Whether the hardware to run it is on-prem at a large company, on local machines or in data centres, it is still a big increase in the demand for that hardware, and that costs money. Your argument doesn’t work, because it relates two unrelated things:
how energy intensive AI is compared to other activities
how hardware intensive AI is compared to current hardware usage
It should be clear that one can be high while the other is low.
The rest of your comment became a highly passionate rant instead of a reply to what I said about energy usage, which I will chalk up as yet more evidence that
people hate it with an irrational passion. So they criticise all aspects of it, whether reasonably or not.
Whether the hardware to run it is on-prem at a large company, on local machines or in data centres, it is still a big increase in the demand for that hardware, and that costs money. Your argument doesn’t work, because it relates two unrelated things:
It should be clear that one can be high while the other is low.
The rest of your comment became a highly passionate rant instead of a reply to what I said about energy usage, which I will chalk up as yet more evidence that