We fell out of love with voice assistants | Hacker News Reader

We fell out of love with voice assistants

bbc.com

16 pointsSirLJ9 comments

In 2011 you could basically mumble and tap faster than ever into a reactive knowledgebase of everything.

Now you're lucky if Google Maps doesn't have it's own unique separate context that is annoying and never works, or some contractors at Amazon aren't giggling at you.

My favorite case on Maps context, is that the assistant (over the past 3 years!) is unable to tell you what your destination is. I've stumbled across that a few times upon upgrading my phone and thinking the assistant might be worth using, but it's just clunky, and I'd rather just organize things so that I can operate my phone without looking at it.

I don't think I'm becoming a technophobe, I'm still open to the idea of voice assistant, it just seems to be permanently crap / vaporware.

I'm working right now on adding voice control to Home Assistant: https://www.home-assistant.io/blog/2022/12/20/year-of-voice/

Our goal is local voice control of your devices in your native language (no Internet required). Hardware is still an open question, but I believe we will succeed due to our focus on a limited domain (IoT) and large open source community.

I'm a home assistant user that's been following along with the year of voice. I played a little with Almond and some with Rhasspy; what's the best way to get up and running right now? I've been using my arch desktop with a mic (running through pipewire) as my primary hardware, not scared of deploying to a pi or similar but I don't have one to spare right now.

I have this problem with HomePod. I ask it turn the TV off and it sits there thinking. It will then say it’s not online. I’ll yell at it, and maybe the second time it complies or says it’s offline again. At that point I just hit the power button on the TV remote.

I mainly use my HomePods for turning the TV off or setting a timer.

Setting a timer is pretty much the killer feature in my house.

Can we skip chat bots already too?

I'm impressed by the demonstrated capabilities, but I worry that they can't run outside time-sharing central machines. What's your issue?

Why do you want to chat with a bot? Wrapping an otherwise unhelpful decision tree in a bot that echoes vague pleasantries, has no real context of my issue or past interactions, and no direct control over any account/service changes certainly does not lead to a pleasant experience for me.

If it can competently translate language into UI actions and feedback it beats what's already there, though yeah a $41M training session for every interaction step is obviously out of the question atm. Maybe after a few Moore epochs.

They look so cool, unless you have an accent, then it is a more-less random reply / action generator.

Reply on news.ycombinator.com