Estimates for how much an AI image costs in water varies widely, estimates from 0.4 liters to 50 liters with the median I’m seeing at about a 1-5 liters. Even taking the 50 liter estimate which we can assume includes training and electricity generation to get a number that high that is far below a serving of rice which costs 276 liters.
Except agriculture puts food on the table, and ai data centers just spits out soulless art and tells people to kill themselves. I would not consider them to be comparable.
We produce food in such a vast excess compared to what we need that we feed most of it to animals so we can have beef instead despite being vastly less efficient.
And I support degrowth. I do not see how this helps you. I never said we don’t have an excess, I just said we need agriculture and we dont need ai data centers.
Yeah, a process required to make food and a process to make an image that could have easily been done by an artist or simply pulled from one of the countless sites full of existing free images, are totally comparable.
How much energy would have been taken in the time it takes you to draw a poster on a computer?
I would actually be interested in the comparison, how long is a typical model trained for and how much is it used. The actual generation usage is typically a few seconds of a normal gaming PC. Training is more, but its then used by many.
Yet again discounting the training costs…
Estimates for how much an AI image costs in water varies widely, estimates from 0.4 liters to 50 liters with the median I’m seeing at about a 1-5 liters. Even taking the 50 liter estimate which we can assume includes training and electricity generation to get a number that high that is far below a serving of rice which costs 276 liters.
Malthus would be proud of you…
the training costs are real
Even with the training costs, it is very little water compared to mass scale agriculture.
Except agriculture puts food on the table, and ai data centers just spits out soulless art and tells people to kill themselves. I would not consider them to be comparable.
We produce food in such a vast excess compared to what we need that we feed most of it to animals so we can have beef instead despite being vastly less efficient.
And I support degrowth. I do not see how this helps you. I never said we don’t have an excess, I just said we need agriculture and we dont need ai data centers.
Yeah, a process required to make food and a process to make an image that could have easily been done by an artist or simply pulled from one of the countless sites full of existing free images, are totally comparable.
How much energy would have been taken in the time it takes you to draw a poster on a computer?
I would actually be interested in the comparison, how long is a typical model trained for and how much is it used. The actual generation usage is typically a few seconds of a normal gaming PC. Training is more, but its then used by many.