Hacker News Reader

Now just need a plausibility checking AI to vet GPT-3 outputs so the author doesn’t have to read through 10 versions of his potential blog post.

That's how generative adversarial networks work, right? One neural net to create, another to check if it matches the training data.