I have been toying at home with llama2 and mixtral, why?
To poke around at it and get some general experience of it and it's a bit of fun and to somehow rationalize my 4090 card to myself... Beware that running the larger models will require a "fair" (read insane by home standard) amount of gpu memory to be bearable to run.
My toying around have cemented the idea I had before, <sarcasm> that the "AI" craze is a plot by Nvidia and big climate change to sell more gpus and waste more electricity on pointless energy hungry compute. </sarcasm> Sure there are probably some cases where it can be a really good technology, but for 99% of what it's used for today it's pointless.