I have had a talking computer since 2004.
I've had a computer that listens to me since 2012 when I build my first neural network that could recognize my voice and convert it to bash commands reliably enough.
I've had an LLM powered version of that since the gpt 3 api came up and a local version with a llama 8b model today.
It's meh.
What I find ironic is that even though I'm dictating this to a $50k work station I'm still using Emacs to do it.
But for my grandmother, who keeps a little notebook with detailed instructions for how to message people on Facebook? For her this might greatly increase her capabilities.
We have had voice controlled machines since at least the 80s. There were voice controlled robots in MIT decades ago. It's just stupid and awful.
I wrote about it the other day, but every Windows computer since 2000 has included a fully functional, fairly well implemented, and mostly accurate voice control. Nobody uses it except people who have to use it, because voice is a terrible control plane. Humans are WAY better at choosing an option from a 2D grid using a mouse, even people who are afraid of computers. Voice is inherently slower, less accurate, less precise, and inherently ambiguous.
Hell, even as shown on Star Trek TNG, controlling a computer through your voice is awful. They show a few times characters having to do searches through a bunch of data and refine the output by adding more filter criteria, and always what they do is better done by a good excel spreadsheet.
People don't want voice controlled computers FFS, stop trying to force them. Stop trying to make fetch happen.