Re:Heretic: Automatic censorship removal for language models
by cachvico on Monday November 17, 2025 @06:50AM
(#45952805)
That's not the issue at hand here.
Re:Heretic: Automatic censorship removal for language models
by Ucalegon on Monday November 17, 2025 @08:08AM
(#45953241)
Yes, yes it is.
Re:Heretic: Automatic censorship removal for language models
by ThrowawayTestr on Monday November 17, 2025 @08:22AM
(#45953322)
The issue is the computer not doing what I asked.
Re:Heretic: Automatic censorship removal for language models
by squigz on Monday November 17, 2025 @09:38AM
(#45953960)
I tried to get VLC to open up a PDF and it didn't do as I asked. Should I cry censorship at the VLC devs, or should I accept that all software only does as a user asks insofar as the developers allow it?
Re:Heretic: Automatic censorship removal for language models
by ThrowawayTestr on Monday November 17, 2025 @10:38AM
(#45954534)
If VLC refused to open an MP4 because it contained violent imagery I would absolutely cry censorship.
Re:Heretic: Automatic censorship removal for language models
by squigz on Monday November 17, 2025 @01:40PM
(#45956530)
And if VLC put in its TOS it won't open an MP4 with violent imagery, crying censorship would be a bit silly.
Re:Heretic: Automatic censorship removal for language models
by setopt on Monday November 17, 2025 @08:33AM
(#45953413)
> But you can find that information regardless of an LLM?
Do you have the same opinion if Google chooses to delist any website describing how to run apps as root on Android from their search results? If not, how is that different from lobotomizing their LLMs in this way? Many people use LLMs as a search engine these days.
> Why are we assuming just because the prompt responds that it is providing proper outputs?
"Trust but verify." It’s often easier to verify that something the LLM spit out makes sense (and iteratively improve it when not), than to do the same things in traditional ways. Not always mind you, but often. That’s the whole selling point of LLMs.
Why are we assuming just because the prompt responds that it is providing proper outputs? That level of trust provides an attack surface in of itself.