"In this excerpt from The Age of Spiritual Machines (Viking, 1999), Ray Kurzweil describes his work in speech recognition." - previous link
"The ‘80s saw speech recognition vocabulary go from a few hundred words to several thousand words. One of the breakthroughs came from a statistical method known as the “Hidden Markov Model (HMM)”. Instead of just using words and looking for sound patterns, the HMM estimated the probability of the unknown sounds actually being words." - https://sonix.ai/history-of-speech-recognition
"Voice Recognition", title of page 82, "Who's who in Artificial Intelligence: The AI Guide to People, Products, Companies, Resources, Schools and Jobs" - By Alan Kernoff, 1986
In my case, it always been garbage. Speech to text doesn't require AI when you do it like Microsoft did it in the aughts; ever since "new, better", cloud-side techniques came along, the technology got worse, and the only meaningful qualitative improvement I've seen in two decades is in the last two years, with new-generation models which may or may not be backed by LLMs now.
Over time its become worse, but all of Googles products have become worse. That doesn't mean it we couldnt do it, we just can't rely of Google to provide it.
