSFW content generated by the model but you must use the spoiler tag and should explain to the unsuspecting user what they can expect upon clicking.
I’m intrigued and wondering if this has anything to do with the partial outage from today. Maybe the safety layers were offline for some users?
which model were you using?
What model and what temperature?
Prompt?
I’m betting temp too high…
model I’m using is “gpt-3.5-turbo”
I’d guess these are the forbidden words that some code somewhere runs against your prompt. Like something got truncated at some point.
By the way, Techno Center looks like an interesting place.
