[jshen]

Co2 levels are the best metric for progress, and progress would be a steady lowering of the rate of increase. We are not seeing that, and AI's power hunger is likely to make it worse despite all of the positive things you mentioned.

[Retric]

CO2 levels tell you nothing about future progress, the legacy of soon to be removed infrastructure built 40+ years ago still impacts it. Meanwhile the vast majority of power brought online in the last decade is ultra low carbon. We know how that story plays out.

AI powered by renewables has ~zero impact on the climate.

[jshen]

I hope you're right, but I'm not optimistic. The demands for power by AI is immediate and extremely large. We can't bring nuclear online fast enough so it will likely keep co2 output high.

Assuming you are right, when would you expect the rate of atmospheric co2 increases to start to decline?

[Retric]

I expect the global peak annual CO2 emissions from fossil fuels and industry to occur sometime in the next 5 years largely dependent on economic activity. I’d give it something like 90% odds.

CO2 from land use (deforestation etc) has already dropped by half since the 1960’s, but I know less about that so I’m unsure of the details. That said it’s well under 10% of total emissions so probably not a major factor for now. https://ourworldindata.org/grapher/co2-land-use?mapSelect=~C...