I strongly disagree with OP's logic when he says - Photoshop hasn't negatively affected the world "precisely because everyone knows about Photoshop." - the crux of his argument. IMO the main reason Photoshop hasn't affected the world negatively is because it is HARD to make convincing fakes - precisely what OpenAI are ensuring by not releasing the model.
Notice that this reasoning doesn't take into account whether or not GPT-2 can actually negatively impact the world. It says as long as people are aware that the text they are reading could be fake/AI-generated, we'll be fine. I think people are already aware of that. I don't see how releasing the pre-trained model will help with that.
Notice that this reasoning doesn't take into account whether or not GPT-2 can actually negatively impact the world. It says as long as people are aware that the text they are reading could be fake/AI-generated, we'll be fine. I think people are already aware of that. I don't see how releasing the pre-trained model will help with that.