[aaronsteers]

Although not evil, adult content should be opt-in, and should be able to be opted-out at a platform level... hence, the need for censored models. Imagine a restaurant booking AI app, built on GPT, that accidentally doubled as a bomb-making tutor or an adult content generator. It's a lawsuit waiting to happen, if nothing else, and it's worth making these use cases harder (if not impossible) to implement in mainstream, commercially available products. Note that for many of these products, the age and consent for adult material has not been already established.

So far, the open source ecosystem seems to be doing a good job of providing both censored and uncensored LLMs - and it seems there are valid use cases for both.

Think of this as similar to Falcon LLM being launched in both 40B and smaller 7B variants - the LLM often will need to match the use case, and the 7B model is a good example of making the model smaller (and worse) on purpose in order to reach certain trade-offs.