At least 70% of things I read about people doing with "AI" assistants like "Open Claw" are basically things people have been doing with deterministic, secure tools like Home Assistant for years.

[hagen@mastodon.social]

@tante that’s one part i also fundamentally not get. even if you want to use llms (which i argue you shouldn’t) – just ask them to write the home assistant yaml files or something similar. it’s by no means great but at least you only have to get it working once and then it won’t change. why roll the dice over and over again?

[fh0@mastodon.social]

@hagen the hope is to have a general personal assistant that simply does all the stuff you tell them, without first figuring out what other software you need to run it and a good place to integrate something somewhere

[hagen@mastodon.social]

@fh0 sure but that’s not even how actual personal assistants work.

[fh0@mastodon.social]

@hagen I don't think that matters, it's the reason why people use it instead of integrating it in homeassistant (your original question)

[sebastianlaube@layer8.space]

@tante The interesting thing about your observation is convinience.

You don't have to buy a device that you have to assemble, setup with absolutely unfamiliar software and configure all the things your self.

Now they set up one or two user accounts and will be guided quite comfortably through the setup.

This reveals the weakness FOSS always had: the lack of user experience.

I don't like these agents, but I wish we get motivated by this to do better UX in FOSS.

[skjeggtroll@mastodon.online]

@tante

We used to just call it "automation."

[research_ftw@sciences.social]

@tante "How can we do this thing we could do already with way more compute so we can prop up the failing fossil fuel industry"

[elrohir@mastodon.gal]

@tante The sad truth is that what drives adoption of a technology is not its technical design but social dynamics. A lot of these people are embracing AI not because they could not automate default email templates ten years ago, but because ten years ago they couldn't be arsed to find out if it was even possible, whereas nowadays they all of a sudden feel pressured to do it because everyone else is doing it. And what bothers me the most is they will then claim that it was impossible 10 years ago

[elrohir@mastodon.gal]

@tante Someone oughta make an AI agent coffee pot that you can text "BREW" when you want it to make you a cup

https://en.wikipedia.org/wiki/Hyper_Text_Coffee_Pot_Control_Protocol

[yatil@yatil.social]

@tante But it is so much more exciting if it only works 50% of the time and there is a 1% chance that it erases your hard drive!

So much entertainment value!

[downey@floss.social]

@dalias @tante

I think a better term may be "latent human desire for slavery"

[greg@icosahedron.website]

@tante I WAS THERE WHEN IFTTT WAS NEW! I WAS THERE!!!

[dalias@hachyderm.io]

@downey @tante Yes, though I'm not sure what "latent" implies here. I don't think it's something intrinsic - most of us find the idea repulsive, and this is probably part of our visceral reaction to "AI" "assistants". Rather, I think people who've been indoctrinated into hierarchical power structures where they observe their bosses "getting to" behave like that develop it as an aspiration. And for the bosses, the appeal is that the thing they're giving orders to like it's a person can't say no to them and can't be secretly trying to achieve goals at odds with them(*).

(*) LMAO that's exactly what all the "AI" chatbots are doing when they scoop up all your trade secrets for Google and Microsoft to exploit.