The Importance of Reinforcement Learning in Chat GPT

The bot has been trained to respond to this kind of, can you please tell me what the capital of Egypt is versus capital of Egypt go. The foundation model before I had been fine tuned before have been instructed. You could get the same response out of it, but you would have to construct a really elaborate prompt that is not just asking for an address. And so when you start out with stable diffusion 1.5 and you don't know what you're doing, you're just going to get pretty trashy looking images. But mid journey is like better at taking a short garbage prompt and making something spectacular with it.