The open source models can and do generate those harms. The harms themselves are probably overblown. There are plenty of deepfakes of various celebrities. Mo

The open source models can and do generate those harms. The harms themselves are probably overblown. There are plenty of deepfakes of various celebrities. Mostly people can tell the difference or they just don't care. I think the reality is that training these models and paying ML engineers is incredibly expensive. Not a good fit for the open source model, thus OpenAI had to convert to SaaS.