Yeah. They did exactly that
I disabled the Google App back when “Google Now” was still a thing. Remember when it would give you directions to get where you were going after you were already on your way? I’d be on the train, it’d tell me “oh, you wanna go somewhere? Get off the train, take a cab to the nearest train station, get on the train…”
They removed everything but sports score tracking, I kept using it for a while, and then I realized that I could just fucking use my browser for search, since that’s where I wanted to read search results anyway. And that’s what I did.
They’re going to keep doing this again and again, making their app worse and worse.
No idea why I would ever want the Google app back.
Only realized how Terrible Google assistant was when I started to use it with Android Auto.
To be fair, to some degree this is exactly the use case stochastic parrots (the thing we call ‘AI’ as a buzzword) can truly excel at:
- Interpeting the bullshit we stammer out when we try to give a verbal command while totally not adhering to any reliable command structure.
- Formulating a reply that sounds like fairly natural language despite how inane the sources used might be.
So yeah, we’re finally at a real use case. Gimme! And from briefly trying it, Gemini is better at figuring things out from impresice input than Assistant was.
Make no mistake, it’s ultimately the same backend. They just swapped the processing layer between audio-to-text parsing and running inputs from them (and again on the way back). Sadly no Google Now smartness at all, we’ve lost that forever. But hey, at least this improves stuff.
Probably not the direction they’re going, but having this run offline like most of Pixels “AI” features would be great. I think the hardest part is the dataset training though, just having an offline assistant that works would be a win even if it wasn’t a LLM.
I’m trying to put distance between me and these data collection points and it’s hard.
It’s also so slow to answer basic questions.
lEaDERShIp
So Bob… We meet again
Hi Bob!
I’d do unspeakable things to get Google Now and Now on Tap back.
worse? What did I miss - it was never good to start with. Alexa, Siri, Cortana, Google Assistant - all they were ever used for was set timers and play songs.
AI is the only hope to make them marginally more useful than they are.
deleted by creator
Sometimes it can’t connect to the server (which is a completely stupid necessity).
That’s where it does the voice processing. The only processing it does on-device is the wake word and taking commands. Actually figuring out what you mean is done in The Cloud. Doing that on-device would not only make the devices significantly more expensive, but they would also rapidly become outdated.
The rest of your complaints are valid and I’ve experienced them all myself to boot.
deleted by creator
You can get a Pi that can do voice to text and NLP, maybe it can even do it with reasonable speed.
You can get a Pi that is $35.
But you won’t get a Pi that is $35 which can do voice to text and NLP with reasonable speed.
When you say “Android devices” you’re talking about devices that are like, 5-10x the price of the nest mini. Of course they’re capable.
deleted by creator
We saw a glimpse of some great mobile awareness in Cortana. Contacts and location reminders were amazing on windows phones.
“Next time I talk to my wife, remind me to ask bout the dogs medication.” Phone call, text, or email, I’d get a reminder.
“Next time I’m at the store, remind me to buy bread.” “Which store?” “Any grocery store.”
Phone assistants have basically been going down hill since windows 10.
Location-aware reminders is almost literally all I want from an assistant these days. “Remind me of x next time I’m at y, or by z time at the latest.” Is this an impossible task? I can imagine how I would code it, but maybe I’m missing something.
I think they just haven’t figured out how to monetize it, really. I agree that it’s totally codable. I could do it with tasker if it were my job, IE, I was paid to and had 8h a day to do it.
Where’s the monetization in alarms, though? Or time-based reminders? Surely those are no more lucrative than what I’m asking for, yet they’ve existed for years.
set timers
This broke for me a few months ago. It just randomly… won’t start, despite saying otherwise.
My wife cannot set timers on our Nest Hub. It just doesn’t understand her command. I’ll say the exact same sentence right after and it’ll work. We did reset her voice profile, remove/add her back, checked all settings possible, nothing worked. Such a decent piece of hardware (speakers are actually pretty good, and the screen is decent and bright) that’s ruined by shitty software. It’s been unplugged for the last month and I didn’t even care. It’s going on Marketplace next week lol
My wife has similar issues. I think it was just trained on male voices.
I think this happened to me once when the assistant was trying to use another clock app than the native one.
“Set a timer that goes off at 9:15 am”
*It proceeds to lecture me on the difference between an alarm and a timer, also, sets neither. *
Ah, I think the wording confuses it.
Timers are set for a duration. Alarms are set for a time. Which makes sense btw, you can’t set an egg timer to 9:15 either, you set it for, say, 21 minutes (if it’s 8:54 right now). And you don’t set your alarm clock for “in 6 hours”, you set it for 8:00.
