Reasoning engines like ChatGPT used something called a large language model to help the system come up with a human$-$like response. How do these large language models generate text?

They string together several search engines to create one massive response database.

They crawl the web in real time to train the response to match your prompt.

They rank the words in a response and then go through them in a linear fashion.

They use probability to predict the next words in response to a given prompt.