I say to a Next Hub Gen 2:
"Hey Google, one hundred fifty divided by thirty four"
It recognises, "150/30 for" (shows that on screen) and gives me the result for 150/3. At no point did I say "zero" or "naught."
I re-tried it multiple times, saying "34" quickly, or more quickly.
I tried this on a Pixel 7 with a Tensor 2 chip. It heard "50/34"
There are a many issues:
- There's no reason it should lack an adequate voice model, or have trouble recognizing what I say. I have long paragraphs I have dictated to Google Assistant stored in my Google account from the past. I speak English fluently. I know this because when I setup the Hubs, it didn't ask me to create a voice model. It just used what it had stored.
- Assistant voice recognition window is not long enough, so you're constantly having to rush what you say, or if you pause, it will take what you said even if you weren't finished, and give you a bad result.
- Why is their no option to retain the voice model on specific, common sets of words like numbers?
- Why is the AI so dumb, that it hears me say "150/3 for", and it assumes I literally said "for"?
- When I say the same thing multiple times, it doesn't, smartly, say "you've said thing multiple times. What am I getting wrong?" At least with Google Assistant on Pixel, you can edit the text that it detects. Why can't a Nest Hub do that? It has an on-screen keyboard. It could use that to improve it's voice model.
Why does Assistant voice recognition typing on Android let you pause, and stay open, but Assistant on Nest Hub or Mini will act on your words the moment you pause? Why is there no "continued input window" like there's a "continued conversation" option? I'd rather have a manual word to say to get it to process what I said, then have it try to auto-determine that and get it wrong constantly, or cause me to rush, or have to think through what I want to say before I say it, instead of while I say it.
Every time I want to report an issue like this, I have to boot up a computer, write it up, and post it. I can send feedback on the device, but how can I do it accurately if it can't interpret what I'm trying to report? Why isn't there some way to flag an interaction, or manually correct feedback reports with an on screen keyboard after dictating them, like you can on Chromecast?
- None of these are hardware issues. The Pixel has a Tensor 2 chip. The Hub is a Gen 2. These are software problems.
I hope you are beginning to see that EVERY DAY, these Google Home devices fail at doing basic things and I am forced to either
- avoid it's issues and do things a harder way
- not do them because I can't rely on it
- do it manually.
Which means there's no point to having smart home devices.
