Oh that’s why I get my weather from yr.no. Nice table, hour by hour for next few days and ranges for the later ones. I also mostly visit from the desktop website so this is what it looks like to me:
(I’ve even learned a few words in Norwegian from it since I’ve never bothered to switch it to English)
The one in the screenshot, I’m assuming, loops through preset ranges with 4-5 PM being a special one because that’s around where people start going home from work, and just gives you a string based on the weather conditions. So morning till 4, 4-5, then 5 and onwards.
Now if you actually wanted to do it with an LLM (not that you should, it’s pretty wasteful if you consider that it’s pretty much a prompt per every person looking at the weather, maybe you can cache for an hour per location, but there are millions of locations), you’d give it the dataset for the entire day and, being an LLM, it’ll at least make the text flow much better than it is on the screenshot. An on-device tiny LLM as an opt-in feature would perhaps make sense here, but as you go smaller, the quality goes down too, so again not much point IMO.
Idk, this is the kind of thing an LLM would actually be good at. I think this is much simpler automation at work.
But it’s like… a couple of numbers vs more text. I reckon looking at the numbers is a lot quicker lol
Oh that’s why I get my weather from yr.no. Nice table, hour by hour for next few days and ranges for the later ones. I also mostly visit from the desktop website so this is what it looks like to me:
(I’ve even learned a few words in Norwegian from it since I’ve never bothered to switch it to English)
The one in the screenshot, I’m assuming, loops through preset ranges with 4-5 PM being a special one because that’s around where people start going home from work, and just gives you a string based on the weather conditions. So morning till 4, 4-5, then 5 and onwards.
Now if you actually wanted to do it with an LLM (not that you should, it’s pretty wasteful if you consider that it’s pretty much a prompt per every person looking at the weather, maybe you can cache for an hour per location, but there are millions of locations), you’d give it the dataset for the entire day and, being an LLM, it’ll at least make the text flow much better than it is on the screenshot. An on-device tiny LLM as an opt-in feature would perhaps make sense here, but as you go smaller, the quality goes down too, so again not much point IMO.
Not sure why you’re being down voted… Pretty sure you’re right…