 So you want to master the right prompts. Not a problem. I cannot stress enough how important it is to understand that Dolly is a super pattern recognition machine. In the machine's mind, every single word exists in relation to other words. If you give it a word like coffee, it goes through and remembers all the pictures of coffee that I'd seen before, and the other words that usually go together with coffee. For example, a coffee will almost always exist within the confines of a cup or a mug on a table with somebody sitting in a chair and usually the color brown. All of these words are closely related to coffee because nearly all the images of coffee include these words in some other way. And while many of these relationships make sense if you think about them for half a second, a lot of these patterns show up in ways you really wouldn't expect. When I was looking up awesome gymnastics, I generally got what I was expecting. But when I looked up awesome parkour, most of the actors changed from female to male, which makes sense because the overwhelming majority of videos and people doing parkour tend to be male. Another example is when I was experimenting with sci-fi dragon action figures, I got this really cool image and I really liked it so I wanted to run it through a few mutations to see what I would get. And one of the interesting things that happened was I got a picture that looked like this. Now do you notice anything peculiar about this picture? Doesn't it look suspiciously a lot like a horse? The original image was a dragon, so why did it become a horse? And it's here where I realized the algorithm for mutation permeates the original image through a cluster of related words. Dragons are related to fantasy, which is related to the swords and shields medieval genre with kings and queens and knights and shining armor and knights almost always ride horses. So essentially from the AI's perspective, a horse is the closest related animal concept to a dragon. That's probably why it didn't morph into something like a hippopotamus, because I've never seen a hippo in a dragon in the same picture before. So this is cool and all, but you're probably wondering why it matters and how it's useful. So here's why. Sometimes the machine identifies relationships between words and concepts that make sense that you don't necessarily want. For example, I'm a big fan of fantasy witches and wizards. I love spending a few minutes at the end of every day just browsing Pinterest and Artstation and Twitter to find some inspiration for new beautiful witch designs. I have a real soft spot for magic girls, so one of the first things I tried to look up on Dolly was magical fantasy witches, but I encountered a problem. You see, one of the relationships that Dolly has also noticed is between evil and ugly. If you add the word evil into your prompt, it will more than likely generate a character with an ugly face. And the other relationship that I noticed was between the word witch and evil, which makes sense because throughout most of artistic history witches were portrayed as evil. So every time I tried to create a beautiful witch, it would often default to giving her ugly facial features, which I really didn't want. I was really imagining something more like this. So in order to get what I wanted, I had to find a way to describe a beautiful witch without using the word witch, because I didn't want to trigger the ugly evil face that would come with it. And the answer was actually pretty simple. If you type in beautiful witch, it's gonna be pretty hit or miss on getting what you want. But if I type in a young beautiful girl wearing a big magical hat, then that is how you get the classy hot beautiful witch archetype that we want. So you really have to start thinking about these concepts in a way that the machine does in order to navigate and control it. But yeah, hope that was helpful for you guys. And as always, I hope you have a fantastic day and I'll see you around.