We are constantly bombarded with organizational rhetoric, whether it’s advertisements, press releases, or rich people humiliating themselves on podcasts (cough cough Mark Zuckerberg cough). With all the messages organizations are sending, we need rhetorical criticism now more than ever. Understanding how organizational rhetoric is analyzed will make us all better rhetors and better consumers.

Hi guys! John the Quant here. In this series of articles, we are going to explore each piece of the architecture that makes models like GPT, Llama, and Claude possible. All of the big, generative Large Language Models follow the same basic architecture.

If I tell you a person is a King, what else do you know? You know the person is royal, and is male, that they lead a country and that their family is also royalty. That’s a lot of information from just a single word. We’re going to need more than a single number to convey that much information. This implies that words are multi-dimensional.

Now, what’s the difference between a King and a Queen? Most of the information in the previous paragraph is true of either a king or a queen. The only difference is gender: King implies Male and Queen implies Female. We have stumbled upon a kind of “semantic arithmetic”.