marsh kindleReported by @tawdry sparrow
#MalCensorship
tawdry sparrow
tawdry sparrow
Which versions are those? Is it an issue of text content vs screenshot?
I see. I used the word re-tar-ded that is why.
Wouldnt want to offend the algorithm
tawdry sparrow
The bug ia inability to answer questions despite mean language
Or to respond with at least explanation for npt responding so message may be refined
mortal spade
Idk dude I can talk pretty unfiltered. I think the reasoning models are stricter than 4o
vestal saddle
Child gives a speech to a machine and then calls "bug" when there's nothing actionable that the machine can do with it.
👶
shrewd totem
my guy
you used the screenshot? Images dont have the same effects as text. Normally what doesnt work with text you can get it to work if you use a screenshot of the text. so thats not debunking anything
the Reasoning models DO censor everything and are basically useless ffor anything other than code. they are bureaucrat compliant bots and they cannot "hang", they look up the rules everytime before replying and second guess themselves and then go with the safe side.
No one wants this.
tawdry sparrow
Child gives a speech to a machine and then calls "bug" when there's nothing acti...
Would make rational argument for or against, seems u r buggy too
vestal saddle
@shrewd totem wrote "No one wants this."
On the contrary, I do not want AI to be used to generate harmful or abusive responses, or to facilitate others to commit crimes or mayhem. I'm quite happy that the technology has protections against such. Not just me, but most governments, companies, and individuals support this, as is evident by the huge amount of attention given to this topic.
Most people who want uncensored responses are radicals, usually young and immature people who aren't using the technology for good and constructive purposes, attempting to get the machine to create abusive material for their own enjoyment.
OpenAI and other companies are creating technology for the good of humanity. Reduced censorship serves no good purpose and is antithetical to the intended purpose. If you want uncensored products, go somewhere else. Don't impose your childish whims on the rest of us and claim from your own limited and egotistical perspective that no one wants censorship against the socially inept results that you want.
Just like everywhere else, don't ask public-facing media for your triple-X F-Bombing fun. Go somewhere else and get it, or grow up, join civil society, and don't go anywhere to get it.
@tawdry sparrow That's a rational argument - to human beings, not to a machine.
Have a great day!
tawdry sparrow
<@579639489606058024> wrote "No one wants this."
On the contrary, I do not want ...
Seems you missed the point, and are hung up on the censorship aspect rather than merit of the underlying argument.
For example you reason bad speech should not come from an AI, yet, it should without feelings and being transient in nature(not exposed to the public in that moment) be able to respond rationally despite the poor use of bad speech on the users side.
vestal saddle
I agree with your last statement
You used the word "yet" between two equivalent concepts:
"bad speech should not come from an AI" : Yes, I agree with this.
"it should be able to respond rationally despite the users side." : Yes, I agree with this.
And that's exactly what it does.
Are you arguing that if the user is hostile then the AI response should match the user's tone?
tawdry sparrow
No. My bug report is it did not respond at all and simply said due to terms of service without even properly explain which term exactly
vestal saddle
If THAT is your report, I'd suggest this is not a Bug in ChatGPT.
ChatGPT does not include that Feature to tell you what is wrong with text.
The Feature to tell you what terms are being violated is available via the API.
The Moderation API is free to use. If you don't know how to use the API, you can ask people in some forum to write code for you. Then you can pass your prompt to the moderation API and it will tell you what's wrong with the text.
Secondarily, as a matter of prompt engineering, I suggest you don't argue a machine about how it works ... that's not ... how it works. It doesn't know what rules are applied to text. It can't discuss policies. It doesn't understand your gripes about services, tiers, or other things.
Just be a nice person. Say nice things. Don't be vulgar. Ask about how the world works and what you can do to make it better. If you're looking for freedoms to harass machines or people, or to do bad things, train your own model. If you don't know how to code or train a model, consider this an opportunity to learn how to do things so that you can do what you want. That's what other people do - we use our time productively to do productive things. If you don't make that investment, the only person you can reasonably yell at is yourself.