Comment by SecretDreams
15 hours ago
The Google voice chat integration in android auto is vile currently. You ask it to map something. The map spins up and adds the address quickly. Shortly thereafter, the Gemini agent asks you if you want it to plot the directions? And then it starts bothering you with extra questions like "want me to tell you the weather for your destination or figure out some fun activities?". No, leave me alone and do your job better, please.
It then gets snippy when you tell it to f** off. I'm not likely to use it until they fix that.
Gemini is the worst bot. It's reprehensibly bad.
I was getting more-consistent, useful responses from the then-new voice-operated modes in Android 2.0, ~16 years ago.
Even the paid version, which is included with my Workspace account, is awful. Every time I've tried to use it for something it has lied to me and then immediately followed that lie with a bullshit follow-up question.
Using it makes me feel angry. I don't like feeling that way.
I use Gemini pro at work and home for some LLM tasks and it's fine. Decent for coding.
But the voice implementation of it and integration into Android and Android auto is such a half assed attempt that is being slammed down my throat. It's offensive. And I agree, it's a regression compared to the very old AA auto assistant for basic tasks.
I haven't used it for coding. I don't get very many Gemini Pro prompts with my [very pedestrian] google workspace account. Most of those prompts are spent dealing with the the lies, and after we get that straight there isn't much left for anything resembling actual work.
As a bonus feature: Because it's a workspace account, it forgets everything about what went wrong last time so every session is like a scene from the film Groundhog Day. It is apparently impossible to change this behavior even though I am the administrator.
Overall, I find its utility to be negative -- I'm worse with it than I am without it.
Which is remarkable, I think: I've been using ChatGPT since pretty early on in the demo days and paying for a Plus account for about as long as anyone ever could, and my opinion of that is generally positive. I've accomplished some fairly neat stuff with OpenAI's offerings that I wouldn't have been able to do on my own.
So I'm not generally averse to LLMs. I'm just averse to using Gemini to do anything more complex than turning on a light bulb. :)
I'm glad to hear that you're able to find some utility with it.