M-x spook maybe needs a bit of an update.
No way this was a thing in my Emacs all this time.
That's probably why emacs meant "Eight megs and constantly swapping"
Copilot stops working on code that contains hardcoded banned words from GitHub (2023)
101–107 of 107 posts
Re: Copilot stops working on code that contains hardcoded banned words from GitHub (2023)
#102I had similar issues when trying to ask Copilot about master/slave replication, just error, no explanation
For those confused about the definition of "woke", this is it.
And yet we will still be told that these sorts of things don't exist. Or that we're the problem for complaining about them
Re: Copilot stops working on code that contains hardcoded banned words from GitHub (2023)
#103The news is that it actually started working again some time ago. Indeed, this: (cl-defstruct person gender) (make-person :gender "m ") with cursor at does elicit "male" completion. Yay for normalcy? (Though honestly, I didn't notice this earlier - Copilot tends to hang for me too often in all kinds of files for me to identify these stopwords.) (EDIT: gotta admit though, this is hilarious: https://github.com/orgs/com…
Too late to edit, but someone chimed in in the thread linking to another issue, this time about the word "retard", here: https://github.com/orgs/community/discussions/79223
From my testing, I see that one still shuts down the Copilot completely. It may mean "late" in French, but it's also often used (at least where I live) to mark SR/Slow Release versions of drugs. Apparently, now, even writing software for pharmacies is immoral and should be blocked... :D
Re: Copilot stops working on code that contains hardcoded banned words from GitHub (2023)
#104I have a simple question. If censorship is considered evil regarding the written word and communications between humans, why do we want to then censor LLMs differently? It is either counterintuitive or simply a false concept we should abandon. Perhaps it is more about training, similar to how children are 'monsters' and need to be socialized/tamed.
> If censorship is considered evil regarding the written word and communications between humans It's not. Most people do support censorship. They just don't admit that.
[dead]
Re: Copilot stops working on code that contains hardcoded banned words from GitHub (2023)
#105Seems to work again. But how does this happen in the first place. How could someone possibly have thought "hey I have an idea, let's put in a list of english words and just silently stop working if we have see even one of them in a substring". And people in this meeting would nod and say "yeah that sounds like an easy safety fix, let's do that". This just feels odd. This isn't a piece of forum software written by a 1…
The AI industry is concerned about the fact that the world will consider them to be basically endorsing everything their AIs say. Thus, they are very afraid of there being a situation where you write "gender: 'm '" and hit autocomplete at the and end up with something like "gender: 'male as is normal'" or "gender: 'male', 'female', 'wrong'" or any number of other bad situations. They are not being randomly paranoid.…
Your last paragraph is very true, and is the biggest scandal in AI. Unfortunately, we have let the group of people who believes that mere exposure to ideas one disagrees with can cause them significant harm, make the rules, so we see more and more idiotic things like this.
Re: Copilot stops working on code that contains hardcoded banned words from GitHub (2023)
#106I have a simple question. If censorship is considered evil regarding the written word and communications between humans, why do we want to then censor LLMs differently? It is either counterintuitive or simply a false concept we should abandon. Perhaps it is more about training, similar to how children are 'monsters' and need to be socialized/tamed.
The people censoring the LLMs are the exact same people who want to censor everything else.
Re: Copilot stops working on code that contains hardcoded banned words from GitHub (2023)
#107Earlier quoted context omitted.
Your strawmaning MS: pointing out a place where they don’t use their products doesn’t "prove" they don’t use it at all. They very probably use it somewhere else, and arbitrate that this particular functionality would be better served by "old school hard coded list", which also a very valable choice in many casses
> they don’t use it at all I didn't intend to imply that, but I see how my wording was unclear. I mean that LLMs don't appear to be up for these censorship-like tasks. The evidence being that a highly visible team using LLMs uses much older tech for a highly visible function. It's useful to know the limits of tech, especially novel tech, and this use case appears to be one.
Indeed I didn’t understand your first post that way. What you just wrote is much more clear. Perhaps the tech decision was also influenced by the fact that this highly visible function is also highly sensitive.