Barbara's AI is an autoregressive transformer, trained on The Pile: a diverse dataset of 800+ gigabytes of natural, unprocessed text from the internet. A transformer is a type of neural network that generates output one word at a time by predicting the next word, so if you give it the string "Hi my name is," it'll run it through the network and decide that a name should be the next word.
This is just the prompt you want the AI to continue.
Temperature works like this: The AI knows millions of names, but the way it'll decide on just one name is based on how many times it shows up in the training data. For example, if the name Jonah appears in the training data 10,000 times but Callum only appears 200 times, it'll choose Jonah over Callum. This is great, but not if you're trying to get it to spit out a creative name. This is where temperature comes in. Temperature is just randomness, making the likelihood for Callum to be chosen closer to the likelihood of Jonah being chosen. This means that instead of just using the most likely next word, Barbara will diversify. Choose the option that best fits the end result you are trying to get.
Choose how long you want your results to be!
Repetition penalty is a value from 0-100, punishing the AI for being repetitive. The higher this is, the more likely it will be to talk about new topics.
Return Sequences is how many results the AI returns. The default is one, but if you want multiple outputs change this.