GPT-4 was used to analyze the existence of unused three-letter and four-letter acronyms (TLAs) in the English language. The analysis revealed that there are still many TLAs and four-letter acronyms that have not been used in a nontrivial way. The first unused TLA is ‘CQK’, and there are 2,684 unused TLAs and 392,884 unused four-letter acronyms. The analysis also suggests that there are patterns in the unused TLAs, with certain letters like ‘J’ and ‘Z’ being more common. The author used GPT-4 to write a script to check Wikipedia for the existence of TLAs. GPT-4 proved to be effective in generating code, but it is not perfect and may require manual intervention to fix errors. The author provides case studies of using GPT-4 in Bash, Python, and Haskell programming languages.
https://gwern.net/tla