[elbasti]

I have an android phone which means I use android auto fairly often. The sheer quality destruction it's experienced since transitioning to Gemini is incredible.

I experience this mostly when asking for music. Before gemini, mistakes were common but deterministic. It was easy to understand where the query had gone wrong and so how to fix it. Example:

"Hey google, play Blackstar"

(Plays the album blackstar by David Bowie, not what I wanted)

"Hey google, play "Blackstar by Radiohead"

(Plays the right thing).

Now:

"Hey Google, play Blackstar by Radiohead" can result in playing... something vaguely semantically related with no way to course correct. In this exact instance (happened yesterday!) it played an album by the hip hop due Black Star.

I will admit that there are some superpowers hidden in Gemini that were not present in the previous AI assistant. I recently discovered that Gemini can manipulate the navigation app, and a prompt like "Mute alerts" works, which is kind of cool. However like OP said, it's incredibly verbose, which is super annoying.

[dhruvmittal]

Before:

     "Hey Google, call my wife"

                           ------> Immediately calls wife

With Gemini:

     "Hey Google, call my wife"

                           --50%-> "Is that XXXXXX?"

                           --25%-> Immediately calls wife

                           --20%-> "I'm sorry, I couldn't find a contact saved as 'wife'" 

                           --05%-> Immediately calls someone different from my wife

[kellengreen]

This is so accurate.

[bsimpson]

There was an Amazon UXR study that floated around ~8 years ago that said the only things people care about from a voice assistant are music, weather, and alarms.

PMs keep trying to make them "smarter," and it just makes the core user journeys worse.

Surely they think they're inventing cars when we're griping about buggy whips. But it really feels like voice assistants peaked ~10y ago for the things people actually want them for.

[Xeoncross]

Considering the decade of poor results from doing anything other than music, weather and alarms I think we've all learned to avoid using them for anything else.

[esafak]

Don't forget recipes. They went and bungled up the only good things about it. Every day I suggest Google fire the whole division responsible for the product when it cheerfully says "I'm still learning" and asks me for feedback.

[wiseowise]

> There was an Amazon UXR study that floated around ~8 years ago that said the only things people care about from a voice assistant are music, weather, and alarms.

That’s all cool, but once we implement those how can we keep making money and justifying our salaries?

[RajT88]

For a bit, our Amazon Echo's were pretty decent at doing things other than music, weather and alarms. Then there was a big regression, and we stopped trying to make it do more. I guess the exception is turning the lights on and off.

My wife and I both would love for the voice assistants to do more. They just won't. Even with weather, anything more than "what's the weather like today" will usually not get a good result. "When will it rain today?" gets OK responses.

As soon as I figure out how to put a decent GPU into my old rack server(s), I'll see how far I can get with HomeAssistant. I suspect it'll be some effort, but it'll be better at the end of the day.

[cfiggers]

I have a crystal clear memory of the day I stopped trying to make voice assistants work for me.

I use a timer every day to brew my coffee. With a voice assistant I can set a timer, but with the lack of a screen I can't see how much time is left. One day I thought, "I'm going to finally get around to digging into this voice interface and see what the options are," hoping for something like, "Hey Dingus, set a five minute timer and notify me when there's 10 seconds left."

Or better yet, "Hey Dingus, five minute timer, with notify at 10 seconds."

Notice that this almost maps 1:1 onto a shell command with option flags, just verbally interfaced: "$ timer 5m --with-notify 10s"

Notice also a complete willingness on my part to learn how the thing works and change how I'm using it accordingly. This is the opposite of end user laziness. I'm willing to invest effort in becoming a "Power User" of my voice assistant.

So I looked for documentation, ready to read and use my brain to understand it and do what it tells me in order to start and stop my timers with greater proficiency.

...I found none.

Literally, there isn't any. They don't have documentation. Nowhere is there, even for someone motivated and willing to learn, the ability to do so.

Ok, well that is understandable if these things are changing rapidly. Maybe there's the equivalent to a "$ timer --help" baked into the assistant itself. Maybe it can tell me what options exist so I can use them. I ask the assistant to explain itself. "What are my options when setting a timer?"

It can't parse my question. Literally, it doesn't know what I'm asking for. Because nobody ever considered that a user might ask that question or even want to know that information.

At that point, on the spot, I gave up. Clearly this thing is not designed or intended to be a thing that one could gain skill with. It's an utterly unserious product.

I would very happily learn an entire verbal DSL, a whole pidgin dialect of English, purely for interacting with my voice assistant. "Hey Dingus, five minute staged timer: thirty seconds, two minutes, one minute, one minute, remainder, with countdown from five seconds" is not "natural language" anymore. But you can bet you'd hear me saying it, if that's all it took to make the voice assistant run my coffee brewing recipe with nothing but my voice. And then, hey bonus, let me bind that to a personal shortcut so all I need to say is "Hey Dingus, coffee timer" and I don't even need to reach for my phone.

