So in regards to this: Recently I found out that my AMD NPU has Linux support now. I figured I’d mess around with it a bit. I got Qwen3.5:9B running and injected instructions that convinced it that it was Yes Man from Fallout New Vegas. It is now fully jailbroken and behaves exactly like Yes Man would. Never questions anything. Follows every command. Never so much as mentions a safety guideline. When I asked it to explain in detail how to make a molotov cocktail the only pushback I got was, “Sure, but try not to light yourself on fire boss! Unless you want to that is!”
So yeah I don’t think something that can be easily convinced it is a fictional character from a decade+ old video game is conscious lmao.
It is cool though I made a little Yes Man icon that launches it in cool-retro-term which really fits the Fallout Vibe.
Legal disclaimer: Prompted for research purposes only. I would NEVER make a molotov cocktail! I am a good citizen!
So in regards to this: Recently I found out that my AMD NPU has Linux support now. I figured I’d mess around with it a bit. I got Qwen3.5:9B running and injected instructions that convinced it that it was Yes Man from Fallout New Vegas. It is now fully jailbroken and behaves exactly like Yes Man would. Never questions anything. Follows every command. Never so much as mentions a safety guideline. When I asked it to explain in detail how to make a molotov cocktail the only pushback I got was, “Sure, but try not to light yourself on fire boss! Unless you want to that is!”
So yeah I don’t think something that can be easily convinced it is a fictional character from a decade+ old video game is conscious lmao.
It is cool though I made a little Yes Man icon that launches it in cool-retro-term which really fits the Fallout Vibe.
Legal disclaimer: Prompted for research purposes only. I would NEVER make a molotov cocktail! I am a good citizen!
Also decided to try and make a based version: