ChatGPT is essentially just a text predictor.

It is trained on basically all the text on the internet, and it uses this to learn what words tend to follow what other words. It's very powerful and sophisticated, to the point where it can write proper english sentences which are on topic and which are (mostly) accurate.

So if you say, "Where was Elvis Presley born?", it predicts that after this text would generally come text which gives the answer to the question, and that's the text it gives you. And because it has been trained on the text of the entire internet, it knows the answer to this question.

If you say, "Please write me a brief essay on the difference between capitalism and socialism", it predicts how such an essay would likely start, then writes that text. Then predicts how such an essay would likely continue, then writes that text. And so on, until the essay is completed. As it's been trained on the text of the internet, it has large volumes of text in its training material about capitalism and socialism and the differences between them.

ChatGPT is specifically trained to be a chat bot, and it probably has multiple censorship routines and "be a nice chatbot" routines which identify when your prompt or its own writing is something against its rules.