undefined | Better HN

0 pointsVLM11y ago0 comments

Your post has 47 words. A 95% success rate means 1 in 20 failure, so you're going to have to stop, swear, fix it, and restart dictation about two or three times in your post. Even the somewhat optimistic 99% still gives only a systemic fifty fity odds of your post being error free.

Its just too annoying for most people to use.

What voice recognition reminds me of, is a disobedient / distracted little kid, the kind you have to tell everything twice.

0 comments

3 comments · 1 top-level

tomhenderson11y ago· 2 in thread

> A 95% success rate means 1 in 20 failure

I'm not sure my success rate is any better than that typing on my iPhone.

TeMPOraL11y ago

But the correction is simpler and less distracting. I can backspace the word I misspelled and fix it without breaking my thought flow, whereas when I try that with dictation... well, it doesn't work that way. Especially that the whole point of dictation should be that you're not actually looking at the screen and double-checking if what's recognized is what you said.

VLMOP11y ago

"Especially that the whole point of dictation should be that you're not actually looking at the screen"

I've noticed UIs always get "needier" over time. Call it VLM's Law. In the old days a dumbphone could be operated tactile in total darkness without looking at it at all. Then smartphones came along which needed occasional glances at the screen to verify the screen and touch screen position/orientation (portrait vs auto-rotated landscape). Now you need to stare continuously at the screen to watch mistakes in the speech to text dictation system as you slowly, tediously, agonizingly argue with a machine.

How clumsy of a UI, like going thru life wearing oven mitts. And VLMs Law is it always gets worse over time. It'll take a lot of work, but we'll find a way to make UIs even more painful.

1 more reply

j / k navigate · click thread line to collapse