(GitHub) Copilot doesn’t work if profanity exists | Hacker News Reader

(GitHub) Copilot doesn’t work if profanity exists

github.com

40 pointsgwbas1c12 comments

I've noticed similar refusal to auto-generate anything if certain other kinds of words exist in a line.

I'm working on adding some AI text generation into an RPG, trying to liven up the NPCs a little bit. For flavor text, the model gets fed information about the name, race, class and gender of the character they're playing. I've noticed that Copilot will happily autocomplete lines about the character's name and class. But the words 'race' and 'gender' completely halt the auto-completion mechanism. I have to type the entire line out by hand, copilot won't suggest a thing, even when talking about Elves and Orcs. The next line it goes back to normal.

I can't say I'm confused why they did it, though. There's definitely a certain kind of person who would read way too much into whatever hallucinations the model comes up with in those areas.

It also doesn't help that lots of previous attempts at AI have been gamed by 4chan into spouting super racist rants out without provocation when user content is used as part of the learning model. (see MS tay for a highly visible example)

If you're doing this for a game, I suggest trying a camelid like the alpacino model, vicuna uncucked, or wizard uncensored.

You'll get much more interesting and creative results than any OpenAI model. The open models are really outperforming GPT-4 in this area.

I think I might have phrased my original message poorly. I'm not using Copilot/GPT to directly generate the text in the game. Copilot is only helping me to write the code that generates a prompt, which then gets sent off to whatever other model I want.

But the suggestions are appreciated, though! I have a framework that allows me to swap out arbitrary model 'back-ends' on the fly. So I can definitely take a look to see if these work well for my case.

I confirmed this is legit. Though you can mask your profanity slightly and it gets right around the filter (ie, fuck -> fu*k, etc). Still, if you pay for a product it should work as intended... we're all mostly adults.

If only we could deny Copilot from pulling your code by including # Fuck off Copilot #

for years and years people have been getting into trouble for explicit comments in source code -- little did we know that they were time-travelers trying to keep their secrets out of the hands of BigAI.

I was hoping that including profanity would prevent code from being hoovered up by Copilot, but that doesn't appear to be the case. That's too bad.

So Microsoft's Windows team can't use it then.

Allegedly.

Yeah not a bug and seriously annoying, literally just one word is enough for it to stop generating.

It senses the dev is upset and doesn't want to get into trouble so it keeps quiet. Yet another example of emerging, human-like behavior ; )

My shitty code is safe then.

Reply on news.ycombinator.com