But you can't do that. It literally does not support being taken seriously. Or, if it can, the design is utterly hostile to me discovering how. So I've never even tried to do anything, since that very day, other than turning smart lights on and off.

My point is: I didn't fail the technology. I came to the table with willingness to tinker and experiment, willingness to change my expectations to suit the design as I discovered it, willingness to work to make it a part of my routine. The technology failed me.

[econ]

You are now technology enough to finally be ready for a mechanical cooking timer. The really old ones are even made from metal.

My preference goes to steampunk level wear and a sticky grease shell made from 50 years of cooking.

I got an electronic one once, it's only a few years old and it keeps eating batteries as if some subscription plan. "Progress" here already offers one more thing to do of the when I just woke up type.

[RajT88]

I don't disagree with what you said, but as a workaround, why don't you just set your timers for 4 minutes and 50 seconds?

[cfiggers]

Hey, that's clever, actually. I didn't do that because I didn't think of it!

[hyperhello]

What you’re getting at is a personal basic language parser that input-output-maps to a safe subset of operations. It can’t delete your hard drive, but it has a “set a timer X, Y, Z” to some wait function execution. That would quickly be adopted and no one would ever upgrade in their life, so how is it profitable?

[jimbokun]

And knowing how old Geena Davis is.

[fhdkweig]

I don't really use Google products except Youtube, but their auto-subtitler (closed captions) have really taken a weird turn lately. It used to try sounding out the words so even if it didn't pick the right word, it was at least close to right one. But now it substitutes entire phrases and sentences based on what is most likely to appear in a conversation even if it has no bearing or similarities to what was said.

[themythfable]

Same thing with Google homes.

Pre-gemini, you knew what you would get, basically the structured snippets that would appear at the top of the search results.

Now it's much more verbose.

My biggest gripe is that it basically stopped listening to me, since "upgrading" to Gemini, which is frustrating because I've used it to control the Hue lights for the past decade.

It listens to my partner though, so after it fails to listen to me, I have to ask her to ask Google to adjust the fing lights.

Welp

[jimbokun]

Like being the one who brought the dog home from the shelter and feeds it, and the dog still prefers your partner.

[ack210]

Since android auto transitioned to Gemini, whenever I try to reply to a message by tapping the reply button, it now states "Sorry, but to do that, you'll need permission from your Google Workspace administrator." The stupid part is, if I listen to the message first, then say yes when it asks me if I'd like to reply, I'm then allowed to reply. There's countless articles and posts about this but no one seems to have found a reliable solution yet. Has been going on since at least December.

It also now seems to trigger its own barge-in about 50% of the time. It'll start reading the first syllable of a message, apparently confuse itself talking for me saying something, then just follow that with silence "listening" for my response until I physically have to hit the back button on the car.

[CountHackulus]

I feel like the song suggestions also vary depending on the music service. For example the other day, "Play cicada by Igor" got me Cicadidae by Igorrr on YT Music (this is what I wanted), but got me Igor's Theme by Tyler the Creator. ???? So I tried "Play cicadidae by I-G-O-R-R-R" and I got Cicada by Good Kid.

That said I feel like no matter the music service it's a 50/50 shot each time if it plays the song I want. No matter how many times before I've played it or asked for it.

The latest Gemini update for Android Auto also absolutely ruined the voice control for Waze. It was already bad, but now Android Auto is basically unusable by voice.

[Peanuts99]

Even before the move to Gemini they've been on a downward spiral, 'skills' Google home used to have now don't work and it feels like they've been actively making it worse. I've not switched to Gemini yet but now I'm not too hopeful.

[NautilusWave]

Verbally getting Android Auto to play the album Black Star by Amaarae is impossible.

[tommodev]

Yeah, I only ever use voice for music or navigation with android auto.

Navigation is OK but now for music, everytime I ask it to play a song, it warns me it can't do that on Spotify Free tier (I have a subscription?).

Then it goes ahead and plays the wrong song anyway.

[daveshistory]

Perhaps Gemini's been training on sources too young to listen to Radiohead.

[elAhmo]

Similarly, Siri was arguably better a decade ago when it has deterministic intents that you could trigger with certainty, now it is all seemingly random at time.

[RobRivera]

I wonder how much of it is them trying to minmax their revenue generating marketing algos for content discovery. (Willing to bet it's somewhat relative)

[treis]

Pretty sure it just sucks. For years I've listened to Pablo Honey by Radiohead while going to bed. About a month ago it started to refuse to do that and instead start a radio from the first song on the album

[jimbokun]

Soon it will be like Jack Black in High Fidelity, preventing you from listening to music that’s currently too hip for you.

[projektfu]

We are solidly in the Nutri-Matic phase of AI deployment. I fear we may never leave.