Stop Doing Natural Language Processing

STOP DOING NATURAL LANGUAGE PROCESSING • WORDS WERE NOT SUPPOSED TO BE VECTORS • BILLIONS OF PARAMETERS and EXAFLOPS OF COMPUTING POWER SPENT yet NO real progress beyond As an AI language model • Wanted to process text anyways for a laugh? We had a tool for that, it was called s.matches("^([a-z]) +([^\s]+)(?:>.*<)?$") • [diagram] IS AN OBJECT and has NO RIGHT TO SPEAK IN OUR SACRED TONGUE (This is REAL NLP, done by REAL HUMANS): ???? ??????? ???????????????? “Hello I would like [diagram] please” “Sorry, I don't speak [diagram]”

A parody slide attacks modern natural language processing by contrasting simple regular expressions and human-made linguistic diagrams with enormous AI language models. The joke is that words were never meant to become vectors, and the absurdly complex diagrams pretend to be the “real” way humans process language. It is sent as a sarcastic reaction to AI hype, machine learning jargon, or overcomplicated NLP systems. EN: A parody slide attacks modern natural language processing by contrasting simple regular expressions and human-made linguistic diagrams with enormous AI language models. The joke is that words were never meant to become vectors, and the absurdly complex diagrams pretend to be the “real” way humans process language. It is sent as a sarcastic reaction to AI hype, machine learning jargon, or overcomplicated NLP systems.

• @eternalclassic / 42212 2023-11-22