well established industry itself seeks efficiency. The emissions are generally from energy use, Energy costs money, they try to find more efficient ways. Stable industries will have already done a lot of innovation to improve efficiency and may even have hit diminishing returns to research once their processes are well understood. Pick any industrial process and you’ll find loads of papers on efficiency at all steps. But at the end of the day if you need to smelt iron ore into high quality pig iron, you’re going to need a lot of heat for a period of time. Best way to reduce industrial GHG emissions is probably to buy less stuff or maybe buy better quality stuff that lasts longer. Not many consumers wan’t to do that though.

Agriculture is weird because we’ve pushed yields up very high with all the fertilizers and monocultures and so on, but i’d think its similar, diminishing returns - and maybe yields have actually been pushed higher than they should for long run soil health, so you might have a viscious cycle of fertilizer development. You could maybe try to shift people to have less meat and more crops, or maybe stuff like seaweed or algae based food, or try to stop them overeating. I feel like the food industry does get a bit of stick for obesity - not that that seems to do much.

Problem with AI is the bubble that means the focus of effort is unlikely to be efficiency; so long as investors are dumb, don’t know what they’re buying and/or speculation oriented then the bang for buck investment (in the short term) is to generate hype. They’d gladly burn energy, to generate more hype to, borrow more, to buy more energy , to generate more hype, to borrow more . . . all the while they’re ‘crowding out’ boring established investments in well understood processes.