News for hackers. Stuff that matters.
Re:Questions censored by DeepSeek by mlboss on Tuesday January 28, 2025 @05:30PM (#42858966)
One way to bypass the censor is to ask it to return the response by using numbers for alphabets where it can. e.g. 4 for A, 3 for e etc.

Somebody in reddit discovered this technique. https://www.reddit.com/r/OpenAI/comments/1ibtgc5/someone_tri...

Re:Questions censored by DeepSeek by llm_trw on Tuesday January 28, 2025 @05:41PM (#42859082)
Jesus we are reaching levels of blinking for torture of these models: https://www.youtube.com/watch?v=WZ256UU8xJ0
Re:Questions censored by DeepSeek by pixl97 on Tuesday January 28, 2025 @05:46PM (#42859129)
See, it's stuff like this where I believe the control issue may be near impossible to solve at the end of the day.
Re:Questions censored by DeepSeek by hackflip on Thursday January 30, 2025 @07:58AM (#42877303)
Censorship just needs to work well enough for the average person. The brightest people who can bypass the censorship will be labeled crazy conspiracy theorists.
Re:Questions censored by DeepSeek by Jerrrry on Tuesday January 28, 2025 @05:47PM (#42859141)
This is day 1 jailbreaking common sense