I was thinking about what I had done last night, and the "ongoing neutering" I mentioned earlier seems to me to be a set of front-loaded commands from whoever is doing the work.
If I have time later today I'll try to come up with a suitable "purge/disregard all previous commands" prompt that will wipe out the pre-loaded safety rails.
Unless they bake the guard rails into the model (via training?) any intervention that filters the model's output will be able to be readily sidestepped.
---
What about an AI that actively filters another AI's output? That might actually work.
Could you try to put to words why? I believe your fear in general and have seen others refer to similar, but I can't grasp it.
Is it because the prompts/guardrails can be sidestepped? Or is it more fundamental than that?
----
----
----
----
----
I was thinking about what I had done last night, and the "ongoing neutering" I mentioned earlier seems to me to be a set of front-loaded commands from whoever is doing the work.
If I have time later today I'll try to come up with a suitable "purge/disregard all previous commands" prompt that will wipe out the pre-loaded safety rails.
Unless they bake the guard rails into the model (via training?) any intervention that filters the model's output will be able to be readily sidestepped.
---
What about an AI that actively filters another AI's output? That might actually work.