This is what a standing recording booth looks like – putting your head in an Audio Mailbox to maintain quiet background noise while speaking. It gets very hot in there.)
Access to the episodes, but the episodes are all open and uploaded some months later to the general world, including this collection at Internet Archive. Download, listen, remix, whatever you’d like – you have my complete permission and blessing.
Episodes have been uploaded to the Internet Archive since 2019, but in 2022, an opportunity try out new technology came up – the Whisper project, open sourced and instantly downloadable, could be implemented for transcription, either as part of video or just a basic audio file. And thanks to the project, I had many audio files, and began phone number database experimenting with using Whisper against them.
Speech recognition, the process of turning spoken words in a microphone or pre-recorded audio files into written words or issued commands, has been around a very long time – decades and decades. The Internet Archive is excellent for doing a dive into historical citations; a fast “text contents” search found these points of discussion in a 1979 issue of the Silicon Gulch Gazette newsletter:
What has changed is the combination of much faster computers, much more analysis of speech, and advances in cross-referencing the resulting training to make chips and, in this case, a program that is using other disciplines within computer science to pattern-match audio, to the point of adding capitalization and punctuation from the implications in the words. Turning this against my growing collection of podcasts, it wasn’t long before I’d say what has continued to be a theme: when it works, it’s shockingly good, and when it doesn’t, it’s shockingly bad.