Comment on Lemmy be like
Ek-Hou-Van-Braai@piefed.social 3 weeks agoI can run a small LLM locally which I can talk to to turn certain lights on and off, set reminders for me, play music etc.
But we could do vocal assistants well before LLMs (look at siri) and without setting everything on fire.
And seriously, I asked for something that’s worth all the down side and you bring up clippy 2.0 ???
Where are the MANY exemples ? why are LLMs/genAI company burning money ? where are the companies making use of of the suposedly many uses ?
I genuily want to understand.
You asked for one example, I gave you one.
It's not just voice, I can ask it complex questions and it can understand context and put on lights or close blinds based on that context.
I find it very useful with no real drawbacks
I ask for an example making up for the downside everyone as to pay.
so, no ! A better shutter puller or a maybe marginally better vocal assitant is not gonna cut it. And again that’s stuff siri and domotic tools where able to do since 2014 at a minimum.
Siri has privacy issues, and only works when connected to the internet.
What are the downsides of me running my own local LLM? I've named many benefits privacy being one of them.
The fact that was the best you could come up with is far more damning than not even having one.
I'm keeping it simple, that's a solid good use case, and what millions of people use ChatGPT for.
I can run a small LLM locally which I can talk to using voice to turn certain lights on and off, set reminders for me, play music etc.
Neat trick, but it’s not worth the headache of set up when you can do all that by getting off your chair and pushing buttons. Hell, you don’t even have to get off your chair! A cellphone can do all that already, and you don’t even need voice commands to do it.
Are you able to give any actual examples of a good use of an LLM?
Like it or not, that is an actual example.
I can lay in my bed and turn off the lights without touching my phone, or turn on certain muisic without touching my phone.
I could ask if I remembered to lock the front door etc.
But okay, I'll play your game, let's pretend that doesn't count.
I can use my local AI to draft documents or emails speeding up the process a lot.
Or I can used it to translate.
If you want to live your life like that, go for that’s your choice. But I don’t think those applications are worth the cost of running an LLM. To be honest I find it frivolous.
I’m not against LLMs as a concept, but the way they get shoved into everything without thought and without an “AI” free option is absurd. There are good reasons why people have a knee-jerk anti-AI reaction, even if they can’t articulate it themselves.
It's not expensive for me to run a local LLM, I just use the hardware I'm already using for gaming.
Electricity is cheap and most people with a gaming PC probably use more electricity than they would running their own Ai and asking it some questions.
I'm also against shoving AI in evening, and not making it Opt-In. I'm also worried about privacy and concentration of power etc.
But just outright saying LLMs are bad is rediculous.
And saying there is no good reason to use them is rediculous. Can we stop doing that.
PeriodicallyPedantic@lemmy.ca 3 weeks ago
That’s like saying “asbestos has some good uses, so we should just give every household a big pile of it without any training or PPE”
It doesn’t matter that it has some good uses and that later we went “oops, maybe let’s only give it to experts to use”. The harm has already been done by eager supporters, intentional or not.
Honytawk@lemmy.zip 3 weeks ago
No that is completely not what they are saying. Stop arguing strawmen.
PeriodicallyPedantic@lemmy.ca 3 weeks ago
It’s not a strawman, it’s hyperbole.
There are serious known harms and we suspect that there are more.
There are known ethical issues, and there may be more.
There are few known benefits, but we suspect that there are more.
Do we just knowingly subject untrained people to harm just to see if there are a few more positive usecases, and to make shareholders a bit more money?
How does their argument differ from that?