Two or more LLMs are running at once, one does the generation that you want and the other watches the output closely to see if it triggers one of their no-nos to do the censorship retroactively. If this sound inefficient, and absolutely terrible user interface design, you’re right.
If you’re running it on your own hardware you take the liability directly for doing any dumb thing it tells you. The first LLM is there to provide you customer service, the second LLM is exclusively to keep you from suing the hosting company. If you want to sue yourself for the output of your own self-hosted open weights LLM you’re more than welcome to!
Two or more LLMs are running at once, one does the generation that you want and the other watches the output closely to see if it triggers one of their no-nos to do the censorship retroactively. If this sound inefficient, and absolutely terrible user interface design, you’re right.
Seems sort of old that they would then release open weight models so the restriction can just be removed.
If you’re running it on your own hardware you take the liability directly for doing any dumb thing it tells you. The first LLM is there to provide you customer service, the second LLM is exclusively to keep you from suing the hosting company. If you want to sue yourself for the output of your own self-hosted open weights LLM you’re more than welcome to!