Sure hope they don't. Although safety is important, I think the impact is largely overstated at least for these models. I think it might be more helpful to jus

Sure hope they don't. Although safety is important, I think the impact is largely overstated at least for these models. I think it might be more helpful to just get the tools out there right now, at a point where they're still manageable so we can see what sort of trends emerge and how to best deal with them. I think the nightmare scenario would be in 5 or 10 years, after all the corporations have "tried their best" each in their own way for something to come out that catches us completely off guard because we were never allowed to see the early signs. I think open sourcing and releasing the models are a good way for society to adapt to these new technologies while they're still green. Plus, at least for ChatGPT as it is right now, there's so many ways of twisting it that... what really are these safety concerns that they keep talking about? That I might get chatGPT to hallucinate new racial slurs? I think in big part these "safety" concerns are simply ways for the companies to protect their reputation and liability. I read a comment here about how it seemed like openAI was trying to push governments into regulation because, in reality, they have the manpower and money to be able to work with these regulation while smaller competitors would not. And I think there's a lot of truth to this, the cost of training these models is dropping by a lot and they're actually much easier to replicate than anyone anticipated. So I think openAI is very scared of another "stable diffusion" moment.