4 ms·
How would you explain the dozens of other open source versions of computer vision and language models that didn't generate all those harms, even ones that were
by aabhay 5y ago
How would you explain the dozens of other open source versions of computer vision and language models that didn't generate all those harms, even ones that were trained to recreate the exact models that OpenAI withheld due to those concerns?
- aeternum 5y agoThe 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.