Yes, obviously AI is emitting way too much. It shouldn’t even be producing 0.2% of global emissions, let alone 2%. My main grievance is that no one ever talks about improving industrial and agricultural processes even though they produce around 29% of emissions and 20% of emissions respectively.
How is CO2 equivalent measured based on altitude and since methane will eventually degrade to CO2?
This seems really problematic for comparing different types of GHG emissions and gauging what type has the most contribution to global warming.
Lots and lots of math and analysis.
My understanding is it is fairly well settled on a chemical & lifespan basis. I am not sure of what impact initial altitude has.
https://en.wikipedia.org/wiki/Global_warming_potential
So it’s an educated guess that has lots of flaws just like I said. GHG emissions comparisons between output methods is ridiculous. Because again, water is a GHG emission.
So your alternative is what? Just say a tonne is a tonne?
It’s adequate for the purpose at hand.
Not group up all emissions and compare them for the sake of clicks. Look at the original post, where agriculture and livestock are being set as significantly worse than AI data center emissions in 2021. These two aren’t really to be compared, they’re both problems and need solutions but it’s not a pie chart, you don’t have to pick a slice and not the other.
It’s not adequate, it is a tool for shit clickbait articles, memes, and discourse when the actual problem is GHG production slowed/stopped as efficiently as possible. Banning private jets for instance is fairly easy to do, it has no massive drawbacks other than the rich people being upset. Not building more datacenters for a while, not really difficult, adding infrastructure is the hard part not pausing more. As opposed to this worthless “NOOOO AGRICULTURE IS WORSE THAN AI BECAUSE MY ONE GRAPH WITH MISLEADING DATA”
Cost-to-benefit analysis, sure. But you still need a realistic comparison of the costs side of the equation to do that.
People were whining about the energy costs of regular data centers long before AI came along.
That invites a lot of questions like is it lower carbon to have a zoom call than fly out for a meeting? Do the travel emissions of an imported tomato offset the heating emissions for a local out-of-season hothouse tomato? If I’m going to make one personal sacrifice, is it more effective to give up red meat, bike to work, or make my next holiday less far away?
Intentionally ignoring evidence is just dumb; decisions made purely on vibes are often going to be wrong.