It’s a bit arbitrary, but this is exactly where I feel models such as Gemini or ChatGPT can actually improve things, because they can more readily leap from the keyword “timer” expecting a duration to that you actually meant “alarm” from the rest of the input, you just said timer instead.
Yeah I understand, I got the lecture from Siri.
The point is all timers are alarms, the end result of a timer going off is an alarm. If I’m cooking and I realize the rice has been on for about 7 minutes so it should finish up at 9:15, then that’s how I’m thinking about it, not doing the math to figure out what the specific number of minutes is between now and 9:15. That’s the goddamned robot’s job.
I think if you realize youve been cooking rice for about 7 minutes you will definitely think in terms of time LEFT and not at what time o’clock it should be ready. “Oh it’s been cooking for about 7 minutes then it needs another 8”
Ok google, make it stop
Well they are different, so why would it set one if you didn’t specify what do you mean exactly
Ok I’ve tossed this comment around in my head a few times, and I can’t fathom why you bothered to make it. What the fuck is the difference between an alarm that goes off at 9:15 and a timer that goes off at 9:15?
The setup.
Timer counts down time and can be paused; an alarm goes off at particular time and can only be snoozed after it goes off. Alarms take into account timezones and time changes, timers are absolute and independent of “clock” time
Yeah in theory but not if I tell it when to set the alarm off. It’s just useless pedantry. Like your virtual assistant is a redditor or something
The enshitification of Assistant is what prompted me, a few months ago, to embark on a quest to remove Google (and other cloud-based services) from my home automation setup. I’ve since swapped over to Home Assistant using Zigbee for almost everything.
I had to keep the Alexa integration going, or the other half would lose their god damned mind because apparently, that’s the only way on the entire planet to turn the light by the couch on and off.
But yeah, next up is just replacing all the light switches with zigbee-enabled ones so I can go full scary motion detection in a room thing. It’s going to be super futuristic in here, like 1998!
Home Assistant actually has its own voice control now. You can even set it up to trigger on the wake word “Alexa”
Enabling it completely breaks Reminders, btw. Or at least it has so far for me. It won’t let me set a reminder and previously set reminders don’t send a push notification. Other assistant things like timers and playing songs seems to hand off to Assistant correctly.
thanks, that’s 90% what I use assistant for
History always comes back at some point
except the AI has even less features
Also, how does it work? Is it even as good?
It’s not, basic functionality has been lost, and instead of completing your requests, it initiates a search/conversation.
A few commands did work and it routed it through Assistant, but after failing my daily tasks I promptly uninstalled.
Don’t you have to install the Gemini app for all the features now? The app that specifically states it’s “experimental”
When you asked the answer was yes, but, then after a day or two they actually force enabled WITHOUT the app being installed. At least for me and a few others online.
Have you found it better than it used to be? I guess I’m not surprised, but Gemini also does not seem super well tested
Classic Google!
Google is using an order of magnitude more electricity to basically copy Siri. Neat.
Nah, this is Google’s version of ChatGPT with speech to text and text to speech. Unlike Siri, it can hold a real conversation, the problem is that it’s worse as an actual assistant for the moment.
In particular, it can’t control media devices.
Google have really enshittified their services, to the point where I’ve seriously thought about using Bing as my search engine instead.
I already switched
I use duck duck go. IIRC, it’s based on bing. It’s better.
I also use perplexity though. It’s very useful.
In my use case, my privacy is already compromised from using mainstream tech, also cached pages are kinda a big deal.
I struggle to find very obvious things when using ddg, unfortunately
Feels badCheck out Ecosia. It’s the same thing, but your searches plant trees in a responsible manner.
I have. Duck duck is more reliable.
Will try again tho. I like trees.
Kagi. Not free, but the best.
Kagi is tempting, and for how much I use search I may want to genuinely use it.
It was always AI, what are you talking about?
Yeah but it wasn’t AI (the buzzword)
I think that’s the funniest part. Like, as far as I know, the regular Assistant uses the same approach to handling data that buzzword AI things use, a neural network. But branding (and potentially internal company politics) is weird, so they decided to kneecap Assistant in order to make Gemini look better on release.
Generative AI is different, it can generate better responses that aren’t just programmed canned responses. However I fully agree with you, they’re trying their hardest to do a bullshit rebrand with it. They could have just swapped it out with assistant and I would have been ecstatic. By rebranding it I get that same bad taste in my mouth whenever marketing elbows themselves into the conversation.
So… Literally the same voice input, but now with more time overhead waiting for responses so that it can be a bit more human sounding?
Yes this bothered be as well.
And only for the US too.
